Generation of naturalistic and critical boundary scenarios: A bi-level adaptive deep reinforcement learning method.

The complexities of real-world driving environments, coupled with a limited availability of naturalistic and critical test scenarios, have long hindered unbiased and effective comprehensive performance evaluations. In this work, we propose a bi-level…

Preview only. Read the full abstract at the source