Bayesian inference of wall torques for active Brownian particles

TL;DR

This paper develops a Bayesian inference method to estimate wall torques acting on active Brownian particles, improving model accuracy without needing detailed orientation data, and compares it with traditional least-squares approaches.

Contribution

It introduces a Bayesian framework for inferring wall torques on active particles from trajectory data, enhancing model validation and applicability to experimental data.

Findings

Bayesian inference accurately estimates wall torques from trajectory data.

The method outperforms least-squares fitting in scenarios lacking orientation data.

Application to experimental data demonstrates practical utility.

Abstract

The motility of living things and synthetic self-propelled objects is often described using Active Brownian particles. To capture the interaction of these particles with their often complex environment, this model can be augmented with empirical forces or torques, for example, to describe their alignment with an obstacle or wall after a collision. Here, we assess the quality of these empirical models by comparing their output predictions with trajectories of rod-shaped active particles that scatter sterically at a flat wall. We employ a classical least-squares method to evaluate the instantaneous torque. In addition, we lay out a Bayesian inference procedure to construct the posterior distribution of plausible model parameters. In contrast to the least squares fit, the Bayesian approach does not require orientational data of the active particle and can readily be applied to experimental tracking data.