Coupled autoregressive active inference agents for control of multi-joint dynamical systems
This work addresses control challenges in mechanical systems, but it appears incremental as it builds on existing active inference methods with coupling modifications.
The authors tackled the problem of controlling multi-joint dynamical systems by proposing a coupled autoregressive active inference agent, which learned the dynamics of a double mass-spring-damper system and drove it to a desired position, outperforming uncoupled agents in surprise and goal alignment.
We propose an active inference agent to identify and control a mechanical system with multiple bodies connected by joints. This agent is constructed from multiple scalar autoregressive model-based agents, coupled together by virtue of sharing memories. Each subagent infers parameters through Bayesian filtering and controls by minimizing expected free energy over a finite time horizon. We demonstrate that a coupled agent of this kind is able to learn the dynamics of a double mass-spring-damper system, and drive it to a desired position through a balance of explorative and exploitative actions. It outperforms the uncoupled subagents in terms of surprise and goal alignment.