Perfect Prediction or Plenty of Proposals? What Matters Most in Planning for Autonomous Driving

2025-10-17

AI Summary

Despite extensive exploration of integrated prediction and planning (IPP) in autonomous driving, its practical benefits remain unclear. This paper reveals that even with perfect prediction, planning performance does not improve significantly; instead, the quality and diversity of candidate trajectories are decisive. To address this, we propose a “proposal-centric” paradigm: prediction outputs are used exclusively for collision checking and trajectory filtering—not for direct planning decisions. We enhance the PDM framework by integrating imitation learning and lane-following policies to generate more realistic and kinematically feasible trajectory proposals. Evaluated on Val14 and the challenging interPlan benchmark—featuring highly interactive and out-of-distribution scenarios—our method achieves state-of-the-art performance, notably outperforming existing approaches on interPlan. Our core contributions are twofold: (i) empirically establishing proposal quality—not prediction accuracy—as the primary determinant of planning efficacy; and (ii) introducing a novel prediction-assisted filtering mechanism that decouples prediction from decision-making while leveraging it for safety-critical validation.

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

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

Abstract

Traditionally, prediction and planning in autonomous driving (AD) have been treated as separate, sequential modules. Recently, there has been a growing shift towards tighter integration of these components, known as Integrated Prediction and Planning (IPP), with the aim of enabling more informed and adaptive decision-making. However, it remains unclear to what extent this integration actually improves planning performance. In this work, we investigate the role of prediction in IPP approaches, drawing on the widely adopted Val14 benchmark, which encompasses more common driving scenarios with relatively low interaction complexity, and the interPlan benchmark, which includes highly interactive and out-of-distribution driving situations. Our analysis reveals that even access to perfect future predictions does not lead to better planning outcomes, indicating that current IPP methods often fail to fully exploit future behavior information. Instead, we focus on high-quality proposal generation, while using predictions primarily for collision checks. We find that many imitation learning-based planners struggle to generate realistic and plausible proposals, performing worse than PDM - a simple lane-following approach. Motivated by this observation, we build on PDM with an enhanced proposal generation method, shifting the emphasis towards producing diverse but realistic and high-quality proposals. This proposal-centric approach significantly outperforms existing methods, especially in out-of-distribution and highly interactive settings, where it sets new state-of-the-art results.

Problem

• Evaluating how prediction integration affects autonomous driving planning performance
• Investigating why current methods fail to exploit perfect future behavior predictions
• Developing enhanced proposal generation for better planning in interactive scenarios

Innovation

• Enhanced proposal generation for diverse realistic plans
• Uses predictions primarily for collision checking
• Shifts focus from prediction to proposal quality