Unbiased Active Inference for Classical Control

TL;DR

This paper introduces an extended unbiased active inference controller (u-AIC) that overcomes limitations of the standard AIC, demonstrating improved performance in robotic control tasks through simulations and real experiments.

Contribution

It proposes an unbiased version of the active inference controller that removes biases and implicit assumptions, enhancing control accuracy and robustness.

Findings

u-AIC outperforms standard AIC in simulations and real robot experiments.

The approach reduces bias in state estimation and control actions.

Improved control performance demonstrated on a 2-DOF arm and a 7-DOF manipulator.

Abstract

Active inference is a mathematical framework that originated in computational neuroscience. Recently, it has been demonstrated as a promising approach for constructing goal-driven behavior in robotics. Specifically, the active inference controller (AIC) has been successful on several continuous control and state-estimation tasks. Despite its relative success, some established design choices lead to a number of practical limitations for robot control. These include having a biased estimate of the state, and only an implicit model of control actions. In this paper, we highlight these limitations and propose an extended version of the unbiased active inference controller (u-AIC). The u-AIC maintains all the compelling benefits of the AIC and removes its limitations. Simulation results on a 2-DOF arm and experiments on a real 7-DOF manipulator show the improved performance of the u-AIC with respect to the standard AIC. The code can be found at https://github.com/cpezzato/unbiased_aic.