Development of test system for soft robotic gloves

(1)

IN

DEGREE PROJECT MECHANICAL ENGINEERING, SECOND CYCLE, 30 CREDITS

STOCKHOLM SWEDEN 2018,

Development of test system for soft robotic gloves

SVEN-RUBEN BÖCKER SIMON MALMSTRÖM

KTH ROYAL INSTITUTE OF TECHNOLOGY

SCHOOL OF INDUSTRIAL ENGINEERING AND MANAGEMENT

(2)

Master of Science Thesis

Production Engineering and Managament Course: MG212X

Supervisor: Andreas Archenti

Swedish title: Utveckling av testsystem för mjuka robothandskar TRITA ITM-EX 2018:558

Development of test system for soft robotic gloves

SVEN-RUBEN BÖCKER

SIMON MALMSTRÖM

(3)

2

Sammanfattning

Bioservo Technologies tillverkar mjuka robothandskar. En växande produktion har gjort att företaget har uttryckt ett behov av en inspektionslösning som ger kvantitativa, objektiva och tillförlitliga data om handskarnas förmåga. Den första uppgiften bestod av att bestämma vilken eller vilka parametrar som ska mätas och hur mätningar ska gå till. Huvuduppgiften var att designa och tillverka ett testsystem för Bioservos produkter som skulle kunna användas i deras nuvarande produktion. Två parametrar, fingerstyrka och fingersensorkalibrering, har tillsammans med en testmetod utvecklats från tillgänglig forskningslitteratur då inga standardiserade test finns för denna typ av produkt. En prototyp av testriggen har skapats, men den är inte lämpad för produktionen på grund av valda sensor- och datainsamlingskomponenter.

Resultaten från testmätningar visar däremot på att systemet har potentiell nytta för företagets kvalitetsarbete. Rekommenderade komponenter och framtida arbete har beskrivits för att underlätta för Bioservo om de väljer att gå vidare med utvecklingen. Varför de dyrare rekommenderade produkterna inte användes diskuteras också eftersom det kan vara relevant för andra företag och organisationer som är intresserade av att göra tillförlitliga mätningar för första gången.

(4)

3

Abstract

Bioservo Technologies is a producer of soft robotic gloves. Growing production volumes has made the company express the need for a product inspection solution that provides quantitative, objective, and reliable data regarding the gloves’ capabilities. The initial task consisted in determining which performance metric or metrics should be measured, as well as how measurements should be obtained. The main task was to design and manufacture a testing system for Bioservo’s products that could be implemented in their current production. Two metrics, finger strength and finger sensor calibration, along with a test method have been derived from available research literature as no standardised tests exist for this type of product.

A protoype testing rig was created, but the sensor and data acquisition components used do not make it suitable for application. However, the results from test measurements do show that the system may have potential benefits for the company’s quality work. Recommended components and future work have been described, should Bioservo wish to proceed with development of the system. It is also discussed why the more expensive recommended products were rejected, which may be relevant for other companies and organisations that are interested in performing reliable measurements for the first time.

(5)

4

Acknowledgements

First of all, we would like to thank our university supervisor, Dr Andreas Archenti, for his inspiring enthusiasm and help with keeping the project within the field of production engineering.

This project would not have been possible without the knowledge and support from Anton Kviberg at the Department of Production Engineering at KTH. His experience with manufacturing was of tremendous help during the manufacturing of our prototype.

We would also like to thank everyone at Bioservo Technologies in Kista who contributed to this project, especially our supervisor at the company, Martin Ascard. His input regarding everything from adapting the test method for Bioservo to the design of the prototype has been invaluable.

(6)

5

1 Introduction

This section explains what the background and goal for the thesis project were. It will also list its starting requirements and limitations.

1.1 Background

Bioservo Technologies is a Swedish company with around 30 employees that develops and manufactures soft robotic gloves. Founded through research collaborations in medicine and robotics, the company has created a product called the Carbonhand, seen in figure 1, for people with weakened grip force or impaired hand functionality. They can also be used to prevent work-related injuries. Because of a growing market interest in the professional sector, the company has developed a new glove system aimed at industrial applications [1]. This new product is called the Ironhand. Compared to the Carbonhand model, it is more powerful and features five fingers instead of three. Both are now part of their product range.

Figure 1. Bioservo’s Carbonhand [2].

Assembly of products and components is currently done at the company office in Stockholm, Sweden. With increasing production volumes and a high market potential, Bioservo now wish to improve their inspection process to ensure quality for their customers. When the glove systems are assembled, only a subjective operator test is performed. The company has expressed the need for a measuring system that will test the products and provide quantitative, objective, and reliable data regarding their capabilities. This can also aid in further development of their products and production.

(9)

8 1.2 Problem definition

The first aim of this thesis is to decide what type of glove characteristic should be measured and describe how to do it. This is necessary to reach the main goal of creating a physical testing system that can be implemented in Bioservo’s current production.

1.3 Initial requirements and demarcations

A discussion was held at the start of the project together with staff from different areas at Bioservo regarding which characteristics and attributes to evaluate. The purpose was to clarify what desires and expectations the company had, as this would narrow down the literature search.

It was also important to discuss the relevance of each request to production engineering, as it is the scientific field of this project.

The main interest was in characterising the glove by measuring the load exerted by it on fingers.

Measuring the total load on a finger was set as a requirement, while measuring the load on each individual joint was considered highly desirable. Regarding the loads, it was decided that measuring static loads would suffice. Dynamic load measurements, i.e. measuring how the loads change during travel or time, were also discussed but not considered crucial. This criterion could be altered if a more suited performance metric was discovered in the literature review.

Another requirement was to be able to attain data from each finger of an assembled glove, as they are individually controlled units. Because the thumb adds a different level of complexity to the system, it was decided that it would not need to be incorporated. However, it was still deemed desirable.

Evaluating the output of the force-sensing resistors in the gloves was decided to be a secondary objective, only to be included if it presented little added complexity.

Data gathered from a measuring system should also be possible to store in the company’s own database. For practical reasons, it was decided that the system would only have to provide data in a format suitable for transfer to the database, e.g. text files.

Any software used or programming made should be simple enough for someone with basic engineering education to understand and be able to develop the system further.

No monetary budget was set. Instead, any products or materials needed would have to be approved by the company before purchase.

Because of the scope of the project and the final aim of creating an operational measuring system, the time available for an initial literature review would need to be limited. This meant a higher risk of missing valuable information at the start of the project, but it was a decision discussed and made together with both the company and university supervisors.

1.4 Disposition of report

The report is divided into six sections. The introduction presents the background, goal, and limitations of the project. In the literature review, key concepts are explained, possible benefits of product inspection mentioned, and the question of what to measure is answered. Next, the method section details how the theoretical framework was used and adapted, and how the

(10)

9

physical system was designed and manufactured. The results section describes a product inspection method that Bioservo could use and shows the manufactured prototype. Difficulties and obstacles encountered during the project are then brought up in the discussion section, along with an evaluation of the system and how it can be used. Recommendations for future improvements to the system and potential thesis work are listed in the final section.

(11)

10

2 Literature review

This part of the report briefly presents the scientific field of the thesis, key terminology and theory, as well as potentially useful practices. It also describes research that the project is based on.

2.1 Metrology - the science of measurements

As explained by the problem definition, this thesis project revolves around measuring some of the gloves’ parameters. Therefore, a condensed description of the field of metrology and important concepts is appropriate.

Metrology is defined as the science of measurement by the International Bureau of Weights and Measures (BIPM) [3]. Measurement, in turn, is the “process of experimentally obtaining one or more quantity values that can reasonably be attributed to a quantity” by way of comparison, where a quantity is a property with a numerable magnitude [4, p.16]. Metrology can be divided into three subfields: scientific, industrial, and legal.

Scientific metrology involves defining and reliably realising measurement standards such as the meter, as well as creating and upholding chains of traceability [5, 6]. This type of work is done by international organisations such as BIPM. Traceability connects the different fields and is important in all of them, as it relates to the reliability of measurements. It is the most relevant part of scientific metrology for this thesis and will be explained separately later in this section.

Industrial metrology, also called technical metrology, deals with measurements and instruments used in manufacturing or other production processes, such as product testing to ensure that manufactured parts are compliant with design tolerances. Focus is on the quality of the measurements based on the known uncertainty of a measuring system, but also on the ability to prove the traceability of instruments to international standards [7, 8]. This thesis project is in the subfield of industrial metrology.

Legal metrology is concerned with measurements where there are legal requirements on measurements or instruments related to finance, safety, health and environment. Examples of applications are weighing scales used in trade, where the aim is to ensure unbiased economic transactions, or instruments used in the medical field, where an incorrectly measured dose can negatively impact a patient’s health [9]. The International Organization of Legal Metrology (OIML) is an intergovernmental organisation that creates regulations and standards, publishes recommendations for use in industry and society, as well as control a certification system used for different types of products [10]. This thesis project is not as concerned with legal metrology as with industrial or scientific, but it is appropriate to mention it since OIML certifications can be used to classify many products. One of the products used in the project is an example of this.

Traceability is the link between an end user’s measurement and the International System of Units (SI). Because it would not be practically or financially feasible to have every measuring system in the world calibrated at the BIPM headquarters, national measurement institutes (NMIs) exist who provide calibration services in their respective country. These are far more accessible than BIPM, but at the cost of a higher uncertainty. Calibration laboratories reference their equipment to NMIs and are more accessible than them, although with increased

(12)

11

uncertainty. Part of the calibration service is often to acquire the proper documentation of an unbroken chain to the base units. Where organisations should calibrate their measuring systems depends on the tolerated level of uncertainty. Hospitals might need to calibrate some equipment at an NMI, whereas a manufacturing company may deem it adequate to calibrate against a purchased working standard (e.g. a gauge block or calibration weight) [5, 6, 11]. Although not always an explicit requirement, knowing the traceability chain of the measuring equipment means that the measurements can be trusted within the specified uncertainty levels.

Different actors in the traceability chain are presented in figure 2 with increasing measurement uncertainty from top to bottom.

Figure 2. Traceability pyramid [5, 6].

2.2 Why product inspection and measurements are useful

To find out how a measuring system could prove useful for Bioservo, literature on quality management was reviewed. This is because measurements are integral components of any quality management system, and product inspection is discussed heavily in the field too. Also, the authors of this thesis both have prior knowledge of quality management theory, making it easier to utilise the literature in this project considering the time available. However, this does not imply that we urge Bioservo to devote themselves fully to a quality program. Rather, we want to highlight some of the principles that are common in industry.

As was described in the background section of the thesis, product inspection is used to verify that products meet specified requirements to ensure quality and keep defective products from being shipped to customers. Apart from deciding on which parameters and metrics to measure, the potential monetary benefits should also be evaluated in the planning phase of any quality program by performing a cost-benefit analysis based on quality costs [12]. No such analysis was made available by the company, and it was not done as part of this thesis due to limited

Standardisation of units BIPM

National primary standards NMIs, e.g. RISE Research Institutes of

Sweden

Reference standards Calibration laboratories

Working standards Industry, university

Measurement equipment End user

(13)

12

information about their production. Bioservo’s specific request for a physical test system, however, indicates that the company believes there are benefits. The costs involved will still be discussed in a general approach as the purpose of the measuring system is to reduce some of these.

Quality costs are described by Feigenbaum [13, pp.109-119] as a way for businesses to compare different quality strategies and tools in quantitative financial terms. These are either cost of control or cost of failure of control (called cost of conformance and cost of nonconformance, respectively, by the Project Management Institute [12, p.238]). The former is comprised of prevention and appraisal costs, and the latter internal and external failure costs. These are structured with examples in table 1 below.

Costs of control Costs of failure of control

Prevention

Keep defects from occurring in the first place by building a quality product or production system

• Quality engineering

• Process control

• Employee training

• Equipment development

• Documentation

Internal failure

Failures found before delivery

• Scrap

• Rework

Appraisal

Assessing the quality of products or production

• Testing and inspection

• Destructive testing

• Measuring equipment maintenance and calibration

• Quality audits

External failure

Failures found after delivery

• Warranty

• Liability

• Lost business

• Product recall

• Product service

Table 1. Quality costs summarised [12, p.283, 13, pp.111-119].

Based on the problem definition in the beginning of report, the project is aimed at reducing external failure costs at the expense of added appraisal cost. It is generally more effective to focus on preventing problems rather than just detecting them [12, p.274], which is why Bioservo should use the potential measuring system for more than characterisation. Some

(14)

13

quality engineering tools that Bioservo could implement using measurements will be mentioned in the discussion section of the report.

A noteworthy theory is the Taguchi view on quality costs. It differs from the more binary perspective on quality as a product being either conforming or nonconforming to tolerances by introducing the quality loss function. In short, losses increase as the variation between target value and the actual value of a product characteristic increases, even within tolerances [14]. It is well beyond the scope of this thesis to calculate and describe a specific quality loss function for the products in question, but it is the authors’ opinion that the principle may be of use for Bioservo’s production, depending on what parameter is to be measured.

2.3 Finding a test method for soft robotics gloves

Test methods for products are often described and regulated by standardisation institutes, so the first step was to see if an industry standard for this type of product already existed. Exoskeletons for use in manufacturing is a relatively new technology and not widely available, and no relevant standards were found. The same applies to soft exoskeletons. However, as stated by Bostelman, et al. [15], while standardised test methods for many wearable robotic systems do not yet exist, these could be derived from test methods designed for conventional robots.

Because the article is recent and two of the authors are employees at the National Institute of Standards and Technology (NIST), it was deemed appropriate to search for performance tests designed for manufacturing robots. Note that NIST is currently part of ongoing work to create performance standards for exoskeletons [16], but nothing applicable has been published yet.

Bioservo are strongly recommended to follow this work as it progresses.

A similar type of product to Bioservo’s gloves is a robotic hand. As with exoskeletons, no applicable benchmark standards currently exist. It is because of this that NIST is leading a project on evaluating relevant performance metrics as well as designing benchmark tests [17].

In a journal article about the project [18], the authors explain that they believe a robotic hand’s grasping capability is of interest for both manufacturers and customers and that any measuring systems used to compare robot hands should be independent from their respective control systems. Of the different performance metrics suggested in the report, the one best suited for this thesis work is finger strength in terms of force. It is described as the most basic metric that will indicate grasp and hand strength, as these are ultimately composed of the capabilities of the fingers. This performance metric complies well with Bioservo’s request of evaluating the load of each finger and was selected as a theoretical framework for the thesis largely because of NIST’s function and reputation as a measurement and standardisation institute.

Another test described by NIST that may be used in this thesis is force calibration, which aims to evaluate the capabilities of force-based tactile sensors on the robot fingers. The setup is similar to the finger strength benchmark and it is said that these two tests be can combined if desired. In short, readings from the finger sensors are compared to the readings of an external load cell [18]. It does require more access to the glove control system and better knowledge of the how the glove’s sensors are used than is available for this project, but a combination of these two tests would fulfil Bioservo’s request for a sensor-evaluation feature.

(15)

14

2.3.1 NIST finger strength test for robotic hand

Finger force is affected not only by actuator capabilities, but also motion controllers and the configuration of the finger and the object that a force is exerted on. This makes it a good performance metric as it is possible to compare assembled robot hands that appear mechanically similar but yield different force readings [18].

To make a test that is easily replicable for most models of robotic hands, the finger should be evaluated near full extension. A force sensor (load cell in practice) is placed just beneath the fingertip, illustrated in figure 3, and the finger is commanded to close completely using full power, as this setting can be performed on any robotic finger. It is recommended that the sensor can measure force in three directions to measure the total force exerted [18]. Because of the long moment arm, this configuration will show the lowest maximal force the finger the finger can exert (a more bent finger would yield greater force readings) [19].

Figure 3. Finger-object configuration for 3-axis force sensor [18, p.130].

If a 3-axis force sensor is not available, a single-axis sensor will suffice. The finger-object configuration will need to be altered, though, to increase the force in the measured direction and reduce contact forces in the other two dimensions. The finger should be bent slightly as seen in figure 4, but no definite angle has been decided [19].

Figure 4. Finger-object configuration for single-axis force sensor [19, p.2].

To protect both the finger and sensor as well as facilitate testing, a rigid column can be attached to the sensor, as shown in figure 5.

(16)

15

Figure 5. Rigid column between robot fingers and load cell [19, p.3].

The quasistatic force region, marked in figure 6, should be extracted from the collected data and averaged to get the final force value. Several cycles should be made and the mean and 95%

confidence interval values calculated [18].

Figure 6. Example of force readings from NIST test [18, p.130].

(17)

16

3 Method

This section will describe how the information from the literature review has been used to create a test system, motivate design choices, and describe the manufacturing and assembly of a proof- of-concept prototype.

3.1 Adapting the NIST finger strength test to suit Bioservo’s needs

The test method developed by NIST can be altered to be better suited for Bioservo’s products and current production. The orientation of the rigid column attached to the load sensor in NIST’s finger strength test does not appear to be very efficient for an inspection process as the sensor would have to be moved several times during each glove test. The sensor could be moved using a servo or stepper motor which would necessitate more electronics and a control system, but also increase the margin of error due to moving parts. A better solution for Bioservo would be to have a system where each finger can load the sensor without changing the setup at all. If the fingertip forces affect the same plane in the same direction, a single-axis load cell should be able to provide accurate readings even if the contact points differ, given that the plane does not deflect or bend. A fixed sensor position should be preferable, as many different glove sizes will be tested. The downside is that the readings would only show the force components in the measured direction, and not the total force exerted by the finger.

Bioservo expressed a greater interest in integrating an evaluation of the glove sensors in the test method after the literature review was done. As the fingers are to be close to full flexion, this did not appear to be too complex and would be considered in the design of components.

3.2 Establishing system requirements

Together with employees at Bioservo Technology, some specific requirements of the test system and testing rig were established:

• The testing rig needs to fit the glove in a correct manner. Failing to do so might affect the glove in a way that it would no longer perform as intended. This will also help in finding possible manufacturing errors related to the fit and shape of the glove.

• The testing rig to be manufactured as part of this project will only need to fit one size, but it should work for both glove models. As discussed in section 1.3 Initial requirements and demarcation, the thumb would not need to be included.

• The testing rig needs to be small and portable. It should be possible to be set up and removed fast and with little difficulty, as available space may be a concern.

• The mechanical system does not need to be actively driven, but it does need to be able to reset the fingers to their starting positions. This requirement is essential since the gloves do not have any incorporated mechanism to perform the rebound of a finger to a flexed position (this is done by the wearer).

• Ease of use for operator is an important criterion, as the test system should be intuitive and simple enough for most people to use. It should also not prevent the operator from performing other important tasks.

(18)

17

• Ease of development is directed more towards the engineers at Bioservo. Anyone with an engineering background should be able to understand how the system works with little effort, making it possible for other thesis workers to continue the project. This is also a practical requirement for this project, as the time available is limited and a lot of it must be allocated to the design and manufacturing phases.

• Changes to the system’s software and monitoring of tests should be possible to do with a laptop via a USB connection, since many computers today only feature this port for serial communication.

The last three requirements are denoted as plug and play-criteria. They are important to the feasibility of making a practical test system that produces reliable data within the time plan of the project.

3.3 System modules

With the requirements set, the measuring system was broken down to its main functions. Three modules could then be created: sensor and data acquisition, hand, and fixture.

Sensor and data acquisition includes all functions regarding the conversion of a load to a digital quantity value that can be stored and analysed.

The hand module components will act as a skeleton or base, ensuring that the desired motions are obtained when actuated by one of Bioservo’s gloves. As the glove is not able to reset to its starting position on its own, the mechanical hand must perform this function. Reactionary forces normally taken up by the wearer’s wrist also need to be supported. Because the glove wrist dimensions are different for each glove size, this last function was sorted under the hand module as opposed to the fixture module.

The fixture constrains the whole system, acting as a base for mounting of the other modules. It will try to limit the influence of factors other than the desired ones on the force readings. For example, noticeable deflection due to reactionary forces in the system is to be avoided as this may affect the repeatability.

Table 2 presents the modules and functions discussed.

(19)

18

Module Function

Sensor and data acquisition

Translate load to a digital signal

Convert signal to quantity values

Present values and transfer them to database

Hand

Ensure correct motion and fit

Reset mechanical system to starting point

Support the glove’s reactionary forces

Fixture Rigid base for mounting of components

Table 2. Modularisation of the measuring system.

3.4 Demarcations

With a performance metric and test method decided, along with the systems functions and requirements outlined, some practical demarcations needed to be set for the next phase. A portion of the project’s time had already been allocated to answering the questions regarding what to measure and how to measure it. The time remaining now acted as the main limiting factor for the design and construction phase, as the end goal of the project was to build a proof- of-concept prototype. This meant that products and software that were intuitive to use or required little time to learn would be preferred. While some solutions might have been omitted due to this, the negative impact was compensated to some degree because it also meant that continued development of the measuring system would be easier.

Recalling the introduction and literature review of the thesis, neither a clear monetary budget nor a cost-benefit analysis for the project existed. Cost of manufacturing or purchasing of components would need to be discussed with the company, leading to a risk of losing time on creating concepts that could be deemed too expensive.

3.4.1 Manufacturing methods

A review of manufacturing methods that could be considered was made. An important criterion when selecting methods was availability. Availability is based primarily on lead time, as components would have to be designed, manufactured, evaluated, redesigned, and so on, within the project time plan. Manufacturing at Bioservo or the university workshop at KTH IIP (Department of Production Engineering) was considered preferable to using third-party workshops because of shorter lead times but also to reduce the risk and impact of parts being

(20)

19

manufactured incorrectly or arriving late. Methods that the thesis authors or employees at the company had previous experience with were prioritised. For example, Bioservo frequently use additive manufacturing (AM) of plastics in their product development and production.

Nonetheless, the methods chosen should be common enough for Bioservo to have no trouble finding third-party manufacturers able to provide the parts designed. An example of this is CNC milling which is available as both 3-axis and 5-axis at KTH IIP. While more complex geometries can be machined with a 5-axis mill, the machines are more expensive and not as readily available as 3-axis machines. Therefore, it would be better to design for 3-axis machining if possible. Price was also considered in the analysis for the sake of future production, as it is likely that only low quantities of the measuring system and its components will be manufactured. As such, methods suited for large production volumes were excluded. Table 3 below summarises the conclusions.

Manufacturing method Availability

Additive manufacturing (polymer)

Fused deposition

modeling (FDM) FDM printer for ABS and PC at KTH IIP Selective laser sintering

(SLS)

SLS printing of various polymers through third-party manufacturer

Subtractive manufacturing

Turning Manual and CNC at KTH IIP

Milling Manual and CNC (3- and 5-axis) at KTH IIP

Waterjet cutting With or without abrasives at KTH IIP

Welding Different welding systems at KTH IIP

Metal forming Simple manual forming tools at KTH IIP

Abrasive machining Grinding and blasting at KTH IIP

Laser cutting Machine for cutting fabrics, wood, and

plastics at Bioservo Technologies in Kista Table 3. Manufacturing methods available.

Standardised components such as fasteners were to be used in the design when possible, as these can often be purchased off-the-shelf at a lower cost than manufacturing an equivalent component in-house.

(21)

20

3.5 Selection of sensor and data acquisition components

As explained by Bray, et al. [20], measuring force can be quite complicated from an operational point of view. Some practical reference standards must be available to ensure traceability of measurements. Weight, elastic, and electromagnetic forces are the types of forces most often compared with in industrial operations. In the SI system, weight force is based on the mass of a body, a constant physical property. It is measured in newtons (N), where one newton is the force accelerating 1 kg by 1 m/s^2. As force is based on the kg, sensors for measuring mass or weight, such as load cells, are adequate for practical force measurements. This comes at a cost, though, since measurements will depend on the acceleration caused by the Earth’s gravitational pull which has significant variations on different locations around the world. Calibration at the region of application negates this to a large extent, highlighting the benefits of traceability, explained previously in the literature review of this thesis.

Load cells appeared to be more practical for the project. Different types are commercially available, such as piezoelectric or hydraulic load cells, but the most prevalent type in terms of products and literature available is the strain-based load cell. In general, this type of load cell balances good accuracy and durability with low cost, making it ideal for this project.

3.5.1 Strain-based load cells and errors

Strain-based load cells almost always measure the change in electrical resistance of one or more elastic materials, commonly strain gauges or less accurate force sensing resistors (FSR). Strain gauges can be metallic foils on a backing material that is attached to an elastic element. When the element deflects, so does the strain gauge, affecting the resistivity of the material [21]. This is illustrated in figure 7.

(22)

21

Figure 7. How a strain gauge works [22].

The change in resistance is measured by configuring one or more strain gauges in a Wheatstone bridge [20], shown in figure 8. For a quarter-bridge configuration, one strain gauge constitutes one of the resistances in the schematic. The other resistances are known, and the voltage measured across the bridge changes as the strain gauge’s resistance increases or decreases. Two strain gauges can be attached to the elastic element and connected to the same bridge, making it a half bridge. Four strain gauges make a full bridge, and this configuration is mostly used as it better compensates for temperature effects.

Figure 8. One or more of the resistances in the Wheatstone bridge can be a strain gauge [23].

(23)

22

Strain gauge-based load cells come in many different of shapes. Common configurations are cantilever beam structures and S-beams, or variations of these. Load cells are compared by certain operational characteristics, or errors [20]:

• Non-linearity

Signals from load cells do not have a linear behavior, meaning the same increase in load will not yield the same output increase for low and high ranges. The linearity error expresses the maximum values the output signal can deviate from a theoretical straight line (based on the least sum of squares for example).

• Sensitivity

The ratio of how the output signal varies to a variation in load, in effect meaning the smallest weight that can change the signal value.

• Hysteris

Also called the reversibility error, the hysteresis error expresses the difference between two signal values for the same load measured with increasing and decreasing load.

• Repeatability

Expresses the spread of load readings for repeated measurements under identical conditions.

• Reproducibility

Similar to the above, reproducibility expresses the spread of repeated load readings under varying conditions.

• Temperature

Load readings can differ greatly under different temperatures, which is why it is often stated in data sheets how the signal is affected in terms of sensitivity per degree.

• Creep

Expresses how the signal changes when under constant load during an extended period of time.

• Drift

Describes permanent signal variations when no load is applied, also called zero drift.

Following either ISO or OIML standards, some these are extracted through calibration of individual cells, as no two physical load cells are identical due to constructional imperfections and manufacturing processes. The combined uncertainty of all errors within the specified capacity range is often presented as the total accuracy [20].

3.5.2 Data acquisition (DAQ)

Data acquisition (DAQ) systems are needed in many engineering applications. The instruments are used to measure some quantity or condition, but also for automation of industrial processes.

Because of the requirements set, this report will focus on PC-based DAQ only. In short, PC- based DAQ instruments translate the signals from a sensor to something that can be processed

(24)

23

by the computer. The benefit of a PC-based systems is that performance comes at a relatively low cost, as normal computers can be used for demanding calculations and processes.

Accompanying software also makes programming and data logging easier for the user.

Sensor output is often an analogue signal in voltage or current. This must be changed to a digital signal before any information can be interpreted, and it is done using an analogue-to-digital converter (ADC). The signal may also need to be altered before the ADC, as it may be in a different range than the rated input of the DAQ instrument (often 0 - 5 V). A common example of this is amplification, where the output signal from the sensor is boosted. Filtering is also used widely to improve the quality of the signal by reducing different types of electrical noise. These alterations fall under the term signal conditioning, and the components may be external or built into the DAQ instrument [24].

Like the sensors described earlier, no DAQ instrument is perfectly accurate. Describing the many errors are not deemed relevant as they are too complicated for the scope of this report, but the combined uncertainty will be an important factor when reviewing instruments, as well as calibration and traceability. More relevant qualities to mention here are resolution and sampling rate.

Resolution is expressed in bits and shows how many discrete steps that can be made over the rated range. In other terms, it is the smallest change in analogue signal value needed to change the digital signal value by one bit [24]. For this thesis project, the resolution needs to be high enough to get a corresponding load resolution of at least about 1/10 N (it will be difficult to distinguish between loads otherwise). Note that high resolution does not inherently equal high accuracy.

Sample rate describes how many samples can be made per second. For the measuring system in this project, it needs to be high enough to be able to distinguish between the dynamic and quasistatic regions of the load signal. Commercial DAQ systems usually range from a few ten thousand to over a million samples per second.

3.5.3 Recommended solution and low-cost alternative

Strain-based load cells with a capacity up to 10 kg were reviewed based on the maximum force of the gloves actuators. It was not known what the actual load readings would be, so the idea was to not go too low or too high in capacity, as the former could result in a damaged load cell whereas the latter might not be accurate enough. For the sake of traceability, only load cells with certifications of calibration where consider in the initial search. As explained in section 3.1, a single-axis load cell could be used if all finger forces acted on the same orthogonal plane.

Single-point load cells were therefore reviewed, since these are often used in platform weighing scales where the contact point should not affect the reading for a specified platform size. A sensor was chosen based on measurement uncertainty, price, and availability (where it could be purchased and delivery time).

DAQ systems that are certified and appeared simple to use were prioritised in the first review.

USB connectivity was also a requirement that narrowed down the search. The most promising systems were provided by National Instruments, NI. Labview was believed to be a good

(25)

24

software solution for the project as it uses graphic programming, is taught at many universities, and instructions are readily available through the company website or handbooks, meeting the requirement of ease of use and enabling further development. It would also be possible to sort and analyse the data and exporting it to a suitable file format for exporting it to a database, and the Labview virtual instrumentation (VI) could also act as operator interface, showing whether a product has passed inspection or not. After contacting the NI customer support and discussing the project and its requirements, a meeting was held at the NI Stockholm office in Kista. The purpose was to conclude the discussions and demonstrate the products that would be best suited for this measuring system.

The selected components for the first concept are listed in table 4.

Product name Type of product

Variohm AL6N-C3-10kg-3B6 [25] Single point load cell

NI 9237 D-sub [26] Input module for strain or load measurements

NI cDAQ-9171 [27] CompactDAQ chassis with USB connection

Labview NXG Software

Table 4. Recommended sensor and DAQ products.

The load cell is rated for 10 kg and has an OIML C3 classification. According to the manufacturer, there are load cells rated for 3 kg and 5 kg in the same product family and with the same dimensions, making it simple to switch sensors if needed. The input module from NI provides simultaneous signal acquisition, amplification, and filtering for up to four load cells with full, half, or quarter bridge configuration, and is connected to a PC via USB through the CompactDAQ chassis. Both products are delivered calibrated. Labview NXG is a version of the software aimed at industrial data acquisition. The original version of Labview would also work, but Labview NXG has a user interface more suited for this type of project.

Although Bioservo agreed to this being a good solution, it was rejected due to the price. If less expensive alternatives were to be used, the requirement of plug-and-play would not be met, and traceability of the system would also be given a much lower priority. It would not be possible to learn the required programming languages for creating a custom DAQ system in the time available, but an Arduino-based solution was found that would only require some minor configurations. The products of this solution are listed in table 5. Note that the purpose of using these products was only to see if a load signal could be acquired from the test system and its approximate magnitude, and not to be used for inspection purposes.

(26)

25

Product name Type of product

SparkFun OpenScale [28] Load cell signal amplifier and digitiser

TAL220 10 kg [29] Parallel beam load cell

Table 5. Low-cost alternative products.

OpenScale, shown on the left in figure 9, is an Arduino-based board made by SparkFun Electronics with an integrated HX711 analogue-to-digital converter (ADC). The product is designed for static measurements as it has a maximum sample rate of 10 samples per second (80 samples per second if the hardware is altered), which is significantly lower than the other systems reviewed. Basic functions and settings are preinstalled. It can be calibrated and run on a PC by connecting it via USB and using any terminal program [30]. Arduino IDE and its serial monitor were used in this project. An alternative was to purchase an Arduino microcontroller and HX11 breakout board, but it would be more expensive and need more time programming than the OpenScale solution. The TAL 220 load cell, shown on the right in figure 9, is used in the applications guide for OpenScale and was chosen for its convenience, as both products could be purchased from the same Swedish online retailer.

Figure 9. OpenScale board [31] (left), and TAL 220 load cell [32] (right).

3.6 Design of hand components

The hand module can be broken down into two submodules: fingers and joints, as well as palm and wrist. How the concepts were generated and designed will be described in this section.

3.6.1 Fingers and joints

The fingers are the most complex components of this submodule, and as a result, several design iterations were performed. Based on available materials and manufacturing methods, three simple initial finger concepts were created, shown in figure 10. More detailed development of the joints was to be started after the finger design was set to avoid unnecessary design iterations.

(27)

26

Figure 10. Finger concepts 1 (top), 2 (middle), and 3 (bottom).

Finger concept 1 emphasises ease of manufacturing and maintenance. The finger consists of series of fork joints coupled by threaded steel bars. The rightmost joint is connected to the knuckle joints and the leftmost end forms the fingertip. This design composes a skeletal base for the finger. For this concept to fit the glove correctly, some type of material would need to be added onto these components.

Finger concept 2 also focuses on ease of manufacturing. The finger consists of waterjet-cut profiles of steel. These profiles are coupled with rivets and forms a chain with three joints.

Finger concept 3 is aimed at creating a robust and rigid design. It consists of waterjet-cut and milled aluminium profiles connected with turned aluminium axles. The axles are locked in place with a steel circlip on each side. The surface seen in the bottom left corner connects to the palm.

(28)

27

Figure 11 below shows how one of the finger concepts could be assembled to a conceptual palm made of bent steel sheet metal. A certain spread between the fingers can also be incorporated into the design with relative ease.

Figure 11. Finger concept 3 assembled to a palm concept.

Discussions were held regarding the concepts together with employees at Bioservo as well as the university supervisor. The main topic was if it would be necessary or even feasible to use fingers with multiple joints. As previously stated, measuring the load exerted on each individual finger joint was desirable but not a strict requirement. After creating the draft models, it became evident that altering the designs for different glove sizes and manufacturing these might be needlessly difficult. This was largely due to the complexity of incorporating mechanisms for rebounding the finger as well as constraining the different phalanges (the bones in the fingers) to each other. Not having joints in the fingers would reduce the flexibility of the system. For example, some of the benchmark tests described by NIST use a grasping configuration.

However, static load measurements had been set as a requirement as opposed to dynamic load measurements, which were only considered desirable. Therefore, it was decided to simplify the fingers to only using a single joint, this being the joint at the knuckles. This would constrain all movements to one degree of freedom, likely making any designs made more robust.

A forth finger concept was made after the above-mentioned evaluation. This finger has place for only one joint. It is designed for additive manufacturing with ABS plastic, making a finger- like shape easy to manufacture. Because it only features one joint and one piece of material for the finger, it should be more rigid than the first concepts. Test fingers, shown in figure 12, were 3D-printed at the university to get an indication of how practical the concept would be. The durability of the parts exceeded the expectations, and they also proved to be rigid enough to be able to be exposed to the forces in the testing rig without any noticeable deformations. Based on the properties of the smallest printed test finger, the concept appeared applicable to all glove sizes. Together with employees at Bioservo, it was decided that this concept should be further developed to a final design.

(29)

28

Figure 12. Printed test fingers of different sizes.

The final finger design is illustrated in figure 13. It features a construction made of solid 3D- printed ABS-plastic. The shape and thickness of the finger was iterated several times to fit the glove correctly. Having a slightly bent finger proved to be highly advantageous in reducing the risk of the gloves’ tendons pulling a finger the wrong way. The slight bend seen in the figure will make sure that the glove tendon is always pulling the finger downwards. The fingertip consists of a flat cutout that will provide a surface for a soft rubber tape to mimic the texture of a human fingertip. Having a soft fingertip should provide the correct circumstances for the FSR sensors to work in a manner suitable for repeated testing.

Figure 13. Final finger design.

A feature to limit the travel of the finger was incorporated into the joint. This consists of a simple radial slot, seen in figure 14. Also shown is the bearing seat, which consists of deformable ridges that will hold the bearing fixed. The bearing fit should be tight enough to hold the bearing in place, but loose enough for it to be easily removed.

(30)

29

Figure 14. Travel stop and bearing seat.

As previously mentioned in section 3.2, there is a requirement for the mechanical system to perform the rebound stroke of the fingers. This requirement will be met with the use of rubber bands. Figure 15 shows the mounting point for a rubber band on the finger.

Figure 15. Mounting point for rebounding rubber band.

The joint forms the mounting point between the palm and the finger. It is responsible for constraining the fingers movement to one degree of freedom. Figure 16 illustrates an overview of the joint assembly design and the components. The golden part is made of 7075 aluminium which can be initially waterjet-cut in the profile shown in the rightmost view. It is given this colour in the design to help distinguish between parts, but no coating or painting is necessary.

After that, all holes are to be drilled using a milling machine to attain high surface finish and

(31)

30

precise positioning. The axle also consists of 7075 aluminium, that is to be machined in a lathe.

It is locked in place with a steel circlip on each side. The figure also shows the spacers that will lock the finger bearings axially.

Figure 16. Joint assembly overview.

The finger and joint assembly is illustrated below in figures 17 and 18. The chosen bearings are deep groove ball bearings with ZZ-sealings [33].

Figure 17. Finger and joint (exploded view).

Figure 18. Finger and joint assembly (top view).

3.6.2 Palm and wrist

The palm can be seen as the base of the hand module, since this part links the fingers, joints, and wrist to the fixture and all forces induced in the system will pass through this component.

Figure 19 illustrates the final design of the palm. The philosophy behind this design is for it to be as strong and simple as possible, as well as being easy to modify for different sizes and

(32)

31

accessories. It will be manufactured from a waterjet-cut piece of steel sheet metal that is bent into shape.

Figure 19. Palm overview.

The wrist component, seen in figure 20, is the main anchoring point for the glove. All forces generated in the tendons will create reaction forces that will load the wrist. Therefore, it is important for this component to be stiff to reduce deformations that might affect the anchoring of the gloves’ wrist straps. Seen from above, as in the rightmost view of the figure, the wrist has a conical profile to be able to anchor gloves of different sizes. The relatively complex geometry of the wrist requires additive manufacturing to keep the manufacturing simple. It will therefore be manufactured in 3D-printed ABS plastic.

Figure 20. Wrist overview.

The assembled hand module components are illustrated in figures 21 and 22. The fingers are fastened to the palm using one M5 screw per finger joint.

(33)

32

Figure 21. Palm and wrist (exploded view).

Figure 22. Hand assembly overview.

3.7 Design of fixture and final assembly

The aim for the fixture design, illustrated in figure 23, was to make it very rigid, since any deformations might affect the accuracy of the scale. It also needs to be relatively small and light to make it easy to move when it is not used. Figure 24 presents an exploded view of the fixture, as well as a list of all components included in this module. The fixture is designed to work with a single load cell, but also with two parallel load cells (of a different type) that can be connected by a beam.

(34)

33

Figure 23. Fixture design.

Figure 24. Fixture (exploded view) and list of components.

The base plate is designed to be cut from 10 mm thick 7075 aluminium using a waterjet cutter.

It needs to be large enough to house both the hand brackets and the scale assembly.

The hand brackets need to be very stiff. Therefore, they will consist of two waterjet-cut profiles of 10 mm 7075 aluminium.

The scale assembly consists of multiple components, each designed to fit the specifications according to the datasheet from the load cell shown in figure 25. Four small aluminium

(35)

34

cylinders are used to provide a stiff connection between the load cell and base plate, as well between the load cell and push plate.

Figure 25. Load cell dimensions from datasheet [34].

The final assembly of all components is depicted below in figure 26.

Figure 26. Final assembly overview.

(36)

35 3.8 Manufacturing

Due to the limited amount of resources, both monetary and in the sense of manufacturing possibilities, design for manufacturing has been a central philosophy throughout this project.

The aim has been to limit the costs by doing the manufacturing of as many components as possible at KTH IIP. All available manufacturing methods have previously been listed in table 3 in section 3.2.1.

Waterjet cutting is an extremely effective manufacturing method in terms of time and cost.

Therefore, most of the metal components in the fixture were designed to be manufactured using this method. All these components would likely require some post processing after the waterjet cutting since the available machine can only to cut along two axes. Shapes and countersunk holes that require three or more machining axes needed to be post processed in a milling or drilling machine. For parts that requires high tolerances, a milling machine was chosen since it is more accurate compared to a drilling machine.

The palm was sandblasted after being cut and bent. This was to remove any sharp edges that could damage the gloves. The corners on the finger side were also welded for the same reason.

Simple manual abrasion could be used instead, and welding of the corners can be replaced by some other design feature.

All parts could not be waterjet cut. These were the retaining pins for the joints as well as the spacers in the scale assembly. This is mainly because of the strict tolerances set, a requirement that a waterjet cutter cannot meet. A turning machine was used to manufacture these parts instead.

The fingers and wrist were printed at the university workshop machine, which uses the FDM method. The initial plan was to print these parts at one of Bioservo’s suppliers using an SLS machine because they are superior at creating smaller details. As the FDM-printed prototype fingers were deemed durable and detailed enough, it was ultimately decided that printing take place at the university to save time.

3.9 Equipment test

To get an idea of how the system could behave in an inspection situation, some tests of the equipment were performed.

After mounting the scale assembly on the fixture, the load cell was connected to the screw terminals on the OpenScale board. This was then connected to a PC running Arduino IDE. The load cell was calibrated and tared to zero value according to the instructions in the OpenScale application guide [30] using a weight. Note that this was not a working standard weight. Sample rate was set to 5 samples per second.

First, a reference test of the scale was done using the same weight as before. The weight was tied to a string, held slightly above the scale, and then placed on the platform for 5 seconds. It was then removed before being placed on the scale for another 5 seconds. Five repetitions were made.

(37)

36

Next, the hand was assembled to the fixture after which a Carbonhand glove (size medium) was mounted as seen in figure 27. After connecting the glove unit controller to the PC and starting its control software, basic functions were checked. Glove fit was good, the mechanical fingers moved when actuated by the glove and rebound when not, and a signal could be read from the OpenScale system.

Figure 27. Overview of test setup.

Test 1 consisted of five force pulses of varying duration induced on the scale by the actuated fingers. The driving torque of the glove fingers was set to maximum working level through the control software. The length of each pulse was increased from 10 seconds to 30 seconds in 5 second increments. All pulses were separated by a 5 second pause were no load was exerted on the scale. This test was performed on both the middle and ring finger with similar battery levels.

The purpose of this test was to see how the load would change over time and if a dynamic load range, as described in the NIST finger strength test, would be visible. This information could also be of help in deciding load duration and sequence.

Test 2 consisted of three series of four force pulses. Each pulse duration was 10 seconds, separated by a 5 second pause. Between each series, the glove was fully removed and then mounted again. The purpose of this test was to see how mounting of the glove would affect the results and to get an indication of the measuring system’s repeatability and reproducibility.

(38)

37

4 Results

Answers to the questions stated in the problem definition, the prototype measuring system, as well as system test results are presented in this section.

4.1 Description of test method

Finger force is the main metric measured in the test method intended for use at Bioservo. It will only be measured in one axis, meaning that it will not show the total force the glove exerts on each finger. Tolerance values can be set for each finger and glove size to compare the measured value with. This makes an objective characterisation of each glove possible.

Another metric that can be evaluated is how load readings from the glove sensors compare with the load readings of the measuring system. If both measuring planes are parallel, the readings should not differ much as both readings are of the same force component. The function of each individual sensor is tested before they are assembled in the glove, but this test can indicate if they have been mounted incorrectly. Also, it provides added redundancy in the inspection program in the case that the equipment used to measure the sensors is malfunctioning. It can eliminate the need for prior sensor inspection too, as the test can flag sensors that are clearly not meeting the supplier’s specifications.

Testing of a fully assembled glove begins by mounting it onto a mechanical hand, giving an indication of the fit and if something has been sewn incorrectly. The glove is connected to a PC that controls the actuators. The glove will move one of the fingers at full power, making it exert a load on the scale platform. When no force is provided, the rebounding mechanism will return the finger to its normal state. Each finger is tested, although the thumb has not been incorporated as of now. No specific load sequence or duration has been determined as part of this thesis project. The data acquired from the test is evaluated using the same PC running the control program and the operator is told if the glove conforms to tolerances or not. Readings are indexed for each finger and transferred to the company database. Ideally, this should all be done in the same program.

4.2 Prototype measuring system

Figure 28 presents the results of the design and manufacturing of the testing rig with its modules (OpenScale board not shown).

(39)

38

Figure 28. Testing rig prototype.

4.3 Results of equipment test

Figure 29 shows the resulting data from the reference test mentioned in section 3.9. The graphs from each test repetition is plotted in the same datapoint interval to distinguish the fluctuations during static load with a known weight. The start and end points of each graph are not of great interest in this test since these parts are where the scale is loaded and unloaded, respectively.

As shown by the graphs, the fluctuations vary by around ± 0,002 kg.

Figure 29. Reference test results.

The captured data from test 1, where the objective was to see how the load changes over time, is shown in figure 30. As can be seen in the graphs, the load decreases significantly once the servo has reached the specified load but then starts to even out.

(40)

39

Figure 30. Finger force from middle and ring finger with increasing pulse intervals.

Figures 31 and 32 show results from test 2 for the middle finger and the ring finger, respectively.

Each figure has three graphs, illustrating the three repetitions of the test. The data shows that there is a difference in load between each repetition.

Figure 31. Finger force from middle finger with glove removed between each test.

(41)

40

Figure 32. Finger force from ring finger with glove removed between each test.

(42)

41

5 Discussion

This section will discuss the results of the thesis project, such as the prototype and its components. The project in general will also be discussed as the experience gained and realisations made throughout it may aid future thesis workers.

5.1 Test method evaluation

As mentioned earlier, no standardised test methods or similar were found for soft exoskeletal gloves. The methods that were used as a base are not yet standardised either, meaning that they are likely to be improved and updated. This may cause the test method derived as part of this project to become obsolete or need drastic changes. However, it is the belief of the authors that there is strong enough theoretical backing to motivate the use of the test method as a framework for Bioservo. Again, the company is strongly encouraged to follow the progress of both NIST projects described in the report, especially the project on exoskeletons.

Using a single-axis load cell means that a reference value is logged, as opposed to the total load exerted by the finger. This is not a problem for product inspection, given that some load value can be set as reference for all gloves. It does mean that the gloves will not be characterised by their actual capacity, which is desirable for product development. Adapting the test rig for a 3- axis load cell should not be difficult, though, and the station could then be used for other tests.

The recommended DAQ products would also allow for this type of change.

From a practical perspective, the test method does appear to work according to the specifications outlined. One concern raised after the equipment test was regarding the decreasing finger force. If the glove cannot maintain a constant load on the scale, it can be difficult to isolate a quasistatic force region. This may be part of the gloves safety or interface features, in which case it should be possible to override it during testing. Otherwise it might be better to gather data from a predetermined time interval. Also, the spikes visible in figure 31 should be investigated, as these are likely due to the glove’s control system.

5.2 Uncertainty of data

Even though the manufactured prototype is fully operational, the uncertainty of the output data should be questioned. Part of the reason that the low-cost DAQ and sensor solution was inexpensive is that none of the components are individually calibrated to any international or national standards. As explained in the theoretical framework of this thesis, traceability and calibration are needed for measurements to be trustworthy from a metrological point of view.

This thesis is in the field of production engineering and industrial metrology, and the low-cost alternative cannot be recommended in its current state.

The test system could be sent to a calibration laboratory or calibrated in house with respect to the operational characteristics listed in section 3.5.1, but the cost of this in terms of money and time would likely match, if not exceed, that of the recommended system. The latter was chosen because it also met requirements such as plug and play, something an extensive calibration process of the low-cost system would still not include.

The benefit of using the cheaper components was that a proof of concept could be made with little financial impact. It was also possible to test the benchmark method described in section

(43)

42

4.1. To manage the time more efficiently, it might have been better to limit the scope of the project to just designing a simple testing prototype rather than trying to create a system that could be used in product inspection to gather reliable data right away. This is mainly due to the monetary issues encountered, which will be discussed next, but a stricter limitation would have meant more time for the literature review and made it possible to begin on the foundations of a custom low-cost data acquisition solution.

5.3 Lack of monetary budget

Throughout the first steps of the project, little was discussed regarding the monetary budget.

Initial indications gave the impression that equipment cost would not be an issue. As the work progressed, however, it became evident that cost was a major limiting factor.

The main goal of the thesis project was to develop a measuring system suitable for the company’s products and production. The system would need to adhere to some standards and principles of metrology to provide good reliable data. Products with certifications of calibration are priced above the type of components that were ultimately used in this project, but it was believed that this would not be considered a problem. A lot of time was therefore lost in creating a concept that could meet the set requirement. It also meant that some of the more important requirements would have to be omitted and the project became more oriented towards product development and prototyping than production engineering. However, based on the holistic view of the project as the development of a test method and corresponding measuring system for a production application, it is still in the field of production engineering.

More initiative should have been taken from both parties in the beginning of the project. With a known monetary limit, it might have been identified early on that a physical measuring system for product inspection was not possible. Instead, more effort could have been directed towards making a cost-benefit analysis or on the design phase.

The authors of this thesis report would like to emphasise the importance of establishing a monetary budget and securing finances early in this type of project to future thesis workers. If no budget is set, at the very least one should make sure to get an estimate of how much monetary value is expected to be gained from the project. This was not available during this thesis, resulting in a system that could not meet all set requirements and some lost time.

5.4 Quality tools

As mentioned in the literature review, it is better to prevent defects in production rather than just detecting them. This can be done by improving the design of the products or using other manufacturing techniques. Any quality engineering program will rely on measurements to gauge how the quality has changed. To help Bioservo with future quality work, some quality tools that could be implemented using data from the measuring system will be described. Only basic tools that are relatively simple to use and understand have been reviewed, based on the company’s size and production volume.

Some assumptions are made regarding the final measuring system as an inspection station. All products are tested after being completely assembled. Each finger is tested for both load and

Development of test system for soft robotic gloves

Development of test system for soft robotic gloves

SVEN-RUBEN BÖCKER SIMON MALMSTRÖM

Development of test system for soft robotic gloves

SVEN-RUBEN BÖCKER

SIMON MALMSTRÖM

Sammanfattning

Abstract

Acknowledgements

Contents

1 Introduction

2 Literature review

3 Method

4 Results

5 Discussion