Skip to content
Open access

Fine-Grained Behavior and Lane Constraints Guided Trajectory Prediction Method

Mar 2025 · IEEE transactions on intelligent transportation systems (Print) · Vol 27, pp. 14036-14050 · 2 citations · 69 references
Computer Science

Abstract

Trajectory prediction, as a critical component of autonomous driving systems, has attracted considerable attention in recent years. Existing prediction methods typically focus on extracting richer scene representations or introducing trajectory-level guidance through destinations, goals, or motion modes. However, future vehicle motion is continuously influenced by both evolving driving intentions and structured lane constraints. Such dynamic factors are difficult to characterize using coarse trajectory-level representations alone, which may limit prediction accuracy. To address this challenge, we propose BLNet, a novel fine-grained trajectory prediction framework that explicitly models behavioral evolution and lane topology constraints throughout the prediction horizon. Specifically, a dual-stream architecture is designed to generate timestamp-level behavior-state queries and lane queries, enabling future motion to be characterized from both behavioral and lane-constrained perspectives. The generated queries are supervised by dedicated auxiliary objectives and jointly guide multimodal trajectory generation. Furthermore, a two-stage decoder is introduced to first produce trajectory proposals and then perform point-level refinement by incorporating lane continuity and future motion features. Extensive experiments on the nuScenes and Argoverse benchmarks demonstrate the effectiveness of the proposed framework. The results show that BLNet achieves a favorable balance between prediction accuracy and inference efficiency while maintaining competitive forecasting performance.

Read PDF

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.