Prediction-Driven Motion Planning: Route Integration Strategies in Attention-Based Prediction Models

2025-12-03

AI Summary

This work addresses the instability and kinematic infeasibility of predicted trajectories in autonomous driving, stemming from insufficient coupling between navigation goals and motion prediction. To this end, we propose a navigation-aware attention mechanism model. Methodologically, we explicitly integrate vehicle navigation paths and target pose information into a multi-agent motion prediction framework, and design multiple navigation feature embedding strategies to enable joint modeling of prediction and planning. Our key contribution is the first deep integration of high-level navigation commands into attention-based spatiotemporal interaction modeling, thereby bridging the semantic gap between prediction and planning modules. Experiments on the nuPlan dataset demonstrate significant improvements: 12.3% reduction in average displacement error (ADE) and 28.7% reduction in collision rate, alongside enhanced long-horizon trajectory stability. The proposed approach establishes a novel paradigm for end-to-end safe decision-making.

Technology Category

• Humans and AI: Human-Aware Planning and Behavior Prediction
• Planning, Routing, and Scheduling: Planning with Language Models
• Multiagent Systems: Multiagent Planning

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Agentic search

Abstract

Combining motion prediction and motion planning offers a promising framework for enhancing interactions between automated vehicles and other traffic participants. However, this introduces challenges in conditioning predictions on navigation goals and ensuring stable, kinematically feasible trajectories. Addressing the former challenge, this paper investigates the extension of attention-based motion prediction models with navigation information. By integrating the ego vehicle's intended route and goal pose into the model architecture, we bridge the gap between multi-agent motion prediction and goal-based motion planning. We propose and evaluate several architectural navigation integration strategies to our model on the nuPlan dataset. Our results demonstrate the potential of prediction-driven motion planning, highlighting how navigation information can enhance both prediction and planning tasks. Our implementation is at: https://github.com/KIT-MRT/future-motion.

Problem

• Integrating navigation goals into attention-based motion prediction models
• Ensuring stable and kinematically feasible trajectories for automated vehicles
• Bridging multi-agent motion prediction with goal-based motion planning

Innovation

• Integrates ego vehicle route into attention-based prediction models
• Proposes architectural strategies for navigation-goal conditioning
• Enables prediction-driven motion planning with kinematically feasible trajectories