Inverse Reinforcement Learning for Minimum-Exposure Paths in Spatiotemporally Varying Scalar Fields

TL;DR

The paper addresses generating datasets of minimum-threat AV paths in static and time-varying threat fields using inverse reinforcement learning (IRL). It formulates the navigation task as a discrete MDP and learns a linear reward $r(s,u,s') = w_1 c(x_{s'}, t_{s'}) + w_2 \|x_{s'} - x_{bar}\|$ while training a deep Q-learning loop to synthesize paths; the training aligns feature expectations $\mu_{\mathcal{Z}}$ with those of generated paths, starting from $w^0 = \mu_{\mathcal{Z}}$. Experiments show good agreement with true optimal costs in static and dynamic environments and demonstrate generalization to unseen threat fields, as well as the ability to discriminate outputs across different training datasets. This IRL-based data synthesis offers a transparent, data-efficient means to augment AV performance analyses with plausible trajectories and is accompanied by publicly available code.

Abstract

Performance and reliability analyses of autonomous vehicles (AVs) can benefit from tools that ``amplify'' small datasets to synthesize larger volumes of plausible samples of the AV's behavior. We consider a specific instance of this data synthesis problem that addresses minimizing the AV's exposure to adverse environmental conditions during travel to a fixed goal location. The environment is characterized by a threat field, which is a strictly positive scalar field with higher intensities corresponding to hazardous and unfavorable conditions for the AV. We address the problem of synthesizing datasets of minimum exposure paths that resemble a training dataset of such paths. The main contribution of this paper is an inverse reinforcement learning (IRL) model to solve this problem. We consider time-invariant (static) as well as time-varying (dynamic) threat fields. We find that the proposed IRL model provides excellent performance in synthesizing paths from initial conditions not seen in the training dataset, when the threat field is the same as that used for training. Furthermore, we evaluate model performance on unseen threat fields and find low error in that case as well. Finally, we demonstrate the model's ability to synthesize distinct datasets when trained on different datasets with distinct characteristics.

Figures (12)

• Figure 1: Schematic illustration of IRL.
• Figure 2: Illustration of the IRL and DQL training processes.
• Figure 3: Sample static threat field.
• Figure 4: Examples of minimum threat exposure paths in the training dataset $\mathcal{Z}$ for the static case.
• Figure 5: Sample evolution of a dynamic threat field.