Real Time Measurement of Dirt Pick-up by a Robotic Vacuum Cleaner using Light Sensing Technology

Real Time Measurement of Dirt

Pick-up by a Robotic Vacuum

Cleaner using Light Sensing

Technology

SHELSEA TINA MONTEIRO

K T H R O Y A L I N S T I T U T E O F T E C H N O L O G Y

ii

Real Time Measurement of Dirt Pick-up by a

Robotic Vacuum Cleaner using Light Sensing

Technology

Master of Science in Embedded Systems

By

Shelsea Tina Monteiro

Industrial Supervisor

Niklas Windh

Electrolux AB, Sweden

Examiner

Professor Mark T. Smith

KTH Royal Institute of Technology, Sweden

DEGREE PROJECT IN INFORMATION AND COMMUNICATION TECHNOLOGY

iii

Acknowledgement

Thank you everyone 

This has been a very interesting thesis and I have had tremendous fun working on it. Besides the engineering aspects, it is the people that I interacted with who really made the final phase of my studies so much more interesting.

I would like to thank my academic examiner Mark Smith for being so cool and letting me focus on the actual project rather than the formalities. Thank you for all the advice, guidance and support right from the time of the thesis selection till the end.

I would also like to acknowledge my industrial supervisor Niklas Windh for letting me find my own path and advising and guiding me when required. Your guidance helped me stay on track and in getting the thesis completed as scheduled.

But what is a workplace if you do not have friends you can talk to and have fun with? The most interesting part of doing the thesis at Electrolux was having wonderful colleagues who would always advice, motivate and encourage me and who sometimes even created new puzzles for us thesis students to solve! There should also be a special mention of my fellow thesis workers here at Electrolux for the camaraderie and friendship we shared - after all we were in the same boat together!

iv

Abstract

Domestic chores are one of the most tedious and time consuming tasks in a person’s life. A lot of time can be saved and put to better use if these tasks can be automated. One such chore is the routine task of vacuuming the house every day. Robotic vacuum cleaners that clean the house are thus one of the most widely used domestic robots. These robots have the ability to clean the entire house almost autonomously with little or no human intervention. However, most of these robots do not have a system to report the real-time pick up of dirt which can be useful to the user in knowing which parts of his house are dirty and maybe investigate the reason behind it. This information can be useful to the robot as well in determining efficient cleaning patterns based on the dirt localization in the house.

In this thesis a prototype was developed to measure the real-time pick-up of dirt by a robotic vacuum cleaner. It uses light sensing technology to measure the amount of dirt picked up and can thus be used to glean which parts of the house are dirtier than the others. The signals can also potentially be used to understand the size of the dirt picked up by the robotic vacuum cleaner.

Research was done to investigate the different sensing technologies that can be used and to select the appropriate one. The system was tested and conclusions were made regarding its performance. Additional functions that can be implemented and improvements that can be made have also been suggested as future work.

Keywords

v

Abstrakt

Att städa hemmet är en av de mest tråkiga och tidskrävande uppgifterna i en persons liv. Mycket tid kan sparas och användas bättre om dessa uppgifter kan automatiseras. Robotdammsugare som städar golvet i hemmet är en av de mest använda inhemska robotarna, då dessa robotar har möjlighet att rengöra hela huset nästan autonomt med liten eller ingen mänsklig intervention. De flesta av dessa robotar har dock inte ett system för att rapportera realtidsupphämtning av smuts som kan vara användbart, då användaren kan få reda på vilka delar av huset som är smutsiga och då ha möjlig het att undersöka orsaken bakom. Denna information kan också vara användbart för roboten för att bestämma effektiva rengöringsmönster baserat på lokalisering av smuts i huset.

I denna avhandling utvecklades en prototyp för att mäta upptagning av smuts i realtid av en robotdammsugare. Den använder IR-teknik för att mäta mängden smuts som hämtas upp och kan därmed använda informationen för att avgöra vilka delar av huset som är smutsigare än andra. Signalerna kan också potentiellt användas för att förstå storleken på smuts som tas upp. Forskning av olika tekniker utfördes för att kunna välja den mest lämpliga. Systemet testades därefter och slutsatser gjordes avseende dess prestanda. Ytterligare utredningar och förbättringar som kan genomföras har också föreslagits som framtida arbete.

Nyckelord

vi

Table of Contents

1.4.1 Benefits, Ethics and Sustainability ... 2

1.5 Methodology / Methods ... 3 1.6 Delimitations... 4 1.7 Outline... 4 2 Technical Background ...5 2.1 Problem Background ... 5 2.2 Literature Review... 5 2.2.1 Gravimetric Sampling... 5 2.2.2 Microphone... 6 2.2.3 Strain Gauge ... 6 2.2.4 Dust Sensors ... 6 2.2.5 Light Sensors... 7 3 Design Methodology ...9 3.1 Theory of Operation... 9

3.2 Data Collection and Analysis... 9

4 System Design and Experimental Setup ... 12

4.1 Hardware Implementation ... 12

4.1.1 Light Source Board ...13

4.1.2 Light Detection Board ...13

4.1.3 TIA Board...14

4.1.4 A/D Board...15

4.1.5 Processor ...17

4.1.6 Voltage Regulator Board ...17

4.1.7 Computer ...18

4.2 Software Implementation... 18

4.3 Experimental Setup ... 23

5 System Testing and Results... 25

5.1 System Testing ... 25 5.2 Results... 25 5.2.1 Breadcrumbs...26 5.2.2 Rubber Balls ...26 5.2.3 House Dirt...27 6 Conclusions... 28

6.1 Discussion and Conclusion... 28

6.2 Cost Estimation ... 28

6.3 Future Work ... 28

References ... 30

Appendix A: Camera for Dirt Pick-up Measurement... 32

vii

Appendix D: Results with Rubber Balls ... 37

Appendix E: Results with House Dirt ... 40

Appendix F: Detailed Conclusions ... 43

Appendix G: System Cost... 44

Appendix H: Further Improvements ... 45

viii

List of Figures

Figure 2-1. Simple Light Opacity Measurement Set-up ... 7

Figure 3-1. Photodiode Voltage when no dirt is picked up... 10

Figure 3-2. Signals generated by incoming dirt ... 10

Figure 3-3. Integrating photodiode signals ... 11

Figure 4-1. Hardware Implementation – Block Diagram ...12

Figure 4-2. Schematic - Light Source Board ...13

Figure 4-3. Schematic - Light Detection Board ...14

Figure 4-4. Transimpedance Amplifier Configuration ...14

Figure 4-5. Schematic - TIA Board ...15

Figure 4-6. SPI Protocol ...16

Figure 4-7. Interconnection between the ADC Evaluation Board and the Arduino ... 17

Figure 4-8. Schematic of Voltage Regulator Board ...18

Figure 4-9. SPI Signals from the Arduino ...19

Figure 4-10. Software Flowchart ...21

Figure 4-11. Photodiode Signal Sampling Flowchart ... 22

Figure 4-12. LED and Photodiode Placement in Inlet Valve ... 23

Figure 4-13. Test Setup ... 24

Figure 5-1. Types of Dirt Used ... 26

Figure B-1. Calculated Signal Area vs Dirt Picked-up ... 33

Figure C-1. Photodiode Signals – Breadcrumbs ... 34

Figure C-2. Voltage Drop and Pulse Width - Breadcrumbs ... 34

Figure C-3. Dirt Prediction Model Fit - Breadcrumbs ... 36

Figure C-4. Prediction Error ... 36

Figure D-1. Photodiode Signals – Rubber Balls...37

Figure D-2. Voltage Drop and Pulse Width – Rubber Balls... 38

Figure D-3. Calculated Signal Area for Breadcrumbs and Rubber Balls ... 38

Figure D-4. Predicted Dirt Values for Rubber Balls... 39

Figure E-1. Photodiode Signals – House Dirt ... 40

Figure E-2. Voltage Drop and Pulse Width – House dirt ...41

Figure E-3. Calculated Signal Area for Breadcrumbs and House Dirt ...41

Figure E-4. Predicted Dirt Values for House Dirt ... 42

Figure E-5. Predicted Dirt Value vs Actual Amount of Dirt Picked-up ... 42

Figure I-1. Initial Prototype ... 46

Figure I-2. Final prototype with all the Modules Neatly Arranged...47

List of Tables

Introduction

1

1 Introduction

Robotic vacuum cleaners, now days, are commonly seen in homes all around the world. These vacuum cleaners are robots that perform cleaning tasks by using suction power and brushes. They are capable of cleaning floors, wall corners, and some can also mop the floor. As per a press release by the International Federation of Robotics [1], vacuum and floor cleaning robots will make up the lion’s share of robotic units at work in households. Sales are predicted to rise from 3.6 million units (2015) to around 30 million units within the 2016-2019 forecast periods. Furthermore, as per [2], vacuum and floor cleaning robots account for 96 percent of domestic robot sales.

These robotic vacuum cleaners possess a lot of intelligence which enables them to perform their tasks almost without any human intervention. This relieves the user from performing the daily cleaning chores and thus saves time which can be put to better use. They thus have one of the biggest and most practical impacts on our daily lives.

1.1 Background

The house is a sanctuary for the people and families living in it. The inhabitants like to return to a home that not just looks good but feels good as well. Hence many people now invest in robotic vacuum cleaners that clean the house autonomously. By doing this, they have more time for other tasks and they can concentrate on things that are more important like family and their hobbies.

The primary function of most robotic vacuum cleaners is to travel the entire area of the house according to a certain algorithm and simultaneously pick up dirt and dust that is present on the floor. This is done irrespective of whether certain areas of the house are clean or dirty. Having a sensing system that will enable the acquisition of real time dirt pick-up data would certainly be useful-both to the user as well as the robot. Such a feature will make the user feel empowered and cared for. It will enable the user to know which parts of the home are dirtier and thereby investigate the cause of the same. While the robot could use to this data to determine a much more efficient cleaning pattern. This information would also enable better and more frequent cleaning of dustier areas, faster and efficient clean-ups and touch-up cleaning could be possible. This would also result in longer battery life as well as a longer life-span of the robot.

1.2 Problem

Introduction

2 Thus the question that this thesis is trying to answer is how light sensing technology can be used to accurately measure the real-time intake of dirt by a robotic vacuum cleaner.

1.3 Purpose

In reference to the problem discussed in the previous section, a solution can be suggested and formulated. This solution will be designed, implemented and improved during this thesis. The sensing system which will be developed during the course of this project should fulfill the requirements stated in section 1.2. This system will consist of analog and digital modules as well as embedded software which needs to be well described and documented. Good and detailed documentation is required if the work is to be repeated or if someone wants to improve upon its results. Thus the main purpose of the written material in this thesis is to document the work done in this thesis and make it repeatable.

1.4 Goal

The goal of this degree project is to design a system that is able to accurately measure the pick-up of dirt in real-time. In order to do this, first a literature review will be performed so as to assess the feasibility of the task as well as to research the different sensing technologies and find the best one that can be used for the said purpose.

Based on the results of the research, a sensing system will be built into a robotic vacuum cleaner. A huge part of this thesis will be to design, build and test a prototype that is able to measure dirt pick-up in real time. Thus delivering a working prototype is the main goal of this thesis.

1.4.1 Benefits, Ethics and Sustainability

The beneficiaries of this thesis can be classified into two groups as direct beneficiaries and indirect beneficiaries. The main groups of people who will directly benefit from the results of this work will be the companies that deploy this technology in their vacuum cleaners as well as the end users of these products.

If this technology is deployed in the end product, the robotic vacuum cleaners can utilize the dirt pick-up data to have much faster and efficient clean-up patterns. This can result in power savings which will eventually benefit the user.

Since the vacuum cleaners deploying this technology will be having additional features, it will probably result in increased product sales and thus an increase in the profits of the company.

Introduction

3 In terms of ethics, this project will not be tested on people or animals. The main purpose of this thesis is to build a better robotic vacuum cleaner so as to benefit people and society and make life easier for them. We hope our technology will only be used for good purposes. However, we cannot guarantee the misuse of this technology or how the end user uses the product. From a sustainability point of view, the result of this thesis will enable a bett er cleaning experience. It will enable the user to understand the pollution in his house better and find means of reduces it thereby improving family well-being.

From an environmental point of view, if this system is installed in the vacuum cleaner, the dirt intake measurements can be used to make better cleaning algorithms which will result in efficient cleaning and thus be more energy efficient. The improved and efficient cleaning patterns developed by using the dirt distribution data will help conserve battery life and improve the lifespan of the robot thereby saving energy and reducing the electronic waste.

1.5 Methodology / Methods

Research methods and methodologies are processes or techniques that are important to plan and steer the research work to hav e good, scientific and well proven results. They are very crucial while conducting any research or projects since they directly affect the quality of the research.

As per [3], there are two types of research methodologies: Quantitative and Qualitative research methodologies. Quantitative research methodologies are used when the research or the degree project is about proving a phenomenon by experiments or testing a system with large data sets. The hypothesis has to be measurable with quantifications. Whereas qualitative research methodologies are used when the research is about studying a phenomena or an artifact to create theories, products, and inventions by investigating the environment and using smaller data sets. Another methodology that uses both the aforementioned research methodologies is called Triangulation. Since this method involves using both quantitative and qualitative methods, it gives a complete view of the research area and is often used to improve the credibility and the validity of the results.

This thesis involves researching and experimenting with different setups that can be used to monitor dirt pick-up in real time. Since these setups will involve many tests, changes and comparisons in an experimental manner, the philosophical assumption for this thesis will be “Positivism”.

Again, as stated earlier, this study will involve experiments to build a model that can accurately measure dirt pick-up in real-time. Hence, the type of research method used in this study will be “Experimental”.

Introduction

4 generates best results and delivers a better measuring system will be selected and used. Thus the research approach used in this thesis will be a “Deductive” approach.

1.6 Delimitations

The dirt used to test and evaluate the prototype will not be a mixture of different types of dirt but will be a uniform type of dirt. In addition, very fine dust or DMT test dirt will not be used as it will stick to the surface of the sensors and it would result in incorrect readings. Such fine dirt can be used as long as methods to prevent the dirt from sticking to the surface of the sensors are used.

1.7 Outline

The rest of the report is organized as given below:

Chapter 2 introduces the reader to the technical background of the problem to be solved. It also describes the different techniques that are considered to build the sensing system along with the advantages and disadvantages of each. Chapter 3 describes the theory of operation of the system and how the sensing technique chosen is used to fulfill the goal of this thesis. It describes the techniques used for data collection and the analysis of the same.

The hardware design and implementation along with the software algorithms and flowcharts are explained in Chapter 4. The experimental and test setup is also described in this chapter.

The tests performed to evaluate the performance of the system are explained in Chapter 5. The results obtained are also discussed in this chapter.

Technical Background

5

2 Technical Background

This section describes the research area and the problem to be solved in detail. Prior work and research that has been done in this field as well as a discussion of our approach to solving the problem is also undertaken in this section. 2.1 Problem Background

As mentioned in the earlier sections, robotic vacuum cleaners that vacuum homes autonomously are here to stay. Different manufactures of these appliances constantly strive to improve their performance and efficiency in terms of cleaning patterns and power consumption. Initially, the original vacuum cleaner was just a suction device on wheels that would roam around the house picking up dirt in a random pattern and avoiding obstacles using simple sensors such as infra-red sensors. But now days, the new products on the market have highly complex house cleaning patterns and use lasers to map the cleaning area and avoid obstacles.

As technology evolves and electronics get cheaper, it is possible to add several additional features to these vacuum cleaners. One such interesting feature that can be very useful to the user would be to have a sensing system that is capable of monitoring the dirt picked up by the vacuum cleaner in real time as it cleans the house. This technology would also benefit the robot in several ways. The dirt distribution pattern would enable the robot to implement much more efficient cleaning patterns based on the distribution of dirt. The robot could increase or decrease the suction force based on the intensity of dirt in the respective area. It could then clean such dirtier areas more often and avoid cleaner areas thereby saving battery. This would also increase the lifespan of the robot making it better for the end user as well. Thus this thesis focuses on finding a good solution to make this feature realizable.

2.2 Literature Review

In order to build a system capable of measuring the dirt pick-up in real time, the first step is to investigate the various sensing and measurement technologies that can be used, evaluating them and then selecting one that is best suited for the purpose.

2.2.1 Gravimetric Sampling

Technical Background

6 very fine inhalable particles with a diameter of 2.5 micrometres or smaller. In order to know how small these particles are we can compare them with an average human hair which is about 70 micrometres in diameter. Thus PM2.5 particles are about 30 times smaller than the average human hair [6]. Another advantage is that the air sample collected can be analysed for determining the chemical composition. However, gravimetric sampling also has disadvantages that limit its usefulness to our system. The required preparation and analysis of filters results in a significant time delay before the dirt intake data is available. As a result this technique does not give measurements in real-time. It is a time integrated sampling method that does not provide information on the temporal pattern of the dirt pick-up but gives the amount of dirt picked-up at the end of a time interval. Furthermore, it is not possible to use this technique to sample over short periods of time because of the difficulty in preparing and exchanging a new filter for each interval of sampling. Thus the process would not be autonomous and hence this approach is not considered. 2.2.2 Microphone

The next technique that was investigated was the feasibility of using a microphone to obtain the dirt pick-up measurements. In this technique the microphone is mounted in the intake valve of the vacuum cleaner and constantly listens to the sound of dirt flow in the passage. As the amount of dirt picked-up increases the noise in the inlet valve also increases. This causes a proportional increase in output voltage corresponding to the increase in dirt. The advantage of this technique is that it is simple to use and has a simple setup consisting of a microphone and the corresponding amplification circuitry. However, the drawback of this technique is that the microphone will pick up external noise in the surrounding area and give false readings of dirt pick-up. Furthermore, the vacuum cleaner is a very noisy system due to the various motors used in it. It will also be difficult to listen to very fine dust which does not make a lot of sound. Furthermore, over time, the surface of the microphone will get contaminated by the dirt flowing past it and this will contribute to noise and false measurements.

2.2.3 Strain Gauge

One of the easiest methods to measure weight is using a strain gauge or a load cell. However, depending on the type of dirt picked up, the measured weights would vary. For example, a piece of gravel would weigh more than an equal volume of dust. Hence it would not be possible to use this type of sensor with different dirt composition as it would give inaccurate values.

2.2.4 Dust Sensors

Technical Background

7 other stringy material and one of the manufactures of these sensors warns that big and stringy dirt should not enter the sensor due to its structure and mechanism. In addition, most of these sensors have a box shape with a small opening in the centre and in order to measure the dirt coming in across the entire cross section of the inlet valve, several of these would need to be used along the cross section of the valve which would block the path of the dirt flow and make the vacuuming process very inefficient. Hence these sensors are not considered as a solution to the problem to be solved, but rather we look at the sensing techniques used by these sensors and strive to build our own sensing system.

2.2.5 Light Sensors

A very interesting and simple technique that can be used to measure the amount of dirt picked is light sensing.

The idea is that dirt intake measurement can be done either by using the principle of scattering of light or the principle of light absorption or a combination of both. As given in [8], opacity of light can be used to measure the intensity of dirt. This technique measures the decrease in intensity of light due to absorption and scattering as the beam crosses the path of the dirt according to Beer-Lambert’s Law [9]. Simply put, Beer-Lambert’s Law states that the intensity of light at the receiver depends on the distance between the light source and light receiver as well as the concentration of the substance between them.

Figure 2-1 shows a simple setup that uses this technique. The basic operational principle is that a beam of light is directed through a stream of dirt toward the receiver. The reduction in received light intensity is used to obtain the concentration of the substance or material between the light source and light receiver.

Technical Background

Design Methodology

9

3 Design Methodology

This section explains the methodology used as well as the different design stages and decisions that were taken so as to fulfill the goal of this thesis. 3.1 Theory of Operation

This project uses the principle that when light is incident on a photodiode, a current is generated proportional to the incident light. This principle is used in our project as follows: The idea is that if we mount infrared LEDs and photodiodes opposite each other in the inlet valve of the vacuum cleaner, when the room is clean and the robot is not picking up any dirt, the light intensity received by the photodiodes will be maximum and hence the generated current will be the maximum possible for the particular setup. However, when the vacuum cleaner picks up dirt, this dirt flowing through the valve will block some or all the light incident on the photodiodes and as a result, the current generated will decrease in proportion to the amount of dirt picked up. This decrease in current when measured over time will give the amount of dirt picked up by the vacuum cleaner for that particular period of time.

Since most processors work with voltage rather than current values, in our system, the photodiode current will first be converted a proportional voltage and hence the drop in voltage over time will be measured to determine the amount of dirt picked up. This drop in voltage integrated over time will give a Voltage-Time value which will be used to determine the amount of dirt picked up.

3.2 Data Collection and Analysis

Design Methodology

10

Figure 3-1. Photodiode Voltage when no dirt is picked up

However, when the dirt picked up by the vacuum cleaner passes through the inlet valve, the incoming dirt blocks the light and thus the voltage decreases in proportion to the dirt picked up. The drop in voltage for four photodiodes is show in Figure 3-2 below.

Figure 3-2. Signals generated by incom ing dirt

Design Methodology

11 product. Under no dirt condition, the Voltage-Time value will be nearly zero and as the dirt picked up increases, the Voltage-Time value will also increase proportionally.

Figure 3-3. Integrating photodiode signals

Thus, the method used to characterize and evaluate our system will be as follows:

 Spread different quantities of dirt in front of the vacuum cleaner

 Vacuum up the different quantities of dirt

 Weigh the dustbin to obtain the actual amount of dirt picked up

 Calculate the Voltage-Time vale corresponding to each of the dirt quantities.

 Find a mathematical relationship between the quantity of dirt picked up and the corresponding Voltage-Time values to obtain a dirt prediction model.

 Use this dirt prediction model to accurately determine dirt being picked up by the vacuum cleaner in real-time by using Voltage-Time values.

System Design and Experimental Setup

12

4 System Design and Experimental Setup

This section describes the hardware and software implementation of the design in detail as well as the experimental setup.

4.1 Hardware Implementation

The process of building the prototype was done step by step and at each stage, signals were observed and decisions were made regarding the design of the subsequent stages.

The main blocks that comprise the dirt measuring system are as given below:

 Light Source Board: Infrared LED strip

 Light Detection Board: Photodiode strip

 TIA Board: Transimpedance stage consisting of operational amplifiers configured as transimpedance amplifiers

 A/D Board: Analog to digital conversion stage

 Processor: Arduino MEGA

 Voltage Regulator Board: Provides regulated voltage supply to the different boards

 Computer: Displays the results

A block diagram of the interconnection and the signal flow between the different boards is illustrated in Figure 4-1.

System Design and Experimental Setup

13 4.1.1 Light Source Board

The Light Source Board contains the light sources required for the measurement of dirt pick-up. It consists of infrared LEDs from Marktech Optoelectronics with PN# MTE9440M3A [10]. The LEDs have a peak emission wavelength of 950nm. The main reason for selecting these LEDs is because they have a wide viewing angle of 160 degrees. This was required because it would compensate for misalignments in the LED and photodiode orientation. Each LED has a current limiting series resistor of 220 ohms to prevent thermal runaway and to have an LED current of 20mA each. At 20mA, the forward voltage drop across the LED is typically 1.2V and that across the series resistor would be 4.4 V. Thus the board operates at a supply voltage of 5.6V and the schematic of the same is as shown in Figure 4-2.

Figure 4-2. Schem atic - Light Source Board

4.1.2 Light Detection Board

System Design and Experimental Setup

14

Figure 4-3. Schem atic - Light Detection Board

4.1.3 TIA Board

As mentioned earlier, the output of the photodiodes is a current which is proportional to the amount of infrared light incident on it. Depending on the amount of incident light, the output current of each diode varied from around 39µA to 127µA. In order to convert this current variation to a proportional voltage we use transimpedance amplifiers. A transimpedance amplifier (TIA) is an operational amplifier connected as shown in Figure 4-4 below. The output of the transimpedance amplifier is a voltage which is equal to a product of the photodiode current (Ip) and the feedback resistance (Rf). The feedback resistance is selected so that the product of diode current and resistance is 4.97V. Thus when no dirt is being picked up the effective output voltage across all diodes is 4.97V. And when dirt is picked up, the sum of these voltages drops in proportion to the amount of dirt. A voltage of 4.97 V is selected so as to match the voltages of the A/D evaluation board.

System Design and Experimental Setup

15 The amplifier configured as a transimpedance amplifier is an operational amplifier from Texas Instruments with PN# TLV274IN which is a low power, wide bandwidth device with rail to rail output [12]. The TIA board operates at 4.97V and is designed as shown in Figure 4-5.

Figure 4-5. Schem atic - T IA Board

4.1.4 A/D Board

System Design and Experimental Setup

16 Serial Peripheral Interface bus

The Serial Peripheral Interface bus (SPI) [16] is a synchronous serial communication interface specification used for short distance communication.

SPI devices communicate in full duplex mode using master-slave architecture with a single master [17]. The master can communicate with several slave devices through selection with individual slave select (SS) lines as shown in Figure 4-6.

The SPI protocol specifies 4 signals:

 Serial Clock (SCLK): The serial clock is provided by the master to the slave. This clock directly controls the analog to digital conversion and readout processes.

 Master Out Slave In (MOSI): Digital data and commands are sent by the master to the slave on this pin.

 Master In Slave Out (MISO): The converted digital samples are transmitted to the master on this pin

 Slave Select (SS): This spin is asserted by the master so as to select a certain slave device

System Design and Experimental Setup

17 4.1.5 Processor

The processor used in this project is an Arduino Mega 2560 Rev3 [18]. It is a microcontroller board based on the ATmega2560 [19] and has a clock frequency of 16MHz. It operates at 5V and has an input voltage range (limit) of 5-20V. The Arduino Mega can be programmed with the Arduino Software IDE in C. The ATmega2560 has a 256KB flash memory. It also has 8KB of SRAM and 4KB of EEPROM.

The connection between the Arduino and the analog-to-digital convertor board is as shown in Figure 4-7.

Since our system requires an ADC with a resolution of 8 bits or higher, the use of the built in A/D convertor of the Arduino Mega was also considered. However, when using this in built ADC, the time taken by the Arduino to convert one analog input value to a digital value is about 120µsec per channel. Thus for our system with 4 analog inputs it would be a total sampling time of 480µsec or a sampling frequency of about 2kHz. Therefore, although the Arduino Mega had a 10 bit built in ADC, it was not used since the sampling rate for our system was required to be 10kHz or higher.

Figure 4-7 . Interconnection between the ADC Ev aluation Board and the Arduino

4.1.6 Voltage Regulator Board

System Design and Experimental Setup

18

Figure 4-8. Schem atic of Voltage Regulator Board

4.1.7 Computer

The computer used in this system is Dell Precision M4700 with an Intel Core i7-3720QM processor and running the Windows 7 Enterprise operating system. The computer is used to program the Arduino as well as display the results and Voltage-Time values as calculated by it. It also provides the power to the Arduino.

4.2 Software Implementation

In this section we will go into detail about the software implementation and the algorithms of the programs that were used.

The main task of the software is to:

 Read the variation in voltage signals generated by incoming dirt

 Process these signals to generate a Voltage-Time value

 Predict the amount of dirt picked up corresponding to the Voltage-Time value.

Signal Read and Processing of Input Signals

System Design and Experimental Setup

19 generated analog signals have a minimum signal period of about 200µsec which gives a maximum signal frequency of approximately 5kHz. Thus, as per the Nyquist sampling theorem, the sampling rate should be greater than 10kHz.

In general we would like the signals to be sampled as fast as possible to accurately replicate the input signals. Since the ADC uses the SPI protocol to communicate with the Arduino, the only limiting factor is the clock speed of the SPI which in turn is limited by the clock frequency of the Arduino which is 16MHz. The maximum SPI clock that the Arduino MEGA can generate is the clock frequency/2. Thus the SPI clock in our system operates at 8MHz which causes the ADC to sample at a frequency of 500kHz. Since we read 4 c hannels of the ADC and process the read signals, as per our code we read each channel at an approximate frequency of 40KHz, which in the time domain is every 25µsec.

The SPI signals from the Arduino are shown in Figure 4-9 below. The blue signal is the Slave Select (SS), the green signal is the Serial Clock (SCLK) and the yellow signal is the Master Out Slave In (MOSI).

Figure 4-9. SPI Signals from the Arduino

System Design and Experimental Setup

20 Thus the algorithm for each area of the room is as below:

 Set Voltage-Time value = 0 and VoltageSum=0

 Read channel 1 of ADC and convert the decimal value to voltage value

 Add this voltage values to VoltageSum

 Read channel 2 of ADC and convert the decimal value to voltage value

 Add this voltage values to VoltageSum

 Read channel 3 of ADC and convert the decimal value to voltage value

 Add this voltage values to VoltageSum

 Read channel 4 of ADC and convert the decimal value to voltage value

 Add this voltage values to VoltageSum

 Subtract VoltageSum from MaxVoltageSum to obtain VoltageDrop

 Multiply VoltageDrop by the sampling period of the four channels and add this value to Voltage-Time

An explanation of the terminology used is as below:

VoltageSum: It is the variable that contains the sum of voltages of all four channels for one sampling period

VoltageDrop: It is the decrease in voltage due to intake of dirt

MaxVoltageSum: It is the sum of voltages of all the four channels when no dirt is picked up.

System Design and Experimental Setup

21

System Design and Experimental Setup

22

System Design and Experimental Setup

23 Prediction of Dirt Picked-up

The Voltage-Time value is used to calculate the amount of dirt picked up in a specific area. This is done by finding a mathematical relationship between the observed Voltage-Time value and the actual amount of dirt picked. The mathematical equations that define this relationship as well the process of obtaining this relationship is explained in detail in section 5.1

4.3 Experimental Setup

The experimental and test set up consists of the prototype vacuum cleaner connected to a suction generating system. The vacuum cleaner that is used to test and verify our design is the PUREi9 [23] robotic vacuum cleaner form Electrolux AB.

In order to build the prototype, the PUREi9 is modified in several ways. First, so as to mount the LEDs and photodiodes in the dirt intake valve, holes had to be drilled in the valve and then the LEDs were glued with hot glue so as to avoid any leaks. These components were then soldered to their respective boards. This was one of the most critical parts in the design of the system since the LEDs and photodiodes had to be or iented opposite each other for maximum illumination. The boards were designed and built by the author and were made using a Veroboard. The placement of the infrared LEDs and the photodiodes in the inlet valve is illustrated in Figure 4-12.

Figure 4-12. LED and Photodiode Placem ent in Inlet Valv e

The suction force in the robotic vacuum cleaner is generated by a motor in the back of the device. In order to have good control and maintain a precise flow rate, this motor has been removed in the modified prototype vacuum cleaner. Instead, the suction force is generated by an external vacuum cleaner connected to the test set up. This required creating mechanical adapters that would fit snugly into the suction tube and vacuum cleaner without any loss of flow pressure.

System Design and Experimental Setup

24 14 photodiodes and 5 LEDs were used, equally spaced in the inlet valve. However, since analyzing multiple channels on an oscilloscope or a group of oscilloscopes was infeasible, for the scope of this project, it was decided to narrow the path of dirt flow in the inlet valve. Hence the cross sectional area of the inlet valve was reduced by sealing the beginning and the end of the valve opening and proportionally reducing the volumetric flow rate so as to maintain a constant airflow velocity in the inlet valve.

This modified vacuum cleaner containing the dirt measuring circuitry and attached to the suction generating source consists of the test setup. A conceptual picture of the same is shown in Figure 4-13 below. Images of the prototype can be found in Appendix I.

System Testing and Results

25

5 System Testing and Results

This chapter describes the tests that were performed to verify the working of the different sub modules and the system as a whole. This chapter also presents the results obtained.

5.1 System Testing

In order to test the working of the system, the house dirt was simulated by using breadcrumbs. The “Ekologiskt Ströbröd” brand of breadcrumbs from Garant was used for performing the tests. The system was characterized for house dirt between 0 and 40g.

To evaluate the system, different quantities of breadcrumbs were picked up by the vacuum cleaner and the corresponding Voltage-Time values calculated by the Arduino were recorded. A plot of the Voltage-Time values versus the dirt picked up is given in Appendix B.

As per the results obtained, a mathematical model was built so as to predict the amount of dirt picked up based on the Voltage-Time value calculated by the Arduino.

The model consists of two equations – one that predicts the dirt between 0 to 10g and another that predicts the dirt picked up between 10 and 40g. These mathematical equations along with their corresponding R-squared values are given in Appendix B.

The R2 _{value is a measure of how close the data is fitted to the regression line}

i.e. how well our models or mathematical equations represent the data. In general, a higher R2 _{value implies a better model.}

5.2 Results

In order to evaluate the performance of the system and the models obtained in the previous section, different quantities of dirt were picked-up by the vacuum cleaner and the values of dirt picked-up as predicted by the Arduino were recorded.

System Testing and Results

26

Figure 5-1. T y pes of Dirt Used

5.2.1 Breadcrumbs

The system and the mathematical models obtained in section 5.1 System Testing, were evaluated using the same type of dirt i.e. the breadcrumbs that were used to obtain these models.

The result of the performance of the system with bread crumbs is given in Appendix C.

5.2.2 Rubber Balls

In order to check the behavior of the system to different types of dirt, the breadcrumbs were substituted with small rubber balls which are larger in size than the breadcrumbs.

System Testing and Results

27 5.2.3 House Dirt

Although testing the system with a uniform dirt type such as breadcrumbs and rubber balls has the advantage of consistency, evaluating the system using real house dirt is interesting because this is the type of dirt that the vacuum cleaner will be cleaning in reality. Hence this design was also tested using real house dirt.

Conclusions

28

6 Conclusions

The inferences made from the results obtained in the previous section will be discussed here. The short comings of the system as well as possible future work will also be described here.

6.1 Discussion and Conclusion

A system to measure the dirt picked-up in real time was designed and developed. It was able to measure the amount of dirt picked up. It was also tested with different types of dirt such as rubber balls and house dirt. The developed solution was also able to measure these different dirt types. A more detailed conclusion of the performance of the system can be found in Appendix F. Appendix F also gives the conclusions made with different dirt types.

6.2 Cost Estimation

The detailed cost estimate of building the sensing system is given in Appendix G. The cost of a processor has not been included in this estimation since the existing processor already used in the vacuum cleaner could be used for the purpose. The requirement while choosing a processor would be to ensure that it has the appropriate communication interface to connect to the ADC and it should be able to provide the required clock signal to the ADC to have the necessary sampling rate.

6.3 Future Work

Although the prototype is able to detect the amount of dirt picked-up, the system can be improved in several ways to be more accurate:

 Part of the system uses a switching supply and part of the system uses a regulated LDO output voltage. In order to avoid power supply noise and voltage shifts it would be better if all the boards of the system would use a single regulated output voltage so as to have a common reference voltage.

 Since this is a prototype, and it would require fine tuning throughout the project, prototyping boards and jumper cable were used to construct the circuits. In order to avoid noise and parasitic effects due to long wires, a great improvement in the design would be to have all the modules integrated together into a single PCB.

Conclusions

29 and protective measures are used to ensure that such fine particles do not adhere to the surface of the sensing elements. Two techniques that can be used to ensure that fine dirt types do not adhere to the surface of the sensors is given in Appendix H.

 Another aspect that should also be considered here is that different people may have a different perception of dirt. i.e. an area that appears dirty to one person may seem clean to another. Thus an appropriate model should be built by taking into account such scenarios. Performing user trials and surveys can help improve the prediction model to include such aspects and improve the performance of the system.

References

30

References

[1] International Federation of Robotics Press Release. [Accessed on: 31st May 2018]. [Online]. Available: https://ifr.org/ifr-press- releases/news/31-million-robots-helping-in-households-worldwide-by-2019

[2] International Federation of Robotics Press Release. [Accessed on: 9th

May 2018]. [Online]. Available:

https://ifr.org/downloads/press/02_2016/2016-DEC_20_IFR_press_release_service_robots_2019_FINAL_QS.pdf

[3] A. Håkansson. “Portal of Research Methods and Methodologies for Research Projects and Degree Projects”, WORLDCOMP'13, pp. 1-7, 2013

[4] Jee Young Kim, Shannon R. Magari, Robert F. Herrick, Thomas J. Smith, David C. Christiani & David C. Christiani, “Comparison of Fine Particle Measurements from a Direct-Reading Instrument and a Gravimetric Sampling Method”, Journal of Occupational and Environmental Hygiene, vol. 1, issue 11, pp. 707-715, 2004

[5] United States Environmental Protection Agency. [Accessed on: 10th

November 2018]. [Online]. Available: https://www.epa.gov/

[6] Particulate Matter (PM) Basics. [Accessed on: 10th _{November 2018].}

[Online]. Available: https://www.epa.gov/pm-pollution/particulate-matter-pm-basics

[7] Application note of Sharp dust sensor GP2Y1010AU0F. [Accessed on: 9th _{May 2018]. [Online]. Available:}

http://www.sharp-world.com/products/device/lineup/data/pdf/datasheet/gp2y1010au_a ppl_e.pdf

[8] Beatrice Castellani, Elena Morini, Mirko Filipponi, Andrea Nicolini, Massimo Palombo, Franco Cotana and Federico Rossi, “Comparative Analysis of Monitoring Devices for Particulate Content in Exhaust Gases”, Sustainability, vol. 6, issue 7, pp. 4287-4307, 2014

[9] D. F. Swinehart, “The Beer-Lambert Law”, Journal of Chemical Education, vol. 39, issue 7, pp. 333-335, 1962

[10] IR LED MTE9440M3A. [Accessed on: 9th _{May 2018]. [Online].}

Available:

http://www.marktechopto.com/pdf/products/datasheet/MTE9440M3 A_2011_08_11.pdf

[11] Photodiode PD333-3C/H0/L2. [Accessed on: 9th _{May 2018]. [Online].}

References

31 [12] Operational Amplifier TLV274IN. [Accessed on: 9th _{May 2018].}

[Online]. Available: http://www.ti.com/lit/ds/symlink/tlv274.pdf

[13] Analog to Digital Converter ADC128S102. [Accessed on: 9th _{May 2018].}

[Online]. Available: http://www.ti.com/lit/ds/symlink/adc128s102.pdf

[14] Analog to Digital Converter Evaluation Module ADC128S102EVM. [Accessed on: 9th _{May 2018]. [Online]. Available:}

http://www.ti.com/lit/ug/snau167/snau167.pdf

[15] F. Leens, “An introduction to I2C and SPI protocols”, Instrumentation & Measurement Magazine, IEEE, vol. 12, no. 1, pp. 8–13, 2009.

[16] Louis E. Frenzel Jr, “Handbook of Serial Communications Interfaces”, pp. 143–145, 2016.

[17] Serial Peripheral Interface Bus. [Accessed on: 31st _{May 2018]. [Online].}

Available:

https://en.wikipedia.org/wiki/Serial_Peripheral_Interface_Bus

[18] Arduino MEGA 2560 Rev3. [Accessed on: 9th May 2018]. [Online]. Available: https://store.arduino.cc/usa/arduino-mega-2560-rev3

[19] Atmel ATmega2560. [Accessed on: 9th May 2018]. [Online]. Available:

http://ww1.microchip.com/downloads/en/DeviceDoc/Atmel-2549-8-

bit-AVR-Microcontroller-ATmega640-1280-1281-2560-2561_datasheet.pdf

[20] Positive Voltage Regulator LM317. [Accessed on: 29th _{May 2018].}

[Online]. Available: http://www.onsemi.com/pub/Collateral/LM317-D.PDF

[21] M. Parker, “Digital signal processing 101 everything you need to know to get started”, Second edition, Amsterdam ; Boston: Elsevier / Newnes, pp. 21-30, 2010

[22] J. Broesch, “Digital signal processing demystified”, Engineering mentor series, Solana Beach, Calif.: HighText publications, pp. 49-58, 1997 [23] Robotic Vacuum Cleaner PURE i9. [Accessed on: 9th May 2018].

[Online]. Available:

TRITA-EECS-EX-2018:451