Evaluation and Implementation of a Longitudinal Control in a Platoon of Radio Controlled Vehicles

(1)

IN

DEGREE PROJECT VEHICLE ENGINEERING, SECOND CYCLE, 30 CREDITS

STOCKHOLM SWEDEN 2017,

Evaluation and Implementation of a Longitudinal Control in a Platoon of Radio Controlled Vehicles

DANIEL ROSHANGHIAS

KTH ROYAL INSTITUTE OF TECHNOLOGY

(2)

(3)

Abstract

Over the past decades, congestion and emission problems has increased remarkably which escalates the demands on vehicles. The advancements within the field of information and communication systems gives the opportunity to deal with the aforementioned problems. The concept of platooning shows to be an attractive way of reducing both congestion and emissions by having a short inter-vehicle spacing. The findings in studies show that fuel reduction potentials of 5-20 % are viable as a result of the lowered air drag by driving in platoon. This thesis investigates the state of the art within the area of intelligent transport systems (ITS) along with advanced driver assistance systems (ADAS). Furthermore, the prosecuted work results in a proposed control design for a longitudinal control in a platoon of vehicles. The platoon consists of two homogeneous radio controlled vehicles (RCV) which are modelled by taking advantage of system identification methods. The identified plant models are implemented into a Simulink model where the control system is developed. Moreover, the developed control system is implemented into a real-time demonstrator for experimental evaluation. The results shows that the modelled dynamics corresponds reasonably well with the real dynamics of the system. The developed control system proves to work well and agree with the expectations of its performance obtained from simulations.

The performance of the proposed controller has been evaluated by means of simulations and real experiments. The resulting control system consists of PID controllers for both speed and spacing control.

Keywords: Advanced Driver Assistance Systems, Cooperative Adaptive Cruise Controller, Intelligent Transportation Systems, Platooning, System Identification, Vehicle-to-vehicle communication

(4)

Sammanfattning

Under de senaste decennierna har mängden trafikstockningar och problem med utsläpp ökat - därmed även kraven p˚a v˚ara fordon. Samtidigt skapar framstegen inom informations- och kommunikationssystem möjligheter för att hantera ovannämnda problem. Kolonnkörning, eller platooning har visat sig vara en effektiv metod för att minska s˚aväl trafikstockningar som utsläpp som en följd av kortare avst˚and mellan fordon. Resultat fr˚an studier visar hur en bränslereduktion runt 5-20 % är möjlig till följd av det sänkta luft- motst˚andet vid kolonnkörning. Avhandlingen undersöker teknikens st˚andpunkt inom intelligenta transportsystem (ITS) tillsammans med avancerade drivhjälp- system (ADAS). Vidare resulterar arbetet i ett förslag till regleringsdesign för en longitudinell kontroll i en kolonn av fordon. Kolonnen best˚ar av tv˚a homogena radiostyrda fordon (RCV) som modelleras genom att utnyttja metoder för systemidentifiering. De identifierade systemmodellerna imple- menteras i en Simulink-modell där styrsystemet utvecklas. Dessutom im- plementeras det utvecklade styrsystemet i en realtids-demonstration för ex- perimentell utvärdering. Resultaten visar att den modellerade dynamiken stämmer bra överens med systemets verkliga dynamik. Det utvecklade styrsystemet visar sig fungera bra och överensstämmer med förväntningarna p˚a dess prestanda som erh˚allits genom simuleringar. Den föreslagna regulatorns prestanda har utvärderats med hjälp av simuleringar och verkliga experiment.

Det resulterande styrsystemet best˚ar av PID regulatorer f¨or b˚ade hastighets- och avst˚andskontroll.

(5)

Acknowledgement

I should express my gratitude to those various people that have contributed to this thesis in one way or another. First, I would like to thank my ex- aminer and supervisor Jonas M˚artensson for your guidance and enthusiasm throughout the thesis. Professor Cristian Rojas also deserve my gratitude for his kindness and guidance throughout the system identification procedure.

Detlef Scholle, my advisor at Alten Sweden, deserves a lot gratitude for giv- ing me the opportunity to do my thesis at the company. Hani, thank you for helping me get in touch with Detlef which resulted in completion of this thesis!

I offer my deep, sincere and painfully inadequate thanks to my team, Emil, Erik, Hanna and Sanel. Without you guys this thesis would not have been possible. Heartfelt thanks, too, to my friends Christian, Alan, Dara and Ramtin whom made the 5 years at KTH bearable. As both coursemates and above all friends, their tolerance and moral support have played no small part in this endeavour.

Finally, I would like to thank deeply the unconditional love and support of the most important persons in my life: my parents Mosa and Marianna and my beloved Elin. Your love has made this journey enjoyable and your support has been crucial to overcome the difficulties along these 5 years at KTH. This thesis is entirely yours. I love you! Thank you one. Thank you all.

Daniel Roshanghias Stockholm

(6)

Abbreviations

Abbreviation Description

ECU Electric Control Unit HDV Heavy-Duty Vehicle

ICT Information and Communications Technology ITS Intelligent Transport Systems

GPS Global Positioning System ACC Adaptive Cruise Controller

CACC Cooperative Adaptive Cruise Controller ADAS Advanced Driver Assistance Systems

PWM Pulse-Width Modulation FPE Final Prediction Error RCV Radio Controlled Vehicle

(7)

Chapter 1 Introduction

1.1 Motivation

The transport industries are facing great challenges. The European Union has set a goal to reduce greenhouse gas emission by 20 percent by 2020 compared with 1990 [7]. Heavy-duty vehicles (HDV) account for 17 percent of the total CO₂ emissions [8]. In order to meet the set goals it is of paramount importance to improve global efficiency and reduce the fuel consumption of all vehicles. Due to the increment of road freight traffic, CO₂ emissions from HDVs rose by 36 percent between 1990 and 2010 [6]. Freight transport de- mand is escalating quickly and will continue doing so as economy grows. This is clearly incompatible with the goals of reducing the emissions.

Congestion is another problem that arises with the increased freight transport. Along with the increased emission and congestion problem, vehicle manufacturers experience the increase in fuel costs. There is no surprise that the changed oil price and increase in oil consumption is a part of the escalated demands on vehicles. As the fuel prices increases, the strain on operating costs grows for a HDV fleet provider. As the fuel costs increase, the profit becomes smaller and road transport becomes less viable. This is a great disfavor for countries whose economy relies on road transport. It is clear that the transport sector faces great challenges regarding environmental issues but also when it comes to general performance for a sustainable future.

The advancements within the field of information and communications technology (ICT) gives the opportunity to deal with the aforementioned problems by integrating intelligent transport systems (ITS). By improved sensor technology, wireless communication, and GPS more advanced driver assistant

(10)

CHAPTER 1 1.1. MOTIVATION

systems are developed. Technologies such as vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) are becoming more prominent. Furthermore, the number of on-board electronic control units (ECU) and advanced sensors has increased remarkably during the last decades. These technologies enable increased functionality in terms of smart control logics [3].

HDV platooning is extensively studied and is one suggested way of reducing emissions and increase the energy efficiency. By having short inter-vehicle distance to the vehicle ahead will result in reduced fuel consumption as a consequence of lowered air drag. When packing HDVs close to each other, the total road capacity can be increased and emissions reduced. Integrating intelligent transportation systems will benefit all means of vehicles by im- proving the overall traffic flow through utilization of automated driving in traffic jams as one example. However, since the driver might feel uncomfort- able keeping a short distance towards the vehicle ahead, there is a need for a reliable control system that assures safety. Since the actions of one vehicle may in turn affect all the other vehicles in the linked chains, the automatic controller has to communicate effectively to the rest of the system in case of an emergency brake by any of the vehicles in the platoon. Governing vehicle platoons by automatic control can reduce the accidents and the traffic flow can be improved [13]. There are already existing systems which facilitates vehicle platooning, such as an adaptive cruise controller (ACC) that uses radars in order to measure the relative distance and velocity to the vehicle ahead and automatically maintain a safe distance [3]. The system does not make use of satellite nor any cooperative support from other vehicles. Hence, it only obtains information from on-board sensors. ACC works reasonably well for a small number of vehicles, such as two vehicles.

Due to delays a more reliable system is needed from a safety perspective when it comes to platooning. A typical example where a delay occurs could be: the time, from the moment when the brake is pushed, til when the brake torque is applied on the wheels and consequently slowing down the vehicle.

The following vehicle might not be able to reduce its speed in time if the preceding vehicles performs an emergency brake.

As an extension to radar measurements, wireless communication can be uti- lized instead to streamline the information between the preceding vehicles.

Although small delays are still imposed due to data processing, a vehicle can transmit wireless information about events from the preceding vehicles and not only from the vehicle ahead, which is only possible with only radar measurements. As a result, vehicles can operate at a much closer distance

(11)

CHAPTER 1 1.2. PROBLEM STATEMENT

and consequently perform better with a suitable controller since information is gathered almost instantaneously.

1.2 Problem statement

The appearance of vehicles driving in platoon is safety critical; hence, the controller should assure that a certain safe inter-vehicle spacing is maintained. By integrating a cooperative part in the already existing adaptive cruise controller, vehicles can operate at a much closer distance as a result of the wirelessly transmitted information. However, questions arises such as:

how should one construct a cooperative adaptive cruise controller (CACC) for maintaining short inter-vehicle spacing and at the same time guarantee- ing abortion in case of any malfunction, without compromising safety? What happens if there are any delays in the signal, which margins are necessary?

Disturbances and failures in the system have to be detected somehow, but which information has to be transmitted, and how should it be transmitted?

All these questions and thoughts addresses the safety issues with having a CACC-system within the vehicle platoon. This give rise to further research on how to assure safety with the increased complexity by integrating a cooperative and intelligent system, and which precautions that has to be taken in order to prevent accidents. Due to limitations and absence of real vehicles in the project, the research will be evaluated on two remote control vehicles (RCV). In summary, the aim of this thesis is to implement and validate a longitudinal control in a vehicle platoon. The controller shall be designed to be compliant with safety requirements.

1.3 Objectives

The specific objectives of this thesis are to:

• Review the literature concerning the already existing technology

• Investigate novel architectural solutions for advanced vehicles

• Model and analyze the architecture for cooperative adaptive cruise controller from a functional safety perspective

• Implement a longitudinal control system and run simulations to observe the behaviour in a vehicle platoon

• Validate the system in a real-time demonstrator consisting of two RCVs

(12)

CHAPTER 1 1.4. DELIMITATION

1.4 Delimitation

This thesis is a degree project in automatic control, therefore the research will mainly focus on the control theory part of the cooperative adaptive cruise controller. No further research on the actual communication or protocols will be covered since it is out of the scope for this project’s purpose. However, the work will be carried out in a manner so that the proposed controller will be compatible and prepared for the transmitted information that serves as inputs to the controller from the wireless communication. The actual implementation will be limited to a vehicle platoon consisting of two RCV with a partly working CACC system. However, the obtained results will lead to a discussion on whether the strategy approach is suitable for a platoon containing real and normal-sized vehicles or HDVs.

1.4.1 Research questions

Along the project, there are three research questions addressed that are related to the cooperative part of the adaptive cruise controller. These are:

• Which information has to be shared through wireless communication along the platoon for improved performance compared to the already existing ACC?

• What happens to the platoon if there are delays in the transmitted information from the wireless communication?

• How is the inter-vehicle spacing control adapted in case of connection loss between the vehicles, and how is this communicated to the system?

The evaluation will primarily consist of a comparison between the already commercially available ACC system and the novel CACC. Amongst other, their gap regulating error will be evaluated.

1.5 Contribution

This thesis is carried out at Alten Sweden who is a partner in three European projects; EMC2¹, SafeCOP², and AMASS³ whom all have a common goal,

1EMC²—Embedded Multi-Core systems for Mixed Criticality applications in dynamic and changeable real-time environments

2SafeCOP —Safe Cooperating Cyber-Physical Systems using Wireless Communication

3AMASS —Architecture-driven, Multi-concern and Seamless Assurance and Certifica- tion of Cyber-Physical Systems

(13)

CHAPTER 1 1.6. USE CASE

that is to bundle the power of innovation of different partners from embedded industry and research from several European countries. Their unified approach will force the breakthrough and deployment of Multi-Core technology in almost all application domains where real-time and mixed criticality are issues, and therefore strengthen the competitiveness of the European Em- bedded System industry [9]. Furthermore, the projects focuses on advancing the state of art for automotive embedded systems. Increasing the functionality of vehicles, for example, passenger cars as well as commercial vehicles such as trucks and buses, is a recurring task. The business and technical demands are derived from broad scope of applications and standards within functional safety, active and passive safety, communications inside and external to vehicle, emission regulations, fuel consumptions, and many more [9]. The projects have several use cases for which each involved partner can contribute their expertise to push forward the technology within that certain area. This thesis will contribute to one of these use cases that is described in more details in the next section.

1.6 Use Case

This thesis will indirectly contribute to one of SafeCOP’s use cases that targets cyber physical systems-of-systems whose safe cooperation relies on wireless communication [23]. The project have five use cases for different applications such as in healthcare, maritime, and automotive and weather domain. One of those case studies is ”vehicle control loss warning” [22].

1.6.1 Description of the use case

The use case aims to demonstrate how to apply and extend safety assurance framework to automotive cooperative V2X–based systems such as, auto- braking and platooning. This class of applications, namely safety critical applications presents new challenges such as trust of information sources, misuse, security, authentication and above all safety. Those who will benefit from solving the research issue at hand will be automotive suppliers involved in the design and development process, and also applications and ECUs for both cooperative and autonomous vehicles. Among others, all drivers and other road users will also have an advantage in getting the research issue solved. SafeCOP have defined three key research issues that are addressed in the use case. These are [22]:

(14)

• V2X protocols were not designed with safety (or even life) in mind, hence its use in cooperative autonomous vehicles requires additional mechanisms that need to be explored and defined, and these mechanisms need to be cost effective and able to be applied by automotive suppliers in their ECU design and development

• The convergence of V2X communication and sensor based technologies could deliver better safety, mobility and ”self-driving” capabilities than either approach could deliver on its own

• Current standards (like ISO 26262) does not cover autonomous or cooperative vehicles.

The autonomous vehicle industry is becoming more prominent and is ap- proaching commercial maturity in the next 5 to 10 years [22]. Sensors, positioning, imaging, guidance, control systems, mapping and V2X communications technologies are sources of data and technologies at the core of autonomous vehicles. Different ways of communicating, such as between vehicles (V2V) and between vehicle and infrastructure (V2I) allow different road users to share information and reach a unified goal, for example, vehicle platooning; thus adding new capabilities to transport systems.

There are several scenarios where V2X communication can prove to be use- ful. SafeCOP has illustrated a scenario where an established communication between the vehicles can prevent accidents. It follows:

”Consider a scenario where a long platoon of vehicles is travelling along a motorway. Suddenly the brakes fail on one of the cars, so a Control Loss Warning (CLW) should be sent to following vehicles. The vehicle involved broadcasts this CLW to sur- rounding vehicles and infrastructure through V2V and V2I communications (based upon 802.11p protocol). Upon receiving noti- fication of the control loss event, the receiving vehicle and infrastructure determines the relevance of the event and establishes ap- propriate actions: warn the other vehicles’ drivers, triggering full auto brake, communicating the event to relevant authorities (po- lice, emergency services) or other stakeholders (insurance compa- nies)” [22].

This scenario illustrates how one vehicle may affect the behaviour of the whole platoon as a result from one single cars performed action. Hence, this

(15)

application is safety critical and must be designed in accordance with methods for safety assurance. Furthermore, the main objective of the use case is to develop a CLW system. The system should notify all the concerned vehicles downstream in case of a vehicle in the platoon loses some functionality that may affect the whole platoon. The functionality loss can be a cause of mechanical failure or human action. From this warning the vehicles and infrastructures can determine whether actions are required or not; if so, what type of action is needed, for example, full auto brake or emergency service assistance. The vehicles must perform a safety analysis of the warning before taking any action to assure that the information sent is reliable. This can be done mathematically by evaluating safety sets for which the vehicle is in an acceptable safe region.

This use case-inspired scenario serves as an inspiration for the prosecuted work throughout the thesis. Two RCVs with implemented CACC-systems are subject to different traffic situations, such as road accidents and informed through V2I communication where proposed control actions are received to prevent the accidents from growing.

(16)

Chapter 2 State of the art

2.1 Related work

The concept of vehicle platooning proves to be lucrative and different approaches to incorporate new and already available technology are extensively studied. Early theoretical work on the subject was presented in [15], [17]

and [24] where the topic of string stability was raised. Constraints imposed by topography when driving in platoon is another problem that is considered. In an article [2], it has been studied whether it is more fuel-efficient to split or maintain a platoon that is facing steep uphill or downhill segments.

The results shows that an improved fuel-efficiency can be obtained by maintaining the platoon throughout the hill. By utilizing a look-ahead cruise control, which takes advantage of preview information, the velocity change can be initiated at a specific point on the road for all vehicles rather than simultaneously changing the velocity to maintain spacing. There are several control strategies for platooning where different ways of maintaining a certain velocity, constant spacing or sometimes a predefined safe time headway are introduced. The proposed controllers in the literature are mainly the proportional-integral-derivative controller (PID), linear quadratic regulator (LQR), and model predictive controller (MPC). They can all be viewed as an extension to the ACC concept and categorized based upon their intra-platoon spacing policies [3]. Either a constant spacing is used [10], a time headway [5], or a nonlinear spacing policy [27]. However, it has been concluded that the time headway policy is much more energy efficient, due to the fact that spacing policy requires a higher acceleration variability for handling disturbances [19],[3]. On the other hand, the disadvantage of having a constant time headway is that it results in larger steady-state spacing which in turn increases the length of the platoon and thereby decreases the benefits from

(17)

CHAPTER 2 2.2. INTELLIGENT TRANSPORTATION SYSTEMS

the reduced air drag [3].

2.2 Intelligent Transportation Systems

Intelligent Transportation Systems (ITS) are defined as those systems utilizing cooperative technologies and systems engineering concepts to develop and improve transportation systems of all kinds. ITS are of vital importance in order to increase the safety for transportation and tackle the problems with the increased emissions and congestion worldwide. They enable safer, more efficient and sustainable transport by applying various information and communication technologies to all modes of passengers and freight transport. Furthermore, taking advantage of existing technologies can provide new services. The development of ITS is fundamental to provide new job opportunities within the transport sector.

Transportation systems can be perceived as large mobile networks. By introducing smart decision making based on external events, intelligence is induced into the system. ITS, as illustrated in Figure 2.1, provide actors in the transportation network with information based actions. ITS applications can be categorized in five summarizing groups [12]:

• Advanced Traveller Information Systems provide drivers with real-time information, such as transit routes, navigation directions; and information about delays due to congestion, accidents, weather conditions, or road repair work

• Advanced Transportation Management Systems include traffic control devices, such as traffic signals, ramp meters, variable message signs, and traffic operations centers

• ITS-Enabled Transportation Pricing Systems include systems which provide information about electronic toll collection, congestion pricing, fee-based express lanes

• Advanced Public Transportation Systems, for example, allow trains and buses to report their position so passengers can be informed of their real-time status

• Fully integrated intelligent transportation systems, such as V2I and V2V integration, enable communication among assets in the transportation system, for example, from vehicles to roadside sensors, traffic lights, and other vehicles.

(18)

CHAPTER 2 2.2. INTELLIGENT TRANSPORTATION SYSTEMS

Figure 2.1: An illustration of ITS [11]

By utilizing these systems, the transportation network will experience an increase in safety, the operational performance is improved, the mobility and convenience are enhanced, environmental benefits are delivered, and the pro- ductivity is boosted as well as the economy is expanding and the employment grows [12].

HDV platooning is one way of incorporating with ITS. Research within intelligent vehicle platooning, especially HDV platooning, addresses several ITS target issues [4]. In situations where there is a high traffic intensity, vehicles typically tend to keep a small inter-vehicle spacing and hence, naturally form a vehicle platoon. Without any available information from ITS, drivers must rely on their decisions on sounds and their own sight which limits their driving capabilities and thereby resulting in lowered fuel-efficiency and driving comfort as a consequence of harsh acceleration and braking in order to maintain a short distance to the preceding vehicle. Such transient control actions results in increased pollution in the form of emission and road particles, as well as an increase in congestion. Intelligent vehicle platoons empowers different systems to communicate relevant information, such as position, velocity to

(19)

CHAPTER 2 2.3. VEHICLE PLATOONING

other vehicles; in addition to other systems which provide navigation help, information about congestion and road constructions. Human functions can be both replaced and supported by taking advantage of V2X communication devices in order to enhance safety, mobility, operational performance and environmental benefits [4].

2.3 Vehicle platooning

Fuel reduction potentials of 5-20 % has been shown to be viable through experiments conducted [1], [20], [26]. These results indicate that vehicle platooning is an attractive concept. In HDV platooning the mass and road slope plays a significant role on the system dynamics. The first in-depth studies of control for heavy vehicle platooning was conducted in the early 1960s [21]. A vehicle platoon consists of a chain of vehicles following each other in an automated way. The first vehicle in the platoon is usually known as the leader, or lead car and the rest of the vehicles are called followers, or host cars. In the simplest case each vehicle measures its distance with respect to the preceding vehicle with on-board sensors such as radars, lidars or cameras in order to maintain a safe distance to its preceding vehicle by controlling the velocity. An example of platooning in a real scenario can be seen in Figure 2.2. By governing vehicles platoons through the use of an automated control strategy the overall traffic flow is expected to be improved and several advantages arises [14]:

• increased traffic throughput and better usage of road capacity by automated driving or platooning as a result of the reduced inter-vehicle distance between vehicles. The results of a study [25], shows that the highway capacity can be increased up to 43 % if all vehicles in the highway enable platooning using on-board sensors (camera and radar).

The same study shows that if all vehicles use both sensors and vehicle- to-vehicle communication the increase is about 273 %

• reduced fuel consumption as a consequence of the reduced inter-vehicle distance between the vehicles which decreases the aerodynamic drag.

The results of a study [1], shows that the fuel consumption can be reduced up to 7 % by having a platoon consisting of only two trucks.

In addition to decreasing the inter-vehicle distance, platooning can also contribute to reduction in fuel consumption by avoiding unnecessary acceleration and deceleration

• reduced air pollution as an consequence of the reduced fuel consumption

(20)

CHAPTER 2 2.4. ENABLING PLATOONING TECHNOLOGIES

• increased safety since human reaction is naturally subjected to delay and the delay depends on the cognitive status of the driver. Accord- ing to statistics [18] human error accounts for 90 % of road accidents.

Hence, by utilizing Advanced Driver Assistance Systems (ADAS) platooning can mitigate accidents by enabling a safe inter-vehicle distance between the vehicles

• increased comfort by letting the driver be relaxed during the ride or even spend time on more enjoyable task such as reading the news or tune in a radio channel instead of focusing on driving on a congested road.

Figure 2.2: Vehicle platooning [4]

2.4 Enabling platooning technologies

In order to empower platooning at a short inter-vehicle distance without compromising the safety, the whole system demands a delicate engineering design. Interactions between three key enabling modules, namely communication, sensing and control are fundamental. Thanks to the improvements in wireless communication, vehicles are now able to both transmit and receive information to and from another vehicle and ultimately take different control actions. The same applies for communication between vehicle and infrastructure. By utilization of standard communication protocols, such as

(21)

CHAPTER 2 2.4. ENABLING PLATOONING TECHNOLOGIES

IEEE 802.11p, the messages can be sent in real-time and in a reliable way.

The V2X communication enhances the sensing module by providing valu- able information to the control module which cannot be obtained by only the on-board sensors. The sensing module consists of on-board sensors, and positioning devices such as GPS and compass. The transmitted information obtained from the communication module and the measured data from the sensing module both serves as good advisers to the control module responsible for decision making. The control system is responsible for maintaining a certain desired safe distance to the preceding vehicle by controlling both the lateral and longitudinal control, in the platoon whereas only the latter is covered in this thesis. The control action is sent to the vehicle actuators which is the engine throttle and the brake in case of longitudinal control and the steering wheel in case of lateral control. Platooning logics such as handling requests to join the platoon and ego vehicle’s platoon status are performed in the platoon logic block. The world awareness block identifies position and status of traffic lights and speed limits at each region. The supervisor block manages the low-level procedures such as acceleration and deceleration. A general system architecture of a vehicle in a platoon is depicted in Figure 2.3.

Figure 2.3: System architecture of an automated vehicle [14]

(22)

CHAPTER 2 2.5. PLATOON SYSTEM ARCHITECTURE

2.4.1 Challenges in the control of the platoon

As the inter-vehicle distance decreases, the safety becomes a critical issue to consider which leaves high demands on the controller. As previously stated, the whole systems requires a perfect symbiosis between the communication, sensing and control in order to reduce the distance to the preceding vehicle.

The control system has to guarantee safety and comfort for the driver and at the same time take the limitations of the actuators into account. In other words, the performed control actions should be compliant with the limitations of the system, such as the produced throttle from the engine and the maximum braking torque. Apart from the obvious, which is the stability of each individual vehicle, the whole vehicle chain has to remain stable during platoon mode with regard to the overall performance. The stability of a string of vehicles (platoon) is often referred to as string stability in the literature and is defined as the ability to attenuate disturbances in position, velocity and acceleration error as they propagate along the vehicle stream. To achieve string stability with constant inter-vehicle distance it has been shown that vehicle-to-vehicle communication is necessary. It is crucial to investigate uncertainties in the system that may affect the behaviour of the platoon. The uncertainties involve communication delays, noise in the measurements and uncertain actuator model. The behaviour of the platoon should be verified for a given controller in presence of the aforementioned uncertainties.

2.5 Platoon System Architecture

The hierarchical order of the different cruise controller systems can be depicted in Figure 2.4. The arrows in the architecture indicate the direction of the information flow in the system. The arrow from the vehicle to its controllers demonstrate the information obtained from the on-board sensors.

Hence, the states such as the velocity of the vehicle are fed back in a feedback control. The arrow between the preceding vehicle and the ACC indicate the information flow obtained from the radar which consists of the states with respect to the preceding vehicle. The arrows between the CACC and the Wireless Communication Network indicates the two-way information flow between the platooning vehicles, which serves as a platform for the V2X communication. Here, the vehicles can both transmit and receive information about the states of the vehicles itself and the whole platoon.

(23)

Figure 2.4: Platoon system architecture for N vehicles. The information flow is indicated by the arrows [3]

With regard to the different cruise controller systems which have some differences in functionality, their main functionality and purposes are listed below in their hierarchical order:

• The conventional CC improves the fuel economy, increases driver comfort over long distances by maintaining a certain desired speed. It uses the on-board sensors to measure the velocity and use that information in a feedback loop which in turn acts as a PI- or PID-controller to compensate for the errors in the desired speed. For the lead vehicle in the platoon, which do not have to keep a distance to a preceding vehicle, this is typically the used controller

• The ACC, which is an extension to the conventional CC and is available in almost all modern vehicles, aims to maintain a desired distance to a preceding vehicle. It does this by utilizing the relative distance and velocity information from the preceding vehicle obtained from the radar measurements as depicted in Figure 2.5. As for the CC, this system improves the fuel economy and increases the driver comfort even more.

As opposed to the CC, the ACC is allowed to actuate the brake and thus improve the safety since the system can react faster than a driver

• The CACC is the cooperative extension to the aforementioned ACC system. This system forms optimal decisions based on vehicle information within the operating range of its wireless transceiver. By im- plementing the cooperative behaviour in the platoon, the inter-vehicle distance can be reduced which results in lowered air drag and thereby lowering the fuel consumption. The safety is also improved since control actions are based on the behaviour of several vehicles ahead in the platoon and not only on the preceding vehicle. Aspects as safety, comfort and fuel-efficiency can be improved even further if limitations

(24)

of the actuators (engine throttle and braking capacity) of each vehicle are considered as constraints and taken into account. For clarifica- tion, it has to be ensured that the preceding vehicles does not brake harder than what the follower vehicles are capable off and the other way around when it comes to accelerating the vehicles

In the presence of system failure in any of the three layers, it is intended that the control actions to be degraded to the next layer; thus the CACC is degraded to ACC and so on till when the CC fails and the driver is instructed to take full control of the vehicle.

Figure 2.5: An illustration of an ACC system using a radar to measure the relative distance and velocity of the preceding vehicle [14]

(25)

Chapter 3 Modeling of the vehicle system

3.1 Longitudinal dynamics

There are several forces acting on a moving vehicle, both external and internal. The longitudinal dynamics of the vehicle serves as a basis when designing the longitudinal control. The imposed forces can be depicted in Figure 3.1.

The negative longitudinal forces constitutes of the air drag F_a, rolling resistance F_rdue to the resistive frictional force between the road surface and the wheels, the gravitational force F_g that can act either as a positive or negative longitudinal force depending on the incline α of the road. Apart from the external forces, there is the negative braking force F_b obtained if the brakes are applied and a positive longitudinal propulsion force F_p from the engine.

The last two mentioned internal forces acts as controller inputs to the whole controller. By applying Newton’s second law of motion a nonlinear vehicle model can be derived:

mdv

dt = F_p− F_b− F_r(α) − F_g(α) − F_a(v)

= F_p− F_b− C_rmgcos(α) − mgsin(α) − 1

2ρv²C_dA

(3.1)

where m and v is the mass and velocity of the vehicle, respectively. ρ is the air density, C_d is the drag coefficient of the vehicle, C_r is the rolling resistance coefficient, and A is the frontal area. The air drag expression is simplified since it is actually a function of the wind velocity and the air drag coefficient is a function of the distance to the preceding vehicle. Introducing a state space representation where x1 denotes the travelled distance and x2

the velocity the system can be expressed as follows:

(26)

CHAPTER 3 3.2. SYSTEM IDENTIFICATION

(x₁ = x(t) x₂ = v(t) ⇒

(˙x₁ = x₂

˙x₂ = _m¹F_p− F_b− F_r(α(x₁)) − F_g(α(x₁)) − F_a(x₂) (3.2)

Figure 3.1: The longitudinal forces acting on a vehicle in motion [3]

3.2 System identification

Due to the fact that the dynamics of the RCV does not completely corre- spond to the dynamics of a normal sized vehicle, it is essential to model the corresponding dynamics. In order to construct a suitable platooning controller, a good and adequate description of the model of the RCVs has to be created. As a consequence of the limited information about the system and the absence of data describing the properties of the motor and other parameters that give rise to negative longitudinal forces, a system identification of the RCVs is required. The foremost advantage with the carried out system identification experiment is that it captures all the relevant system dynamics in the environment it is driving in, that is, all the imposed forces acting on the RCV such as, the rolling resistance between the road surface and the wheels, the frictional forces in the bearings, the air drag — which is negligible for the speed that the car is travelling in — and the gravitational force are all included in the plant model.

(27)

System identification uses statistical methods to build mathematical models of dynamical systems from measured data. Since no prior model of the system is available it can be seen as a black box model, that is a system which can be viewed in terms of its inputs and outputs without any knowledge of its internal workings. The generated input is a sequence of steps which gives information about the system’s frequency response in all frequencies [16];

the signal has varying time duration but with same amplitude. The time behaviour of the system can be captured and the generated output can be measured as seen in Figure 3.2. For this system identification experiment, the input is a Pulse-Width Modulation-signal (PWM) given to the electric motor on the RCV and the filtered output is the measured rotational velocity from the encoders mounted on the wheels.

0 10 20 30 40 50 60 70 80

-5

0 5 10 15

Rotational velocity [rad/s]

Input and output signals

0 10 20 30 40 50 60 70 80

Time [s]

0 1000 2000 3000

Pulse-Width Modulation

Figure 3.2: Measured input and output signals from the system identification experiment

As seen in the time plot, in the time instance between 7-15 seconds an anomaly can be observed that differs from the rest of the output responses.

This comes from the fact that once the experiment is started and the input is starting to be generated from the computer it takes a while to move the

(28)

experimental setup and prepare the RCV to drive on the road and start mea- suring the output. This is accounted for by selecting a range of data that is accurate for both estimation and validation purposes.

As seen in Figure 3.3, half of the accurate data is used for estimating polynomial models and the other half is used for comparing their results. This is done by validating the models against the validation data set in order to select the model with the best fit and suitability for the purpose of the model. Since system identification is an iterative process, models with different structures are identified and their model performance is compared.

Ultimately, the simplest model that best describes the dynamics of the system is chosen according to the TSTF (try simple things first) principle. The principle suggests that simple, cheap and ready-made models (for example ARX) should be tested first and then go on to more complex models only if the simple models are not sufficient [16].

10 20 30 40 50 60 70 80

-5

0 5 10 15

Input and output signals

10 20 30 40 50 60 70 80

Time [s]

0 1000 2000 3000

Pulse-Width Modulation

Validation data Estimation data

Figure 3.3: Selected data range for estimation and validation purposes When estimating the polynomial functions the work flow is proceeded from the flow chart as seen in Figure 3.4 and TSTF; ARX models of different orders

(29)

are estimated and compared. Other model structures such as ARMAX, Box- Jenkins (BJ) and Output-Error (OE) are tested as well as the nonlinear ARMAX model, NARMAX. When comparing the different model structures and their orders, the trade-off between complexity and accuracy, or price and quality has to be considered. A more complex model structure and a higher order can give high quality, that is, small bias and variance inaccuracy. But the price in the form of modeling, programming, and computational work can be high. The purpose of the system identification for this project is to obtain an adequate model for simulation and control design possibilities. The model quality is based on the ability of the model to reproduce the behaviour of the system, that is, that the model’s simulated or predicted output is in good agreement with the outputs produced by the system. In this case it is desired that the model captures the right system response once it gets a certain input.

Figure 3.4: Identification cycle. The rectangles reflect the computer’s main responsibility and the ovals represents the user’s main responsibility

(30)

Some general point of the different model structures are listed below [16]:

• The ARX model A(q)y(t) = B(q)u(t) + e(t) is the easiest to estimate since the corresponding problem is of a linear regression type. Accord- ing to the aforementioned principle, TSTF, the ARX models is tested at first hand. The foremost disadvantage is that the disturbance model H(q, θ) = 1/A(q) comes along with the system’s poles. This can easily cause an incorrect estimation of the system dynamics because the A polynomial also has to describe the disturbance properties.

• The ARMAX model A(q)y(t) = B(q)u(t) + C(q)e(t) gives extra flexi- bility to handle disturbances because of the C polynomial. This is an often used model.

• The advantage with the OE model y(t) = ^B(q)_{F (q)}u(t) + e(t) is that the system dynamics can be described separately and that no parameters are wasted on a disturbance model.

• The BJ model y(t) = ^B(q)_{F (q)}u(t) + ^C(q)_D(q)e(t) is the considered as the most complete model since disturbance properties are modeled separately from the system dynamics.

Table 3.1: The different model structures and the corresponding model order with their corresponding best fit to the measured data

Model structure na nb nc nd nk nf Best Fit

ARX 1 1 - - 1 - 74.18 %

ARX 4 4 - - 1 - 75.23 %

ARX 5 1 - - 1 - 75.31 %

ARMAX 5 5 2 - 1 - 75.18 %

ARMAX 2 2 2 - 1 - 75.28 %

OE - 2 - - 1 2 74.70 %

OE - 2 - - 1 4 77.62 %

BJ - 2 2 2 1 2 74.68 %

BJ - 2 2 2 1 4 77.07 %

NARX 5 5 - - 1 - 74.28 %

NARX 5 1 - - 1 - 71.71 %

(31)

As seen in Table 3.1, the different model structures are represented with their model order and the corresponding best fit with respect to the measured data.

The evaluation of the different model orders is proceeded from Akaike’s Final prediction error (FPE) among other. This criterion provides a measure of model quality; after computing several different models, the models can be compared using FPE. According to Akaike’s theory, the most accurate model has the smallest FPE [16] which is a measure of prediction error variance.

However, only considering the prediction error can cause incorrect conclusions about which model that is best. A larger model, that is, with higher model order will always give a lower value of the criterion function (3.3), since it has been obtained by minimizing over more parameters.

V_N(θ) = 1 N

N

X

t=1

²(t, θ) (3.3)

The criterion function contains the prediction error, which is defined as,

(t, θ) = y(t) − ˆy(t|θ) (3.4) where ˆy(t|θ) is the prediction of the output, y(t) and θ is the parameter vector. The value V_N decreases since the model includes more and more of the system’s relevant properties. Even though a ”correct” model order has been passed, the criterion function continuous to decrease. The explanation is that the extra — and unnecessary — parameters are used to fit the model to the specific disturbance signals in the present data set. The model becomes

”overfit” and does not serve any purpose since the model will be used when other disturbances affect the system. On the contrary, the model will in fact be worse because of the overfit [16]. The goal is to find the transition from relevant model fit to overfit. Referring back to the TSTF principle, it can be observed in Table 3.1 that the ARX models generate satisfactory results when it comes to best fit. However, it can be discussed whether its better to use a BJ or OE model instead because of their better fit to the model. For this project’s purpose it is considered as sufficient with a simple model such as ARX or ARMAX. Furthermore, the more complex model’s best fit do not differ significantly compared to the simpler ones. In Table 3.1 it can be witnessed that the ARX551 model (na = 5, nb = 1, nk = 1 ) has a best fit of 75.31 % compared to the ARMAX2221 model (na = 2, nb = 2, nc = 2, nk = 1 ) with a best fit of 75.28 %. As mentioned earlier, the foremost disadvantage with the ARX model is that the disturbance model H(q, θ) = 1/A(q) comes along with the system’s poles and as a consequence of that, it is easy to get an incorrect estimate of the system dynamics because the A polynomial also has to describe the disturbance properties. Higher orders in A and B

(32)

than necessary may be required which it proves to be in this case. The ARMAX model with a system order of 2 instead of 5 as in the ARX model, proves to be sufficient without compromising the quality. In fact, this is a great advantage when designing optimal controls where the number of states determine the complexity of the controller. The ARMAX model becomes less complex as the model order is lower. Comparing the final prediction error, ARMAX2221 has a FPE = 0.0265 as opposed to ARX511 which has a FPE

= 0.02667; thus, the ARMAX model is better according to Akaike’s theory.

The identified model is considered as adequate for plant modeling and for this project’s purpose while adhering to the TSTF principle. In Figure 3.5 the measured and simulated data using the ARMAX model can be seen. As seen, the dynamics are different when the rotational speed is negative which is a consequence of the absence of brakes on the vehicles.

40 45 50 55 60 65 70 75 80

Time [s]

-2 0 2 4 6 8 10 12 14

Measured and simulated model output

ARMAX2221 Measured data

Figure 3.5: Measured and simulated data using ARMAX model

(33)

3.2.1 Residual Analysis

-20 -15 -10 -5 0 5 10 15 20

-0.4

-0.2 0

0.2 Autocorrelation of residuals for the output

-20 -15 -10 -5 0 5 10 15 20

Samples -0.2

-0.1 0

0.1 Cross correlation for the input and the output residuals

Figure 3.6: Autocorrelation of the residuals of the output and the cross correlation for the input and output residuals

From the residuals plot in Figure 3.6, especially the autocorrelation plot, there are a few conclusions that can be drawn; the residuals are the difference between the fitted model and the measured data. In a signal-plus-white noise model, if the model is fitted good, the residuals should be white noise.

This comes from the fact that the noise is a sequence of uncorrelated random variables following a normal distribution N (0, 1) [16]. This means that all the random variables have zero mean and unit variance. When inspecting the autocorrelation, it can be determined whether or not there is evidence of autocorrelation. In other words, if the sequence of residuals looks like in the autocorrelation of a white noise process, that is, if the values lie within the 99 %-confidence interval it can be concluded that none of the signal has escaped during the model fit and ended up in the residuals.

(34)

In this case, the autocorrelation values exceed the 99 %-confidence bounds for a white noise autocorrelation at some lags. Hence, it can be concluded that the residuals are not just a white noise sequence. The implication is that the model has not accounted for all the signals and therefore the residuals consists of both signal plus noise. However, it seems that it have captured most of the essential signals and, as previously mentioned, this model fit is still adequate for the purpose of the model.

(35)

Chapter 4 Implementation

4.1 Control design

With a description of the system dynamics of the car, obtained from the system identification, a comprehensive platooning control is designed. As seen in Figure 4.1, the whole system is divided into two subsystems: the lead car and the host car with its corresponding controller. The lead car is equipped with a cruise control (CC) system and intend to keep a constant speed by feedback information. The host car is equipped with both an adaptive cruise control (ACC) system and a cooperative adaptive cruise control (CACC);

these two systems can be switched in between depending on which mode the driver decides to drive in, namely, single mode or platoon mode.

Figure 4.1: Overview of the whole Simulink model where the leftmost box displays the lead car with its cruise control included, the box in the middle displays the sensed data from the lidar and the rightmost box displays the lead car with its longitudinal control (red box)

(36)

CHAPTER 4 4.1. CONTROL DESIGN

The host car has a lidar that measures the distance to the preceding vehicle, as well as the relative velocity of the lead car. When the host car drives in single mode and with the ACC system active, it operates in two modes:

speed control and spacing control.

• In speed control, the host car travels at a driver-set speed.

• In spacing control, the host car maintains a safe distance from the lead car.

The ACC system decides whether to operate in speed control or spacing control, as seen in Figure 4.2, based on real-time lidar measurements. For example, if the actual distance between the lead and host car is too short, the system switches from speed control to spacing control. Similarly, if the lead car is further away, the ACC system switches from spacing control to speed control. Consequently, the ACC system let the host car travel at a driver-set speed as long as a safe distance is maintained. The inputs to the ACC system are:

• Driver-set velocity, v_set

• Velocity of the host car, vi−1

• Velocity of the lead car, v_i (from lidar)

• Actual distance between the lead car and the host car, d_i,i−1

The desired safe distance between the cars is derived from a time headway, that is, the distance between the cars is a function of the velocity of the host car, as well as the time gap. The set time headway policy is defined as,

d_i,i−1^{saf e}= 25 + τ vi−1 (4.1)

where 25 [cm] is the determined standstill distance between the two RCVs in the experimental environment and τ is the time gap. The spacing error,

e = d_i,i−1− d_i,i−1^{saf e} (4.2) determines which mode the ACC system operates in. If e ≥ 0, that is, that the actual distance between the cars is greater than the safe distance then the control goal is to track the driver-set velocity, vset. On the other hand, if e < 0 then spacing control mode is active and the goal is to maintain the safe distance, d_i,i−1^{saf e}.

(37)

CHAPTER 4 4.1. CONTROL DESIGN

Figure 4.2: The two different operating modes of the ACC system depending on the spacing error

When the driver switches to platoon mode and the CACC system is engaged, the control policy is different. In this mode the host car maintains a shorter safe distance and intends to keep the same velocity profile as the preceding vehicle. Compared to when the vehicles operate in single mode and gather their information from on-board sensors and measure the distance and velocity of the preceding car, the different states can be received wirelessly. This is achieved by utilizing the V2V communication between the RCVs, where the first vehicle transmits its velocity, v_lead to the second car which in turn can adapt its own velocity with respect to the lead car.

Figure 4.3: The lead car communicating its states to the host car by utilizing wireless communication

(38)

CHAPTER 4 4.2. CONTROLLER

4.2 Controller

For the purpose of the longitudinal control of the two RCVs, it proves to be sufficient with PID controllers in the system; both for the lead car which travels with a constant speed, as well as for the host car which intends to maintain a safe distance to the preceding car as well as travel at a driver-set speed.

As aforementioned, the lead car is equipped with a PID controller in order to control the reference speed. By utilizing the encoders mounted on each wheel, the car can measure its own ego states (velocity and acceleration) and compensate for the errors from the reference in a feedback loop.

The host car is equipped with two PID controllers, one for speed control and one for the distance control. The distance to the lead car is measured by the lidar and is also fed back to the system to compensate for the errors in the desired reference.

u(t) = K_pe(t) + K_i Z t

0

e(τ )dτ + K_dde(t)

dt (4.3)

The choice of the controller parameters is easily tuned in simulations once the model of the car is acquired. The parameters are tuned manually until all the desired control responses are fulfilled. Beyond the requirement of stability, the control parameters are tuned such that the system responds quickly with a short rise time and settling time. Kp, Ki and Kd differ from each other in each PID controller. These constants are tuned such that the RCVs maintain their set speed and minimum safe distance. The control system on the host car that is equipped with two PID controllers, also have a switch that determines which PID that should be active during run-time. As described in the control design, the spacing error determines which controller that should be active and is automatically set by the switch. If the actual distance between the cars is greater than the safe distance then the switch sets the PID controller for speed control otherwise the other PID controller for spacing control is active.

(39)

Chapter 5 Results

5.1 Simulation

For simulation results, the identified plant model from the system identification, with its corresponding control system is used. The results of interest are the accelerations of the cars, the velocities, and the distances between the two cars with the corresponding spacing errors. The differences between the two driving modes, single or platoon mode can be depicted in the figures below.

The lead car is exposed to different reference velocities with step responses during different time instances; to simulate a realistic driving environment, the lead car both decreases and increases its velocity and the sensed data is subject to noise. The host car has a constant driver-set velocity which it intends to follow unless the lead car is in the range of the desired safe distance for which it switches to distance control.

5.1.1 Single mode

In Figure 5.1, both the cars travel in single mode drive. In order to see how well the ACC system performs and responds to changes in the speed of the lead car, both accelerations and decelerations occur at different time instances. Whenever the host car operates in speed or spacing control can be depicted from the graphs, especially in the velocity and distance graphs.

In the first approximately 55 seconds, the host car operates in speed control as can be observed by the increased distance between the two cars. Both the cars have zero overshoot during the first step response. Once the lead car brakes, which happens after the first 55 seconds, and travels at a lower speed than the driver-set speed the controller switches back to distance control when the actual distance approaches the minimum safe distance, for which it operates in between approximately 60 to 90 seconds. In the step response

(40)

CHAPTER 5 5.1. SIMULATION

from 200 to 150 cm/s, the lead car has an overshoot of 33 % while the host car has an overshoot of 29 %. After 90 seconds the lead car accelerates to a speed of 255 cm/s, which is above the driver-set speed and the distance between the car increases; thus, the host car switches back to speed control and maintains its driver-set speed. In the step response from 150 to 255 cm/s for the lead car, the overshoot is zero. For the host car, which steps back to 200 from 150 cm/s and maintains that speed until approximately 310 seconds, the overshoot is 1 %. After approximately 140 seconds, the lead car brakes from 255 to 150 cm/s for which it travels for a short time and has an overshoot of 70 %. The next acceleration back to 255 cm/s yield an overshoot of zero as well. The last two brakes, from 255 to 200 cm/s and from 200 to 150 cm/s yield an overshoot of 27 % and 33 %, respectively. The last speed decrease for the host car, which occurs at approximately 310 seconds and where the actual distance is equal to the safe distance, the overshoot is 35 %.

(41)

0 50 100 150 200 250 300 350 400

Time [s]

-100 0 100 200

cm/s2

Acceleration

lead host

0 50 100 150 200 250 300 350 400

Time [s]

0 100 200 300

cm/s

Velocity

reference lead host set

0 50 100 150 200 250 300 350 400

Time [s]

0 2000 4000

cm

Distance between the two cars

actual safe

0 50 100 150 200 250 300 350 400

Time [s]

0 2000 4000

cm

Spacing error (actual - safe, greater than zero is desired

space error

Figure 5.1: Simulated results for step responses when the lead car both decreases and increases its velocity. The host car has a constant driver-set velocity and adjusts its velocity when the actual distance has to be adapted to the desired safe distance

(42)

5.1.2 Platoon mode

In platoon mode, the cars travel very closely together at a distance which is shorter than it is when driving in single mode and having a spacing control operating. The reduced distance should be set without compromising the safety. Once communication between the vehicles is established, the lead car can transmit information about its state almost instantaneously to the host car. As a result, the cars can operate at a much closer distance compared to if the host vehicle was only equipped with a radar or lidar. In case of communication loss, the host car disengage immediately from platoon mode and switches back to single mode to guarantee safety.

In the simulation results below, it can be seen how the host car follows the same velocity profile as the preceding car. In Figure 5.2, the results from when the lead car transmits its own reference velocity can be seen. The transmitted reference velocity can be perceived as either that the instant throttle position or brake pedal position is transmitted to the host car. In Figure 5.3, the results from when the lead car transmits its actual velocity can be seen. The main difference between the transmitted information in the two scenarios is that in the first case the reference signal, r is sent to the host car while in the other case the actual velocity, y is sent. The instantaneous errors that occurs right after that the reference velocity is sent is not accounted for when aiming to keep the same velocity as the lead car, therefore the actual velocity is to consider.

Upon initial inspection, the results seems to be identical when considering the velocity profile and the distance between the two cars. However, when investigating the results even further it can be seen that there is a marginal difference in the spacing error. It appears to be smaller when the actual velocity of the lead car is transmitted compared to when the reference velocity is sent. Comparing the overshoots in velocity between the two different cases, shows that the overall performance is better when the actual velocity is transmitted to the host car.

(43)

0 50 100 150 200 250 300 350 400

Time [s]

-100 0 100 200

cm/s2

Acceleration

lead host

0 50 100 150 200 250 300 350 400

Time [s]

0 100 200 300

cm/s

Velocity

0 50 100 150 200 250 300 350 400

Time [s]

0 200 400

cm

actual safe

0 50 100 150 200 250 300 350 400

Time [s]

-50 0 50 100

cm

space error

Figure 5.2: Simulated platooning results for when the reference velocity is sent to the host car

(44)

0 50 100 150 200 250 300 350 400

Time [s]

-100 0 100 200

cm/s2

Acceleration

lead host

0 50 100 150 200 250 300 350 400

Time [s]

0 100 200 300

cm/s

Velocity

0 50 100 150 200 250 300 350 400

Time [s]

0 200 400

cm

actual safe

0 50 100 150 200 250 300 350 400

Time [s]

-100 0 100 200

cm

space error

Figure 5.3: Simulated platooning results for when the actual velocity of the lead car is sent to the host car

Evaluation and Implementation of a Longitudinal Control in a Platoon of Radio Controlled Vehicles

Evaluation and Implementation of a Longitudinal Control in a Platoon of Radio Controlled Vehicles

DANIEL ROSHANGHIAS

Abstract

Sammanfattning

Acknowledgement

Abbreviations

Contents

Chapter 1 Introduction

1.1 Motivation

1.2 Problem statement

1.3 Objectives

1.4 Delimitation

1.4.1 Research questions

1.5 Contribution

1.6 Use Case

1.6.1 Description of the use case

Chapter 2

State of the art

2.1 Related work

2.2 Intelligent Transportation Systems

2.3 Vehicle platooning

2.4 Enabling platooning technologies

2.4.1 Challenges in the control of the platoon

2.5 Platoon System Architecture

Chapter 3

Modeling of the vehicle system

3.1 Longitudinal dynamics

3.2 System identification

3.2.1 Residual Analysis

Chapter 4

Implementation

4.1 Control design

4.2 Controller

Chapter 5 Results

5.1 Simulation

5.1.1 Single mode

5.1.2 Platoon mode