Dynamic Interaction-Aware Scene Understanding for Reinforcement Learning in Autonomous Driving
This work addresses the problem of improving high-level decision-making in autonomous driving systems, offering a novel approach to scene understanding that is interaction-aware, though it appears incremental as it builds on existing methods like Deep Sets and Graph Convolutional Networks.
The paper tackles the challenge of enabling deep reinforcement learning for autonomous driving decision-making by developing architectures that handle variable-length object sequences and capture interactions between traffic participants. The proposed Deep Scenes architecture, implemented as Graph-Q and DeepScene-Q algorithms, outperforms state-of-the-art methods in evaluations using the SUMO traffic simulator.
The common pipeline in autonomous driving systems is highly modular and includes a perception component which extracts lists of surrounding objects and passes these lists to a high-level decision component. In this case, leveraging the benefits of deep reinforcement learning for high-level decision making requires special architectures to deal with multiple variable-length sequences of different object types, such as vehicles, lanes or traffic signs. At the same time, the architecture has to be able to cover interactions between traffic participants in order to find the optimal action to be taken. In this work, we propose the novel Deep Scenes architecture, that can learn complex interaction-aware scene representations based on extensions of either 1) Deep Sets or 2) Graph Convolutional Networks. We present the Graph-Q and DeepScene-Q off-policy reinforcement learning algorithms, both outperforming state-of-the-art methods in evaluations with the publicly available traffic simulator SUMO.