Learning An Active Inference Model of Driver Perception and Control: Application to Vehicle Car-Following

2023-03-27

AI Summary

This study addresses the challenge of modeling driver perception–control coupling in car-following behavior, aiming to construct an interpretable and cognitively consistent model from limited naturalistic driving data. We propose a Bayesian framework grounded in active inference, unifying perception, preference, and control under the principle of minimizing sensory surprise. To bridge cognitive plausibility and data fidelity, we introduce a novel bi-level optimization method that jointly integrates human cognitive priors with data-driven parameter estimation—overcoming the interpretability bottleneck of black-box models. Technically, the framework integrates variational Bayesian inference, dynamic Bayesian networks, and sensorimotor control modeling. Empirical validation on real-world car-following datasets demonstrates that our model significantly outperforms conventional black-box approaches in prediction accuracy while providing explicit, mechanistic explanations rooted in cognitive theory—establishing a new paradigm for trustworthy human–autonomy shared driving.

Abstract

In this paper we introduce a general estimation methodology for learning a model of human perception and control in a sensorimotor control task based upon a finite set of demonstrations. The model's structure consists of i the agent's internal representation of how the environment and associated observations evolve as a result of control actions and ii the agent's preferences over observable outcomes. We consider a model's structure specification consistent with active inference, a theory of human perception and behavior from cognitive science. According to active inference, the agent acts upon the world so as to minimize surprise defined as a measure of the extent to which an agent's current sensory observations differ from its preferred sensory observations. We propose a bi-level optimization approach to estimation which relies on a structural assumption on prior distributions that parameterize the statistical accuracy of the human agent's model of the environment. To illustrate the proposed methodology, we present the estimation of a model for car-following behavior based upon a naturalistic dataset. Overall, the results indicate that learning active inference models of human perception and control from data is a promising alternative to black-box models of driving.

Problem

• Learn human perception and control model from demonstrations
• Apply active inference theory to car-following behavior
• Estimate model using bi-level optimization approach

Innovation

• Active inference model for driver perception
• Bi-level optimization for estimation approach
• Learning from demonstrations with structural assumptions