Angel Molina Acosta, Elektroteknik

Tite​l: Vehicle-in-the-loop validation of autonomous cars
A framework for modelling, analysis, and control of test-scenarios


​Anslut till seminariet från PC, Mac. Linux, iOS elelr Android via zoom

Maila till phdadm.e2@chalmers.se i god tid innan seminariet för att få lösenordet

The seminar can be accessed through Zoom, and it will open shortly before 10:00. We would kindly ask you to keep the video off and mute the microphone during the seminar. At the end of the session there will be an opportunity to ask questions through Zoom. In case there will be any updates about the event, these will be posted on this website.

Angel Molina Acosta är doktorand i forskargruppen mekatronik, Avdelningen för System- och reglerteknik
Diskussionsledare är docent  Jana Tumova, KTH, Stockholm
Examinator är professor Jonas Sjöberg, Avdelningen för System- och reglerteknik
Huvudhandledare är docent Paolo Falcone, Avdelningen för System- och reglerteknik

Sammanfattning
Validation of autonomous driving (AD) cars is a difficult task because of the complexity that results from the integration of multiple systems and the variety of operating conditions. To this end, testing with real vehicles is crucial to ensure a thorough validation of AD cars. However, testing AD vehicles in public roads is not viable in early stages of the development cycle. An alternative is to conduct tests in controlled environments, such as proving grounds.

This thesis proposes a framework for modelling, analysis, and control of tests-scenarios for validation of autonomous cars by exposing the vehicle-under-test to a traffic scenario at a test track, where mobile test-targets represent other road users. The framework is suitable for leader-follower, multi-agent systems where the motion of the followers should be coordinated with the motion of an externally controlled leader. Scenarios are modelled as switched systems. The feasibility of the scenario is investigated using backward reachability analysis. A constrained optimal control problem is formulated to control the state of the multi-agent system through a sequence of goal sets. Simulation results illustrate the usefulness of the framework.
A second contribution in this thesis is a novel method for decentralized computation of backward reachable sets and robust control invariant sets. The method is applicable to large-scale systems arising from the interconnection of multiple subsystems with linear dynamics. Polyhedral constraints and additive disturbances are considered. Compared to the standard centralized procedure for computation of control invariant sets, the proposed method is more efficient for large-scale systems where the coupling among the subsystems is sparse.


Kategori Licentiatseminarium
Plats: EA, lecture hall, Hörsalsvägen 11, EDIT trappa C, D och H
Tid: 2022-06-13 10:00
Sluttid: 2022-06-13 12:00

Sidansvarig Publicerad: ti 17 maj 2022.