Nguyen, Kien (Aalto University), Keurulainen, Oskar (Aalto University), Shintemirov, Almas (Aalto University), Kyrki, Ville (Aalto University)

Intention-Aware Autonomous Driving in Interactive Traffic Scenarios through Amortized Active Inference

Scheduled for presentation during the Video Session "On-Demand Video Presentations" (VP-VP), Saturday, November 22, 2025, 08:00−18:00, On-Demand Platform

2025 IEEE 28th International Conference on Intelligent Transportation Systems (ITSC), November 18-21, 2025, Gold Coast, Australia

This information is tentative and subject to change. Compiled on April 2, 2026

Abstract

This work presents a novel active inference framework for autonomous driving in interactive traffic scenarios. The framework is based on amortizing the planning algorithm in a reinforcement learning fashion, that allows it to be suited to real-time requirements. A revised particle-based belief representation captures richer environmental information such as goals, beliefs or intentions of other traffic agents, while also being able to update the belief of an autonomous agent efficiently in a sequential manner. The simulation experiments of two common vehicle-pedestrian interaction scenarios: a) passing an occluding object, b) passing a pedestrian approaching a crosswalk, demonstrate how the model naturally manages vehicle-pedestrian interactions and provide insight on how such interactions are guided by agent’s preference for information seeking and belief about the pedestrian’s intention. The code for the simulation experiments is available at https://github.com/kien-ng17/Amortized-active-inference-for-autonomous-driving.