Sim-to-Real Transfer for Muscle-Actuated Robots via Generalized Actuator Networks

2026-04-10

AI Summary

This work addresses the challenge of sim-to-real transfer in tendon- and soft-muscle-driven robots, whose nonlinearities, friction, and hysteresis hinder the direct deployment of simulation-trained policies. The authors propose a sim-to-real transfer framework based on a Generalized Actuator Network (GeAN), which learns an actuator model from joint position trajectories and integrates it with rigid-body simulation to capture system dynamics and environmental interactions. Notably, the approach generalizes across diverse actuators without requiring torque sensors. The method enables, for the first time, the successful deployment of a reinforcement learning policy trained solely in simulation onto a four-degree-of-freedom pneumatic artificial muscle arm (PAMY2), achieving high-precision target reaching and dynamic ball-in-cup tasks, thereby demonstrating its effectiveness and broad applicability.

Technology Category

• Intelligent Robots: Manipulation
• Machine Learning: Imitation Learning & Inverse Reinforcement Learning
• 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
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications

Abstract

Tendon drives paired with soft muscle actuation enable faster and safer robots while potentially accelerating skill acquisition. Still, these systems are rarely used in practice due to inherent nonlinearities, friction, and hysteresis, which complicate modeling and control. So far, these challenges have hindered policy transfer from simulation to real systems. To bridge this gap, we propose a sim-to-real pipeline that learns a neural network model of this complex actuation and leverages established rigid body simulation for the arm dynamics and interactions with the environment. Our method, called Generalized Actuator Network (GeAN), enables actuation model identification across a wide range of robots by learning directly from joint position trajectories rather than requiring torque sensors. Using GeAN on PAMY2, a tendon-driven robot powered by pneumatic artificial muscles, we successfully deploy precise goal-reaching and dynamic ball-in-a-cup policies trained entirely in simulation. To the best of our knowledge, this result constitutes the first successful sim-to-real transfer for a four-degrees-of-freedom muscle-actuated robot arm.

Problem

• sim-to-real transfer
• muscle-actuated robots
• nonlinear actuation
• tendon-driven systems
• hysteresis

Innovation

• Generalized Actuator Network
• sim-to-real transfer
• muscle-actuated robots
• tendon-driven actuation
• model identification