Sight Over Site: Perception-Aware Reinforcement Learning for Efficient Robotic Inspection
Richard Kuhlmann, Jakob Wolfram, Boyang Sun, Jiaxu Xing, Davide Scaramuzza, Marc Pollefeys, Cesar Cadena
AI summary
Problem
Traditional robotic inspection reduces the task to point-goal navigation, forcing robots to reach a target's physical location even when visual access is achieved earlier, resulting in inefficient and redundant movement.
Approach
The authors train a map-free, end-to-end reinforcement learning policy that uses egocentric depth and relative target pose to directly optimize for finding the shortest viewpoint that guarantees visual contact with a target.
Key results
- Outperforms state-of-the-art RL navigation baselines in trajectory efficiency
- Successfully deployed on a real-world Boston Dynamics Spot quadrupedal robot
- Introduces a ground-truth shortest inspection path algorithm for rigorous benchmarking
- Demonstrates emergent distance-aware balancing between navigation and inspection
Why it matters
Provides a practical, map-free framework for faster and more energy-efficient autonomous inspection in industrial, search-and-rescue, and service robotics applications.
Abstract
Autonomous inspection is a central problem in robotics, with applications ranging from industrial monitoring to search-and-rescue. Traditionally, inspection has often been reduced to navigation tasks, where the objective is to reach a predefined location while avoiding obstacles. However, this formulation captures only part of the real inspection problem. In real-world environments, the inspection targets may become visible well before their exact coordinates are reached, making further movement both redundant and inefficient. What matters more for inspection is not simply arriving at the target’s position, but positioning the robot at a viewpoint from which the target becomes observable. In this work, we revisit inspection from a perception-aware perspective. We propose an end-to-end reinforcement learning framework that explicitly incorporates target visibility as the primary objective, enabling the robot to find the shortest trajectory that guarantees visual contact with the target without relying on a map. The learned policy leverages both perceptual and proprioceptive sensing and is trained entirely in simulation, before being deployed to a real- world robot. We further develop an algorithm to compute ground-truth shortest inspection paths, which provides a ref- erence for evaluation. Through extensive experiments, we show that our method outperforms existing classical and learning- based navigation approaches, yielding more efficient inspection trajectories in both simulated and real-world settings. The project is available at https://sight-over-site.github.io/