Autonomous Systems is a WASP core course whose purpose is to give a broad understanding of the wide area of autonomous systems and the foundational knowledge in the topic areas required to understand and develop autonomous systems.
The course consists of 4 modules where each module corresponds to 3hp:
- Control and Decision making
- Sensing and Perception
- Learning and Knowledge
- Interaction and Collaboration
After the course, the student will be able to:
- Explain what autonomy is and what makes a system autonomous.
- Explain how an autonomous system works and describe important components.
- Analyze how different components contribute to the autonomy of a system.
- Explain and apply common techniques and the theory behind them in the four topic areas Control and Decision making, Sensing and Perception, Learning and Knowledge and Interaction and Collaboration.
- Analyze the industrial and societal impact of autonomous systems.
- Analyze the ethical and moral issues related to autonomous systems.
In the area of control and decision making, after the course the student will be able to:
- Explain important model classes, key system properties, and basic control strategies.
- Explain optimal control and model predictive control and how they relate to autonomous systems.
- Explain common approaches to classical task planning and relevant extensions to classical planning.
- Analyze the relations between and integration of control, motion planning, task planning and decision making.
In the area of sensing and perception, after the course the student will be able to:
- Explain common types and models of sensors (camera types and geometric camera models).
- Analyze the choice and placement of sensors (triangulation e.g. stereo, Kinect and observability e.g. in multi-camera systems).
- Explain common methods for visual object detection and tracking (background modelling, appearance-based detectors, region tracking, object tracking).
- Explain common methods for visual recognition of objects and actions (recognition pipelines, feature selection, connection to learning).
- Explain common approaches to filtering and data association (outlier rejection of detections, PnP).
- Explain common techniques for sensor fusion including multi-modal input and prior information (bundle adjustment).
- Explain common types of positioning systems (e.g. GNSS, UWB, dead reckoning, inertial navigation, signal opportunity, magnetic field SLAM), their strengths and weaknesses as well as their characteristics in terms of accuracy, integrity, accessibility and coverage.
- Analyze the effect of measurement errors on the accuracy and dependability of positioning systems.
- Explain how motion models and motion conditions can be used to improve the performance of positioning systems and allow for simpler and cheaper sensors to be used.
- Explain common methods and techniques for mapping and situation awareness (visual SLAM, structure from motion).
In the area of learning and knowledge, after the course the student will be able to:
- Explain the strengths and weaknesses of supervised and unsupervised learning.
- Explain common methods and techniques for classification and regression.
- Explain common methods and techniques for reinforcement learning, especially for autonomous systems.
- Explain common methods and techniques for neural networks and deep learning.
- Explain common methods and techniques for knowledge-based systems including the use of ontologies.
- Explain common methods and techniques for representing and handling uncertainty.
The students are expected to have a background in computer science, computer engineering, electrical engineering or similar. The students are expected to have the foundational mathematics found in most engineering programs and basic programming skills.
The course is organized around five two day sessions with physical meetings:
- Control and Decision Making, September 7-8 (Wed-Thu), Linköping
- Sensing and Perception, October 5-6 (Wed-Thu) Linköping
- Learning and Knowledge, October 27-28 (Thu-Fri) Stockholm
- Interaction and Collaboration, November 14-15 (Mon-Tue) Stockholm
- Final examination, December 12-13 (Mon-Tue) Göteborg
Each session starts at 10.30 the first day and ends at 15.00 the second day and consists of lectures, invited talks and seminars. The main content of each module is presented at a session and then examined at a seminar at the following session. The first session will also give an introduction to autonomous systems and autonomy.
Between the sessions there will be local activities at the four main sites (Göteborg, Linköping, Lund and Stockholm). These will be mainly student driven.
The course literature is research papers and tutorials associated with each topic area.
- Control and Decision Making:
- Torkel Glad and Lennart Ljung. Control Theory – Multivariable and Nonlinear Methods. Taylor & Francis,2000.
- Torkel Glad and Lennart Ljung. Reglerteori – Flervariabla och olinjära metoder. Studentlitteratur, 2003.
- IJCAI2016 Tutorial: Introduction to Planning Models and Methods
- IJCAI2016 Tutorial: Deliberative Planning and Acting
- 2014. HDRC3 – A Distributed Hybrid Deliberative/Reactive Architecture for Unmanned Aircraft Systems. In Kimon P. Valavanis, George J. Vachtsevanos, editors, Handbook of Unmanned Aerial Vehicles, pages 849–952. Springer Science+Business Media B.V. DOI: 10.1007/978-90-481-9707-1_118. .
The examination consists of exercises, labs and projects. There is also a final examination in the form of a two-day seminar at the end of the course (officially part of PROJ1). The deadline for all the exercises and the labs are at the end of the course, i.e. December 13. The deadline for each challenges is the session after. You can get credits for each of the three parts (UPG1, LAB1, PROJ1) individually.
- UPG1 4hp: For each module there are exercises related to that module that should be solved individually. These are mainly examined locally. To pass you have to solve more than half of each of the exercises. Each file is a separate exercise.
- Control and Decision Making: control exercises, planning exercises
- Sensing and Perception: vision exercises (view0, view1), localization exercises (Matlab files), slam exercises (optional)
- Learning and Knowledge: learning exercises, knowledge representation exercises
- Interaction and Collaboration: human robot interaction exercises, task allocation exercises
- LAB1 4hp: For each module there are practical lab assignments related to that module that should be completed individually or in pairs. These are mainly examined locally. To pass you have to solve more than half of each of the labs. Each file is a separate lab.
- PROJ1 4hp: For each module there are challenges related to autonomous systems that should be completed in groups of 4-8 students. These challenges will involve programming and integrating components into working autonomous systems. The challenges are examined at the sessions.