Generalizable Trajectory Prediction via Inverse Reinforcement Learning with Mamba-Graph Architecture

2025-06-14

AI Summary

Modeling driving behavior and predicting trajectories in complex traffic scenarios remain challenging due to poor generalizability across diverse environments. To address this, we propose a novel hybrid architecture integrating Inverse Reinforcement Learning (IRL) with Mamba—a state-space model—and Graph Attention Networks (GATs). This is the first work to jointly leverage IRL, Mamba, and GATs for interpretable reward function inference and unified spatiotemporal representation learning. Within an encoder–decoder framework, our method effectively captures both the diversity of cross-scenario driving policies and underlying structural constraints. Experiments on urban intersection and roundabout datasets demonstrate superior prediction accuracy over mainstream baselines. Notably, our approach achieves twice the cross-scenario generalization performance of existing IRL-based methods, significantly enhancing adaptability to previously unseen traffic scenarios.

Technology Category

• Planning, Routing, and Scheduling: Model-Based Reasoning
• Machine Learning: Imitation Learning & Inverse Reinforcement Learning
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Accurate driving behavior modeling is fundamental to safe and efficient trajectory prediction, yet remains challenging in complex traffic scenarios. This paper presents a novel Inverse Reinforcement Learning (IRL) framework that captures human-like decision-making by inferring diverse reward functions, enabling robust cross-scenario adaptability. The learned reward function is utilized to maximize the likelihood of output by the encoder-decoder architecture that combines Mamba blocks for efficient long-sequence dependency modeling with graph attention networks to encode spatial interactions among traffic agents. Comprehensive evaluations on urban intersections and roundabouts demonstrate that the proposed method not only outperforms various popular approaches in prediction accuracy but also achieves 2 times higher generalization performance to unseen scenarios compared to other IRL-based method.

Problem

• Modeling human-like driving decisions in complex traffic
• Improving trajectory prediction accuracy and generalization
• Combining Mamba blocks with graph attention networks

Innovation

• Inverse Reinforcement Learning for reward inference
• Mamba blocks for long-sequence dependency modeling
• Graph attention networks for spatial interactions