Control and Perception in Networked and Autonomous Vehicles

Instructors

Dr.-Ing. Bassam Alrifaee
Lecturer

Patrick Scheffe, M. Sc.
Assistant Lecturer

Key Facts

The Cyber-Physical Mobility Group at the Chair of Embedded Software of RWTH Aachen University offers this course.

We offered this course for the very first time in the winter term 2019/2020.

Due to lab capacity, currently we are able to host only 30 students per class. We grasped this opportunity to provide a learning experience, which is tailored towards the participants and builds on interactions between lecturer and participants.

Students get six credit points for this course.

This course follows the concept ‘Method. Application. Experiment.’. The course teaches methods of networked control, which can be applied to systems consisting of multiple vehicles. The methods are applied in experiments, which results in a high learning factor.

We designed this course for students in the master programs of Computer Science, Automation Engineering, and Computational Engineering Science. The students do lab exercises in teams of two students from two different study programs.

There are no special or necessary requirements to take this course. We start the course with a survey, which we call diagnostic test. The goal of this test is to know the students’ pre-knowledge in various topics of the course, which allows the lecturer to adapt the content and the pace. Click here to go to the test.

Course Description

This course combines theory of multi-agent decision-making with practical exercises in the Cyber-Physical Mobility Lab (CPM Lab, an open source platform for Networked and Autonomous Vehicles).

In the theory part, we focus on distributing the control problem of multi-agent systems coupled via objective function or constraints using Distributed MPC (DMPC). Additionally, the course discusses the application in a multi-vehicle system, which prepares the participants for the practical lab work.

In the CPM Lab, the students implement controllers for vehicle platoons. The CPM Lab supports rapid functional prototyping. It consists of 20 model-scale vehicles for experiments and a simulation environment. Software developed in simulations can be seamlessly transferred to experiments without any adaptions. Additionally, experiments with the 20 vehicles can be extended by unlimited additional simulated vehicles. The CPM Lab allows researchers as well as students from different disciplines to see their ideas develop into reality. The CPM Lab is completely designed and developed by the Cyber-Physical Mobility Group at RWTH Aachen University.

If you cannot access the CPM Lab physically, stay tuned to our remote-access, which is under construction and expected to go online in Oct. 2021.

The course covers the following topics

Vehicle models: mass-point, kinematic, and kinetic
Control engineering and optimization: model predictive control, convex optimization, non-convex optimization
Network and distribution: graph theory, distributed algorithms, cooperative and non-cooperative approaches to networked control
Machine perception: neural networks
Software architectures and testing concepts: service-oriented software architecture, In-the-loop testing concepts for networked systems
Case study: vehicle platoon

We recommend the following literature

R. Rajamani: Vehicle Dynamics and Control. Springer, 2012. DOI

S. Boyd and L. Vandenberghe. Convex Optimization. Cambridge University Press, 2009. pdf

F. Borrelli, A. Bemporad, and M. Morari. Predictive Control for Linear and Hybrid Systems, Cambridge University Press, 2017. pdf

J. Maciejowski. Predictive Control with Constraints. Prentice Hall, 2002.

B. Alrifaee. Networked Model Predictive Control for Vehicle Collision Avoidance. 2017. pdf

Learning Objectives

Knowledge and Understanding

After successful participation in the course, the students

know methods for the development of control and perception algorithms in networked and autonomous vehicles
know the topics relevant to the implementation of control and perception algorithms in networked and autonomous vehicles
understand different methods for modelling, control, optimization, perception, distribution, design and testing in networked and autonomous vehicles and are able to name their fields of application
know the challenges of real-time algorithm implementation for networked and autonomous vehicles
know different approaches and process models for the development of control and perception algorithms in networked and autonomous vehicles as well as their advantages and disadvantages

Skills and Competences

After successful participation in the course, students are able to independently perform the necessary steps for the successful development of control and perception algorithms in networked and autonomous vehicles. In doing so, they independently take into account the different aspects of the development and are able to evaluate to what extent the available approaches, methods and algorithms are applicable. They are also able to synthesize different control and perception algorithms. Furthermore, they can consider practical aspects by testing in the lab.

Lecture Style

The lectures combine the following teaching methods

Classical presentation: We liven it up with many examples and games to introduce new topics
Group discussions: we divide the students into five groups with six members each. The lecturer gives an initial presentation of a topic; the groups discuss the topic for ten minutes and a member of each group presents the results
Flipped classroom: the lecturer introduces a topic; the students prepare it by reading literature and watching videos given by the lecturer. In the next week, a student or two present it before we go to discussions. We offer the flipped classroom five times to give each group the opportunity to present
Practical exercises: we offer optional programming tasks using MATLAB to deepen the understanding of some topics
A case study at the end of the lecture prepares the students for the lab work

Each part of the lecture starts with mapping it to the lab architecture.

Lab Style

The students do lab exercises in teams of two students from two different study programs. Each team gets four complete days in the lab.

Exam

The exams are oral; 20 minutes per student. The exam starts with a 5 minutes presentation (unified template) about the lab results, following by 15 minutes questions on lecture content and lab work.

Evaluation

The students evaluated the lecture of this course on average with 1.6 in the winter term 2019/2020 where 1.8 is the average of the department and 1.8 in the winter term 2020/2021 where 1.8 is the average of the department.

The students evaluated the lab exercises of this course on average with 1.8 in the winter term 2019/2020 and 2.0 in the winter term 2020/2021.

The evaluation scale is from 1.0 (best) to 5.0 (worst).

Statistics

Course Materials

Here are the course materials (lecture slides, exercise sheets, lab tasks, and lecture videos) of the winter term 2020/2021. They are in continuous development. We will offer the course in the winter term 2021/2022 again and will publish the course materials here after accomplishment.