Residual Neural Terminal Constraint for MPC-based Collision Avoidance in Dynamic Environments

2025-08-05

AI Summary

Real-time obstacle avoidance in dynamic environments using Model Predictive Control (MPC) struggles to guarantee safety due to the computational intractability of Hamilton–Jacobi (HJ) reachability analysis for online evaluation of safety constraints. Method: We propose a hybrid MPC planner that decomposes the value function of a time-varying safe set into a signed distance function (SDF) and a non-negative residual term. A lightweight neural network models the residual, while a hypernetwork enhances cross-scenario generalization. This residual model serves as a verifiable, conservative lower bound on the true value function and is embedded as a safety-critical terminal constraint in MPC. Contribution/Results: The approach ensures theoretical safety guarantees without sacrificing real-time performance. In both simulation and physical experiments, it achieves a 30% higher success rate than the optimal baseline, with comparable computational overhead and shorter trajectory execution times.

Technology Category

• Natural Language Processing: Safety and Robustness
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Planning, Routing, and Scheduling: Replanning and Plan Repair

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Security and Privacy: Large-scale security measurements

Abstract

In this paper, we propose a hybrid MPC local planner that uses a learning-based approximation of a time-varying safe set, derived from local observations and applied as the MPC terminal constraint. This set can be represented as a zero-superlevel set of the value function computed via Hamilton-Jacobi (HJ) reachability analysis, which is infeasible in real-time. We exploit the property that the HJ value function can be expressed as a difference of the corresponding signed distance function (SDF) and a non-negative residual function. The residual component is modeled as a neural network with non-negative output and subtracted from the computed SDF, resulting in a real-time value function estimate that is at least as safe as the SDF by design. Additionally, we parametrize the neural residual by a hypernetwork to improve real-time performance and generalization properties. The proposed method is compared with three state-of-the-art methods in simulations and hardware experiments, achieving up to 30% higher success rates compared to the best baseline while requiring a similar computational effort and producing high-quality (low travel-time) solutions.

Problem

• Real-time collision avoidance in dynamic environments
• Learning-based approximation of safe sets for MPC
• Neural network modeling for real-time value function estimation

Innovation

• Hybrid MPC planner with learning-based safe set
• Neural network residual for real-time safety
• Hypernetwork parametrization enhances performance