Safe Beyond the Horizon: Efficient Sampling-based MPC with Neural Control Barrier Functions

2025-02-20

AI Summary

Addressing the critical challenge of ensuring long-term safety for sampling-based model predictive control (MPC) in black-box nonlinear systems under limited prediction horizons, this paper proposes a novel framework integrating neural control barrier functions (Neural CBFs) with variational inference MPC (VIMPC). It is the first work to embed Neural CBFs into sampling-based MPC, leveraging constraint-aware importance sampling and low-variance resampling to achieve rigorous, horizon-wide safety guarantees—albeit with mild relaxation of recursive feasibility. The method enables real-time deployment (sub-millisecond latency on CPU) and exhibits robustness to cost function design: safety is preserved even when the cost function is poorly specified. Comprehensive simulations and real-world hardware experiments demonstrate substantial improvements in both safety assurance and computational efficiency.

Technology Category

• Search and Optimization: Sampling/Simulation-based Search
• Natural Language Processing: Safety and Robustness
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

A common problem when using model predictive control (MPC) in practice is the satisfaction of safety specifications beyond the prediction horizon. While theoretical works have shown that safety can be guaranteed by enforcing a suitable terminal set constraint or a sufficiently long prediction horizon, these techniques are difficult to apply and thus are rarely used by practitioners, especially in the case of general nonlinear dynamics. To solve this problem, we impose a tradeoff between exact recursive feasibility, computational tractability, and applicability to ''black-box'' dynamics by learning an approximate discrete-time control barrier function and incorporating it into a variational inference MPC (VIMPC), a sampling-based MPC paradigm. To handle the resulting state constraints, we further propose a new sampling strategy that greatly reduces the variance of the estimated optimal control, improving the sample efficiency, and enabling real-time planning on a CPU. The resulting Neural Shield-VIMPC (NS-VIMPC) controller yields substantial safety improvements compared to existing sampling-based MPC controllers, even under badly designed cost functions. We validate our approach in both simulation and real-world hardware experiments.

Problem

• Ensuring safety beyond prediction horizon in MPC
• Balancing feasibility, computational tractability, and applicability
• Improving sample efficiency and real-time planning capability

Innovation

• Neural Control Barrier Functions
• Variational Inference MPC
• Efficient Sampling Strategy