GPU-Accelerated Path-Dependent Marginal Information Gain for Autonomous Exploration

2026-09-30

AI Summary

This study addresses the biased gain estimation and high computational overhead in sampling-based planners for autonomous exploration, which arise from neglecting viewpoint overlap along planned paths. To overcome these limitations, this work proposes a depth-buffer-based method for computing path-dependent marginal information gain. The approach innovatively replaces voxel-based representations with depth buffers to encode historical observations, integrating GPU-accelerated parallel ray casting with a depth-ordered planning tree evaluation algorithm to achieve precise and efficient overlap identification and gain computation. Experimental results demonstrate that the proposed method accelerates computation by up to 118 times while constraining estimation errors within 5–10%. Furthermore, real-world deployments show a 30% reduction in coverage completion time, substantially enhancing overall exploration efficiency.

Abstract

Autonomous exploration demands that robots continuously evaluate candidate viewpoints based on their expected information gain and execution cost. Sampling-based planners estimate this gain by volumetric raycasting and, due to its computational cost, evaluate candidates under an assumption of mutual independence, ignoring the overlap between viewpoints along the same path. This work presents a GPU-accelerated method for computing path-dependent marginal information gain, where instead of storing and merging the observed unknown voxels along each candidate path, previous observations are represented using depth buffers. Candidate rays are projected into the depth buffers of their ancestors to identify observation overlap and exclude regions expected to be observed. The planning tree is evaluated in depth order to maintain the dependency between viewpoints and their optimized yaws, while candidate nodes and rays at each level are processed in parallel on the GPU. The proposed method stays within 5-10% of the exact marginal gain computed using voxel hash maps, with speed-ups of up to 118x on a desktop GPU and 28x on an NVIDIA Jetson Orin NX. The method was integrated into two sampling-based exploration planners and evaluated in three simulation environments, where marginal gain reduced the time to 95% coverage in five of the six evaluated planner-environment combinations. Real-world experiments also showed a 30% reduction in the time to 95% coverage, as well as earlier exploration termination times.

Problem

• Autonomous Exploration
• Information Gain
• Path-Dependent Marginal Gain
• Observation Overlap
• Sampling-Based Planning

Innovation

• GPU acceleration
• marginal information gain
• depth buffer
• autonomous exploration
• path-dependent planning