Perceive with confidence: Statistical safety assurances for navigation with learning-based perception

Abstract

Rapid advances in perception have enabled large pre-trained models to be used out of the box for transforming high-dimensional, noisy, and partial observations of the world into rich occupancy representations. However, the reliability of these models and consequently their safe integration onto robots remains unknown, particularly when deployed in environments unseen during training. To provide safety guarantees, we rigorously quantify the uncertainty of pre-trained perception systems for object detection and scene completion via a novel calibration technique based on conformal prediction. Crucially, this procedure guarantees robustness to distribution shifts in states when perception outputs are used in conjunction with a planner. As a result, the calibrated perception system can be used in combination with any safe planner to provide an end-to-end statistical assurance on safety in unseen environments. We evaluate the resulting approach, Perceive with Confidence (PwC), in simulation and on hardware where a quadruped robot navigates through previously unseen static indoor environments. These experiments validate the safety assurances for obstacle avoidance provided by PwC. In our simulation experiments, our method reduces obstacle misdetection significantly compared to uncalibrated perception models. While misdetections lead to collisions for baseline methods, our approach remains safe. We further demonstrate reducing the conservatism of our method without sacrificing safety, outperforming all baselines in success rates in challenging environments. In hardware experiments on a quadruped robot, our method improves empirical safety and obstacle misdetection by significant margins over the baselines, highlighting our approach’s robustness under more demanding conditions.