Safe collaborative behavior planning demonstrated in mixed-traffic on-road scenarios
The focus is on nominal traffic events and thus complementary to the solution strategy and the safeguarding strategy for unexpected situations developed in KoSi. Unexpected situations are not expected to be the normal case for an autonomous vehicle and should only occur singularly. In this nominal case, society expects autonomous vehicles not only to obey the law, but also to integrate optimally into traffic. In this case, no overly passive behavior may be exhibited ("rolling traffic obstacle") and one's own strategic goals must still be pursued with the vigor necessary in urban road traffic.
Our game-theoretical approach inherently considers mixed-autonomous traffic and is demonstrated and evaluated in real-time on the test vehicle.
As a basic prerequisite for the definition of a multi-agent game, the scope and intended use case of the algorithms must be clearly defined and delimited (definition of the Operational Design Domain ODD). From this, we derive which restrictions of the nominal behavior are expected and how unexpected situations can occur despite cooperative planning.
01.02.2021 - 31.08.2023