Boundary-Guided Trajectory Prediction for Road Aware and Physically Feasible Autonomous Driving

2025-05-10

AI Summary

Autonomous vehicle trajectory prediction must jointly satisfy road compliance and kinematic feasibility, yet existing deep learning methods often produce off-road or kinematically infeasible trajectories, relying heavily on post-hoc correction. To address this, we propose a high-definition map-guided boundary-constrained regression framework: for the first time, trajectory prediction is formulated as a regression problem explicitly constrained by left/right piecewise-linear drivable region boundaries; an explicit acceleration temporal profile is introduced to govern longitudinal traversal distance. We further employ hypergraph-based path superposition representation, kinematic constraint embedding during training, and adversarial robustness evaluation to achieve end-to-end physically consistent prediction. On Argoverse-2, our method reduces final displacement error significantly, lowers off-road rate from 66% to 1%, and eliminates all kinematically infeasible trajectories. It demonstrates superior generalization to rare maneuvers and out-of-distribution scenarios.

Technology Category

• Humans and AI: Human-Aware Planning and Behavior Prediction
• Planning, Routing, and Scheduling: Model-Based Reasoning
• Computer Vision: Motion & Tracking

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Accurate prediction of surrounding road users' trajectories is essential for safe and efficient autonomous driving. While deep learning models have improved performance, challenges remain in preventing off-road predictions and ensuring kinematic feasibility. Existing methods incorporate road-awareness modules and enforce kinematic constraints but lack plausibility guarantees and often introduce trade-offs in complexity and flexibility. This paper proposes a novel framework that formulates trajectory prediction as a constrained regression guided by permissible driving directions and their boundaries. Using the agent's current state and an HD map, our approach defines the valid boundaries and ensures on-road predictions by training the network to learn superimposed paths between left and right boundary polylines. To guarantee feasibility, the model predicts acceleration profiles that determine the vehicle's travel distance along these paths while adhering to kinematic constraints. We evaluate our approach on the Argoverse-2 dataset against the HPTR baseline. Our approach shows a slight decrease in benchmark metrics compared to HPTR but notably improves final displacement error and eliminates infeasible trajectories. Moreover, the proposed approach has superior generalization to less prevalent maneuvers and unseen out-of-distribution scenarios, reducing the off-road rate under adversarial attacks from 66% to just 1%. These results highlight the effectiveness of our approach in generating feasible and robust predictions.

Problem

• Prevent off-road trajectory predictions for autonomous driving
• Ensure kinematic feasibility in trajectory forecasting
• Improve generalization to rare maneuvers and unseen scenarios

Innovation

• Boundary-guided trajectory prediction for autonomous driving
• Training network with superimposed paths between boundaries
• Predicting acceleration profiles for kinematic feasibility