Conditional Prediction by Simulation for Automated Driving

2025-02-05

AI Summary

Existing modular autonomous driving systems decouple trajectory prediction from motion planning, hindering explicit modeling of inter-vehicle trajectory dependencies and limiting collaborative planning capabilities. To address this, we propose a microsimulation-based conditional trajectory prediction method: leveraging ego-vehicle-generated candidate trajectories as conditioning inputs, it jointly models surrounding agents’ responsive behaviors. Crucially, we introduce adversarial inverse reinforcement learning (AIRL)-trained behavior models into the simulation framework—enabling cross-agent conditional modeling and online adaptation of ego-vehicle trajectories. This approach breaks the conventional prediction-planning separation paradigm. Evaluated on complex interactive scenarios—including cooperative lane-changing and merging—the method significantly improves prediction fidelity, planning rationality, and safety while maintaining high interpretability.

Technology Category

• Humans and AI: Human-Aware Planning and Behavior Prediction
• Multiagent Systems: Adversarial Agents
• Machine Learning: Imitation Learning & Inverse Reinforcement Learning

Application Category

• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Modular automated driving systems commonly handle prediction and planning as sequential, separate tasks, thereby prohibiting cooperative maneuvers. To enable cooperative planning, this work introduces a prediction model that models the conditional dependencies between trajectories. For this, predictions are generated by a microscopic traffic simulation, with the individual traffic participants being controlled by a realistic behavior model trained via Adversarial Inverse Reinforcement Learning. By assuming various candidate trajectories for the automated vehicle, we generate predictions conditioned on each of them. Furthermore, our approach allows the candidate trajectories to adapt dynamically during the prediction rollout. Several example scenarios are available at https://conditionalpredictionbysimulation.github.io/.

Problem

• model conditional trajectory dependencies
• enable cooperative automated driving
• dynamic adaptation during prediction

Innovation

• Adversarial Inverse Reinforcement Learning
• Microscopic traffic simulation
• Dynamic trajectory adaptation