M-Predictive Spliner: Enabling Spatiotemporal Multi-Opponent Overtaking for Autonomous Racing

2025-06-19

AI Summary

To address the challenge of autonomous racing vehicles struggling to collaboratively overtake an arbitrary number of opponents in high-dynamic, multi-vehicle competitive racing, this paper proposes an intent-driven real-time cooperative overtaking decision-making framework for multiple adversaries. Our method integrates spatiotemporal joint modeling with a closed-loop optimization architecture, enabling, for the first time, joint opponent intent perception and trajectory prediction scalable to arbitrary numbers of adversaries. It unifies multi-target re-identification (ReID), Kalman filter-based tracking, spatial-velocity-coupled Gaussian process regression (GPR), and real-time motion planning and control. Evaluated on a 1:10-scale physical racing platform, our approach achieves a 91.65% overtaking success rate and reduces collision probability by 10.13 percentage points at equivalent speeds—significantly outperforming state-of-the-art methods. The core contribution lies in abandoning the restrictive single-opponent assumption and establishing a scalable, intent-aware multi-agent perception and cooperative decision-making paradigm.

Technology Category

• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems
• Multiagent Systems: Adversarial Agents
• Humans and AI: Planning and Decision Support for Human-Machine Teams

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Unrestricted multi-agent racing presents a significant research challenge, requiring decision-making at the limits of a robot's operational capabilities. While previous approaches have either ignored spatiotemporal information in the decision-making process or been restricted to single-opponent scenarios, this work enables arbitrary multi-opponent head-to-head racing while considering the opponents'future intent. The proposed method employs a KF-based multi-opponent tracker to effectively perform opponent ReID by associating them across observations. Simultaneously, spatial and velocity GPR is performed on all observed opponent trajectories, providing predictive information to compute the overtaking maneuvers. This approach has been experimentally validated on a physical 1:10 scale autonomous racing car, achieving an overtaking success rate of up to 91.65% and demonstrating an average 10.13%-point improvement in safety at the same speed as the previous SotA. These results highlight its potential for high-performance autonomous racing.

Problem

• Enabling multi-opponent overtaking in autonomous racing
• Predicting opponents' future intent for safer maneuvers
• Improving overtaking success rate and safety performance

Innovation

• KF-based multi-opponent tracker for ReID
• Spatial and velocity GPR for predictions
• Validated on 1:10 scale autonomous car