Presentation MPSYS WASP AIwasp-sweden.org/custom/uploads/2018/03/1.3-Lennartson.pdf ·...
Transcript of Presentation MPSYS WASP AIwasp-sweden.org/custom/uploads/2018/03/1.3-Lennartson.pdf ·...
2018-02-27
Systems Control and MechatronicsAI in Chalmers largest master program
Division of Systems and Control, E2
Prof Bengt LennartsonDivision Head: Systems and ControlDepartment of Electrical EngineeringChalmers University of Technology
2018-02-27
Systems and Control (SysCon)IMAGE SAMPLE
Division of Systems and Control, E2 1
• Robotics & Automation Lab• New Autonomous Systems & Collaborative Robotics Lab(5 million SEK investment)
70+ members• Automatic Control• Automation• Mechatronics
Department of Electrical Engineering• Communication and Antenna systems• Signal processing and Biomedical engineering
• Electric Power Engineering• Systems and Control
Aerospace
2018-02-27 Division of Systems and Control, E2 2
Process Industry
Biological Systems
Production Systems Automotive
Power Systems
Embedded Systems
Master program in
Systems, Control and MechatronicsAnalysis and synthesis of complex
computer controlled products and systems
2018-02-27 Division of Systems and Control, E2 3
• Algorithms and Artificial Intelligence
• Autonomous Systems
• Control and Signal Processing
• Electric and Hybrid Powertrains
• Embedded Systems
• Industry 4.0
• Machine Learning
• Mathematical Systems Theory
• Power systems
• Process control
Course packages
2018-02-27 Division of Systems and Control, E2 4
Deep machine learning
Content:• Supervised learning and (some) reinforcement learning
• Components and principles for training neural networks
• Specialized networks for various applications
Overview: Deep neural networks for object detection, speech analysis,machine translation, control, …
Prof Lennart SvenssonSignal processing
2018-02-27 Division of Systems and Control, E2 5
Course introduction
A primer in machine learning
Motivating examples
Course information
Flipped classroom teaching
At home In class
Traditional lectureSolve problems Listen to lecture
[alone] [with teacher]
Flipped classroomWatch videos Solve problems
[alone] [with teacher & peers]
Remarks:
– Material is covered by videos )classes dedicated to active learning!
– Students first meet the materialalone and then analyse it with ateacher and peers.
– We have effectively replacedlectures with two teaching elements. Figure: Bloom’s taxonomy of
learning objectives.
Chalmers University of Technology Deep machine learning Lennart Svensson
Flipped classroom teaching & Adaptive learning
Adaptive learning (Wikipedia)Computerized educational
material is adapted to students' responses to questions, tasks
and experiences.Use machine learning for
adaptive learning!
2018-02-27 Division of Systems and Control, E2 6
• Temporal logic planning and decision making, including learning and adaption.
• Abstraction based state space reduction.
• Model based reinforcement learning (RL) with limited amount of data for Smart Assembly.
• Bridging the gap between formal optimal control and RL.
• Development of RL-based tools, suitable for safety critical applications.
Expert advisor Agent Process
Quality criteria
AI/ML Research