Low-pass sampling in Model Predictive Path Integral Control

2025-03-13

AI Summary

MPPI control suffers from high-frequency noise in sampled trajectories, causing actuator wear and inefficient exploration. To address this, we propose a low-pass filter-embedded sampling mechanism that explicitly incorporates tunable, interpretable low-pass frequency constraints directly into the MPPI control law generation—enabling, for the first time, direct spectral shaping of control trajectories while preserving real-time performance and physical feasibility. Unlike post-hoc smoothing, our method suppresses deleterious high-frequency components at the source, significantly improving trajectory smoothness and exploration quality. Experiments across Gymnasium simulation, quadruped robot simulation, and the F1TENTH real-world platform demonstrate: 32–57% reduction in control jitter, 18.4% average improvement in task success rate, and 41% decrease in acceleration L²-norm (a key smoothness metric). The proposed approach consistently outperforms existing MPPI variants in both stability and performance.

Technology Category

• Search and Optimization: Sampling/Simulation-based Search
• Planning, Routing, and Scheduling: Replanning and Plan Repair
• Intelligent Robots: Motion and Path Planning

Application Category

• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization

Abstract

Model Predictive Path Integral (MPPI) control is a widely used sampling-based approach for real-time control, offering flexibility in handling arbitrary dynamics and cost functions. However, the original MPPI suffers from high-frequency noise in the sampled control trajectories, leading to actuator wear and inefficient exploration. In this work, we introduce Low-Pass Model Predictive Path Integral Control (LP-MPPI), which integrates low-pass filtering into the sampling process to eliminate detrimental high-frequency components and improve the effectiveness of the control trajectories exploration. Unlike prior approaches, LP-MPPI provides direct and interpretable control over the frequency spectrum of sampled trajectories, enhancing sampling efficiency and control smoothness. Through extensive evaluations in Gymnasium environments, simulated quadruped locomotion, and real-world F1TENTH autonomous racing, we demonstrate that LP-MPPI consistently outperforms state-of-the-art MPPI variants, achieving significant performance improvements while reducing control signal chattering.

Problem

• Reduces high-frequency noise in MPPI control trajectories
• Improves control smoothness and sampling efficiency
• Enhances performance in real-time control applications

Innovation

• Integrates low-pass filtering in MPPI control
• Direct control over trajectory frequency spectrum
• Reduces control signal chattering effectively