Dynamic Interaction-Aware Scene Understanding for Reinforcement Learning in Autonomous Driving

TL;DR

This paper introduces Deep Scenes, a novel architecture for autonomous driving that models complex interactions among traffic participants using deep sets or graph convolutional networks, improving decision-making in reinforcement learning.

Contribution

The work proposes the Deep Scenes architecture and two new algorithms, Graph-Q and DeepScene-Q, for interaction-aware scene understanding in autonomous driving reinforcement learning.

Findings

Outperforms state-of-the-art methods in SUMO simulator

Effectively models interactions between traffic participants

Enhances high-level decision making in autonomous driving

Abstract

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.