MAVEN-T: Multi-Agent enVironment-aware Enhanced Neural Trajectory predictor with Reinforcement Learning

TL;DR

MAVEN-T is a novel multi-agent trajectory prediction framework that combines architectural co-design, progressive distillation, and reinforcement learning to achieve state-of-the-art accuracy with high efficiency in autonomous driving.

Contribution

It introduces a teacher-student framework with hybrid attention, adaptive curriculum learning, and reinforcement learning to improve model compression and decision-making in trajectory prediction.

Findings

Achieves 6.2x parameter compression and 3.7x inference speedup.

Maintains state-of-the-art accuracy on NGSIM and highD datasets.

Demonstrates robustness through reinforcement learning-based knowledge refinement.

Abstract

Trajectory prediction remains a critical yet challenging component in autonomous driving systems, requiring sophisticated reasoning capabilities while meeting strict real-time deployment constraints. While knowledge distillation has demonstrated effectiveness in model compression, existing approaches often fail to preserve complex decision-making capabilities, particularly in dynamic multi-agent scenarios. This paper introduces MAVEN-T, a teacher-student framework that achieves state-of-the-art trajectory prediction through complementary architectural co-design and progressive distillation. The teacher employs hybrid attention mechanisms for maximum representational capacity, while the student uses efficient architectures optimized for deployment. Knowledge transfer is performed via multi-granular distillation with adaptive curriculum learning that dynamically adjusts complexity based on performance. Importantly, the framework incorporates reinforcement learning to overcome the imitation ceiling of traditional distillation, enabling the student to verify, refine, and optimize teacher knowledge through dynamic environmental interaction, potentially achieving more robust decision-making than the teacher itself. Extensive experiments on NGSIM and highD datasets demonstrate 6.2x parameter compression and 3.7x inference speedup while maintaining state-of-the-art accuracy, establishing a new paradigm for deploying sophisticated reasoning models under resource constraints.