High-level Decisions from a Safe Maneuver Catalog with Reinforcement Learning for Safe and Cooperative Automated Merging

Reinforcement learning (RL) has recently been used for solving challenging\ndecision-making problems in the context of automated driving. However, one of\nthe main drawbacks of the presented RL-based policies is the lack of safety\nguarantees, since they strive to reduce the expected number of collisions but\nstill tolerate them. In this paper, we propose an efficient RL-based\ndecision-making pipeline for safe and cooperative automated driving in merging\nscenarios. The RL agent is able to predict the current situation and provide\nhigh-level decisions, specifying the operation mode of the low level planner\nwhich is responsible for safety. In order to learn a more generic policy, we\npropose a scalable RL architecture for the merging scenario that is not\nsensitive to changes in the environment configurations. According to our\nexperiments, the proposed RL agent can efficiently identify cooperative drivers\nfrom their vehicle state history and generate interactive maneuvers, resulting\nin faster and more comfortable automated driving. At the same time, thanks to\nthe safety constraints inside the planner, all of the maneuvers are collision\nfree and safe.\n