Dual Online Stein Variational Inference for Control and Dynamics

TL;DR

This paper introduces a novel implicit variational inference algorithm using Stein Variational gradient descent for real-time control and dynamic environment adaptation in model predictive control systems.

Contribution

It develops a dual online Stein Variational inference method that estimates distributions over model parameters and control inputs dynamically, improving MPC performance in complex environments.

Findings

Effective in approximating complex multi-modal posteriors

Enables real-time adaptation to changing environments

Demonstrates superior performance in simulated and real-world tasks

Abstract

Model predictive control (MPC) schemes have a proven track record for delivering aggressive and robust performance in many challenging control tasks, coping with nonlinear system dynamics, constraints, and observational noise. Despite their success, these methods often rely on simple control distributions, which can limit their performance in highly uncertain and complex environments. MPC frameworks must be able to accommodate changing distributions over system parameters, based on the most recent measurements. In this paper, we devise an implicit variational inference algorithm able to estimate distributions over model parameters and control inputs on-the-fly. The method incorporates Stein Variational gradient descent to approximate the target distributions as a collection of particles, and performs updates based on a Bayesian formulation. This enables the approximation of complex multi-modal posterior distributions, typically occurring in challenging and realistic robot navigation tasks. We demonstrate our approach on both simulated and real-world experiments requiring real-time execution in the face of dynamically changing environments.