RNBF: Real-Time RGB-D Based Neural Barrier Functions for Safe Robotic Navigation

2025-05-04

AI Summary

Safe robot navigation in unknown, unstructured environments using low-cost RGB-D sensors remains challenging due to depth noise, lack of obstacle priors, and real-time requirements. Method: This paper proposes an online neural signed distance field (SDF) modeling framework that requires no pretraining and no prior geometric knowledge of obstacles. It jointly incorporates depth noise modeling, online implicit surface optimization, and a first-order differentiable SDF representation to construct a continuous, differentiable, and gradient-stable neural SDF. This SDF is integrated into a neural barrier function (NBF)-based control framework, ensuring strict controller compatibility and formal safety guarantees. Results: Evaluated in simulation and on a physical Fetch robot, the system achieves millisecond-level response, zero collisions, and significantly enhanced robustness in dynamic, cluttered environments. The key contribution is the first realization of noise-resilient, online, pretraining-free neural SDF modeling—eliminating reliance on handcrafted or learned geometric priors of obstacles.

Technology Category

• Natural Language Processing: Safety and Robustness
• Intelligent Robots: Localization, Mapping, and Navigation
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Autonomous safe navigation in unstructured and novel environments poses significant challenges, especially when environment information can only be provided through low-cost vision sensors. Although safe reactive approaches have been proposed to ensure robot safety in complex environments, many base their theory off the assumption that the robot has prior knowledge on obstacle locations and geometries. In this paper, we present a real-time, vision-based framework that constructs continuous, first-order differentiable Signed Distance Fields (SDFs) of unknown environments without any pre-training. Our proposed method ensures full compatibility with established SDF-based reactive controllers. To achieve robust performance under practical sensing conditions, our approach explicitly accounts for noise in affordable RGB-D cameras, refining the neural SDF representation online for smoother geometry and stable gradient estimates. We validate the proposed method in simulation and real-world experiments using a Fetch robot.

Problem

• Safe robotic navigation in unknown environments using RGB-D sensors
• Constructing real-time differentiable Signed Distance Fields without pre-training
• Handling noise in affordable RGB-D cameras for stable gradient estimates

Innovation

• Real-time RGB-D based neural barrier functions
• Continuous differentiable SDFs without pre-training
• Online refinement for noise-resistant performance