Planning on the fast lane: Learning to interact using attention mechanisms in path integral inverse reinforcement learning

Abstract

General-purpose trajectory planning algorithms for automated driving utilize complex reward functions to perform a combined optimization of strategic, behavioral, and kinematic features. The specification and tuning of a single reward function is a tedious task and does not generalize over a large set of traffic situations. Deep learning approaches based on path integral inverse reinforcement learning have been successfully applied to predict local situation-dependent reward functions using features of a set of sampled driving policies. Sample-based trajectory planning algorithms are able to approximate a spatio-temporal subspace of feasible driving policies that can be used to encode the context of a situation. However, the interaction with dynamic objects requires an extended planning horizon, which requires sequential context modeling. In this work, we are concerned with the sequential reward prediction over an extended time horizon. We present a neural network architecture that uses a policy attention mechanism to generate a low-dimensional context vector by concentrating on trajectories with a human-like driving style. Besides, we propose a temporal attention mechanism to identify context switches and allow for stable adaptation of rewards. We evaluate our results on complex simulated driving situations, including other vehicles. Our evaluation shows that our policy attention mechanisms learns to focus on collision free policies in the configuration space. Furthermore, the temporal attention mechanism learns persistent interaction with other vehicles over an extended planning horizon.

AB-Mapper: Attention and BicNet Based Multi-agent Path Finding for Dynamic Crowded Environment

“…In contrast to the multi-headed attention mechanism approach based on graph neural networks, some algorithms do not model the agents as graphs. For example, Rosbach et al [24] applied the attention mechanism to inverse reinforcement learning for predicting the reward function over an extended planning horizon. Shah et al [25] proposed a novel linguistic instruction attention mechanism, where the attention mechanism scored the tokens of the input visual and instruction information.…”

Interaction-Aware Trajectory Prediction and Planning for Autonomous Vehicles in Forced Merge Scenarios

“…An RL-based policy can account for the vehicle interactions in certain scenarios through training in an environment capable of representing such interactions [14], [15]. In order to obtain through RL driving policies that behave like human drivers, several researchers chose to use inverse RL to estimate human's reward function for driving [16], [17], [18], [19]. To be able to model different human driver styles and/or interaction intentions, [20] incorporates cooperativeness into the intelligent driver model and [21] formulates different reward functions for different drivers and performs RL based on the models.…”