Track-to-Track Association for Collective Perception based on Stochastic Optimization

2025-10-24

AI Summary

In intelligent urban environments, multi-vehicle collaborative perception faces challenges including sensor limitations, trajectory ambiguity, and difficulties in cross-node trajectory alignment. To address these, this paper proposes a stochastic optimization-based trajectory association method. It formulates a multidimensional likelihood model integrating both trajectory count and spatial distribution characteristics, and employs Monte Carlo sampling to generate and evaluate multiple association hypotheses—thereby avoiding reliance on heuristic rules or suffering from prohibitive computational complexity inherent in conventional approaches. Extensive validation in both simulated and real-world collaborative perception scenarios demonstrates significant improvements in association accuracy and robustness, particularly under high-density and low-observability conditions. This work establishes a scalable, data-driven association paradigm for multi-agent collective perception, enhancing autonomous driving systems’ capability to model dynamic urban environments.

Technology Category

• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Multiagent Systems: Coordination and Collaboration
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• 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

Collective perception is a key aspect for autonomous driving in smart cities as it aims to combine the local environment models of multiple intelligent vehicles in order to overcome sensor limitations. A crucial part of multi-sensor fusion is track-to-track association. Previous works often suffer from high computational complexity or are based on heuristics. We propose an association algorithms based on stochastic optimization, which leverages a multidimensional likelihood incorporating the number of tracks and their spatial distribution and furthermore computes several association hypotheses. We demonstrate the effectiveness of our approach in Monte Carlo simulations and a realistic collective perception scenario computing high-likelihood associations in ambiguous settings.

Problem

• Optimizing track association for multi-vehicle perception systems
• Overcoming computational complexity in sensor fusion methods
• Resolving ambiguous associations in collective perception scenarios

Innovation

• Uses stochastic optimization for track association
• Incorporates multidimensional likelihood with spatial distribution
• Computes multiple association hypotheses for ambiguity