Adaptation through prediction: multisensory active inference torque control

TL;DR

This paper introduces a multisensory active inference torque controller for robotic arms that enhances adaptation, control accuracy, and noise rejection by integrating prediction, learning, and multimodal sensor data.

Contribution

It presents a novel controller inspired by the predictive brain hypothesis that combines learning and multimodal sensor integration, improving upon existing active inference methods.

Findings

Improved control accuracy in goal-directed reaching.

Enhanced noise rejection through multimodal filtering.

Demonstrated adaptability to dynamic changes and disturbances.

Abstract

Adaptation to external and internal changes is major for robotic systems in uncertain environments. Here we present a novel multisensory active inference torque controller for industrial arms that shows how prediction can be used to resolve adaptation. Our controller, inspired by the predictive brain hypothesis, improves the capabilities of current active inference approaches by incorporating learning and multimodal integration of low and high-dimensional sensor inputs (e.g., raw images) while simplifying the architecture. We performed a systematic evaluation of our model on a 7DoF Franka Emika Panda robot arm by comparing its behavior with previous active inference baselines and classic controllers, analyzing both qualitatively and quantitatively adaptation capabilities and control accuracy. Results showed improved control accuracy in goal-directed reaching with high noise rejection due to multimodal filtering, and adaptability to dynamical inertial changes, elasticity constraints and human disturbances without the need to relearn the model nor parameter retuning.