Semantically-driven Deep Reinforcement Learning for Inspection Path Planning

TL;DR

This paper presents a semantics-aware deep reinforcement learning approach for autonomous inspection path planning that combines semantic object inspection with collision-free navigation, demonstrating robustness and real-world applicability.

Contribution

The paper introduces an end-to-end semantics-driven inspection policy that integrates semantic understanding with navigation using only limited sensory inputs, bridging the sim2real gap.

Findings

Robust generalization in unknown environments

Successful deployment on a flying robot in real-world scenarios

Effective semantic object inspection combined with navigation

Abstract

This paper introduces a novel semantics-aware inspection planning policy derived through deep reinforcement learning. Reflecting the fact that within autonomous informative path planning missions in unknown environments, it is often only a sparse set of objects of interest that need to be inspected, the method contributes an end-to-end policy that simultaneously performs semantic object visual inspection combined with collision-free navigation. Assuming access only to the instantaneous depth map, the associated segmentation image, the ego-centric local occupancy, and the history of past positions in the robot's neighborhood, the method demonstrates robust generalizability and successful crossing of the sim2real gap. Beyond simulations and extensive comparison studies, the approach is verified in experimental evaluations onboard a flying robot deployed in novel environments with previously unseen semantics and overall geometric configurations.