Interactive planning improves autonomous driving in busy urban traffic
INTERACT: Interactive Planning for Autonomous Driving via Anchor-Conditioned Prediction and Trust-Region Refinement
Robotics
Summary
Driving in crowded city streets is tricky because how other drivers behave depends on what you intend to do. The authors present a new method called INTERACT that helps self-driving cars plan better by predicting how nearby drivers will react based on broad driving intents instead of exact moves. They break down the planning into parts: first guessing how others would respond to a few possible intentions, then fine-tuning the car’s plan within those intentions. This way, the prediction and planning parts stay separate but connected, making the car’s decisions more reliable and better suited for interactive situations.
What this means in practice
- •For autonomous vehicle developers: Build interactive self-driving systems that better predict and plan around human drivers’ reactions in complex urban scenarios.
- •For robotics software engineers: Improve multi-agent planning by separating prediction from optimization while maintaining interaction dependencies using anchor-based intents.
Authors
Aron Distelzweig, Andreas Look, Faris Janjoš, Steffen Hagedorn, Luigi Palmieri, Joschka Boedecker
Abstract
Driving in dense urban traffic is interactive: whether a merge or an unprotected turn succeeds depends on how surrounding agents respond to the ego vehicle. Conventional planners predict first and plan second and, therefore, cannot account for this dependency. Methods that integrate prediction and planning either train both jointly, which introduces task interference, or keep them separate and are restricted to a predefined set of proposals. We present INTERACT: Interactive Planning for Autonomous Driving via Anchor-Conditioned Prediction and Trust-Region Refinement. Our key insight is that surrounding agents react to the intent a trajectory expresses rather than to its exact realization, so a single reactive prediction stays valid across an entire family of plans. INTERACT therefore decomposes interactive planning into prediction across driving intents and optimization within each intent. We derive a small set of diverse intents, which we call anchors, from map geometry, query a dedicated ego-conditioned prediction model once per anchor, and refine every anchor with the Cross-Entropy Method under a trust-region penalty that keeps the refined plan close enough to its anchor for the conditioned reaction to still apply. Prediction thus remains a separate model, avoiding task interference, while conditioning on anchors preserves the dependency. Because each anchor is refined continuously, the final plan is not restricted to the anchor set, yet INTERACT requires only one predictor query per anchor rather than one per candidate plan, with all anchors processed in parallel. On the nuPlan and interPlan closed-loop benchmarks, INTERACT sets a new state of the art, with the largest gains precisely in the interactive scenarios that motivate the method. The code will be released upon acceptance.