Neural NMPC through Signed Distance Field Encoding for Collision Avoidance

2025-11-26

AI Summary

This paper addresses mapless, collision-free navigation for aerial robots in unknown environments using only onboard ranging sensors. Method: We propose a neural-augmented nonlinear model predictive control (Neural NMPC) framework featuring a two-stage neural architecture: a convolutional encoder implicitly maps single-frame range images to a signed distance field (SDF), and an MLP embeds this representation into NMPC position constraints—ensuring recursive feasibility and closed-loop stability. The framework requires no explicit mapping or external localization, operating directly on raw range measurements for real-time, velocity-level obstacle avoidance. Contribution/Results: Evaluated in both simulation and real-world forest environments, our approach significantly outperforms state-of-the-art local navigators and demonstrates strong robustness against odometry drift.

Technology Category

• Intelligent Robots: Localization, Mapping, and Navigation
• Planning, Routing, and Scheduling: Planning with Language Models
• Natural Language Processing: Safety and Robustness

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Agentic search
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

This paper introduces a neural Nonlinear Model Predictive Control (NMPC) framework for mapless, collision-free navigation in unknown environments with Aerial Robots, using onboard range sensing. We leverage deep neural networks to encode a single range image, capturing all the available information about the environment, into a Signed Distance Function (SDF). The proposed neural architecture consists of two cascaded networks: a convolutional encoder that compresses the input image into a low-dimensional latent vector, and a Multi-Layer Perceptron that approximates the corresponding spatial SDF. This latter network parametrizes an explicit position constraint used for collision avoidance, which is embedded in a velocity-tracking NMPC that outputs thrust and attitude commands to the robot. First, a theoretical analysis of the contributed NMPC is conducted, verifying recursive feasibility and stability properties under fixed observations. Subsequently, we evaluate the open-loop performance of the learning-based components as well as the closed-loop performance of the controller in simulations and experiments. The simulation study includes an ablation study, comparisons with two state-of-the-art local navigation methods, and an assessment of the resilience to drifting odometry. The real-world experiments are conducted in forest environments, demonstrating that the neural NMPC effectively performs collision avoidance in cluttered settings against an adversarial reference velocity input and drifting position estimates.

Problem

• Develops neural NMPC for mapless collision avoidance in unknown environments
• Encodes range images into Signed Distance Functions using deep networks
• Enables aerial robots to navigate cluttered settings with drifting odometry

Innovation

• Neural NMPC framework enables mapless collision avoidance navigation
• Signed Distance Function encoded from range images using deep networks
• Velocity-tracking NMPC embeds SDF constraints for thrust control