Learning Vision-based Pursuit-Evasion Robot Policies

TL;DR

This paper introduces a supervised learning approach for vision-based pursuit-evasion robot policies, enabling a quadruped robot to effectively pursue evaders in real-world scenarios by leveraging a fully-observable policy for supervision.

Contribution

It transforms pursuit-evasion into a supervised learning problem, highlighting the importance of supervision quality factors like diversity, optimality, and modeling assumptions.

Findings

Successful deployment on a quadruped robot in real-world pursuit-evasion tasks.

The robot effectively gathers information, predicts intent, and intercepts despite sensing constraints.

Supervised learning approach improves pursuit policy robustness in noisy environments.

Abstract

Learning strategic robot behavior -- like that required in pursuit-evasion interactions -- under real-world constraints is extremely challenging. It requires exploiting the dynamics of the interaction, and planning through both physical state and latent intent uncertainty. In this paper, we transform this intractable problem into a supervised learning problem, where a fully-observable robot policy generates supervision for a partially-observable one. We find that the quality of the supervision signal for the partially-observable pursuer policy depends on two key factors: the balance of diversity and optimality of the evader's behavior and the strength of the modeling assumptions in the fully-observable policy. We deploy our policy on a physical quadruped robot with an RGB-D camera on pursuit-evasion interactions in the wild. Despite all the challenges, the sensing constraints bring about creativity: the robot is pushed to gather information when uncertain, predict intent from noisy measurements, and anticipate in order to intercept. Project webpage: https://abajcsy.github.io/vision-based-pursuit/