Real-Time Model Predictive Control for the Swing-Up Problem of an Underactuated Double Pendulum

2025-04-07

AI Summary

This work addresses the global swing-up and stabilization control problem for underactuated double-pendulum systems (Acrobot/Pendubot). We propose a lightweight, real-time nonlinear model predictive control (NMPC) framework grounded in exact nonlinear dynamical modeling, incorporating efficient real-time optimization and explicit constraint handling to guarantee global convergence from arbitrary initial states. Compared to conventional LQR and reinforcement learning baselines, the method significantly improves robustness against initial condition sensitivity and external disturbances: it achieves 100% swing-up success rate on the RealAIGym platform with end-to-end latency under 50 ms. The core contribution lies in an NMPC architecture that jointly ensures theoretical guarantees—such as global stability and constraint satisfaction—and engineering feasibility—namely hard real-time execution—thereby delivering a verifiable, deployable global control solution for underactuated systems.

Technology Category

• Multiagent Systems: Multiagent Systems under Uncertainty
• Intelligent Robots: State Estimation
• Reasoning under Uncertainty: Stochastic Optimization

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work
• Search and Retrieval-Augmented AI: Agentic search

Abstract

The 3rd AI Olympics with RealAIGym competition poses the challenge of developing a global policy that can swing up and stabilize an underactuated 2-link system Acrobot and/or Pendubot from any configuration in the state space. This paper presents an optimal control-based approach using a real-time Nonlinear Model Predictive Control (MPC). The results show that the controller achieves good performance and robustness and can reliably handle disturbances.

Problem

• Swing up and stabilize underactuated double pendulum
• Develop global policy for Acrobot/Pendubot control
• Real-time Nonlinear MPC for robust performance

Innovation

• Real-time Nonlinear Model Predictive Control
• Swing-up underactuated double pendulum
• Robust performance with disturbances