Multimodal Belief-Space Covariance Steering with Active Probing and Influence for Interactive Driving
Devodita Chakravarty, John M. Dolan, Yiwei Lyu
AI summary
Problem
Existing methods treat prediction and control as separate, reactive modules, overlooking how an autonomous vehicle's actions can influence human drivers in uncertain, multimodal traffic.
Approach
We introduce a hierarchical Bayesian belief model updated in real-time, paired with an active probing strategy that identifies ambiguity and plans actions to disambiguate intent while gently steering human decisions, all bounded by a runtime CVaR risk layer.
Key results
- Hierarchical Bayesian belief model for coarse-to-fine intent and motion reasoning
- Active probing strategy that reduces multimodal ambiguity and influences human behavior toward safer outcomes
- Runtime CVaR-based risk layer ensuring probing actions remain within human risk tolerance
- Higher success rates and shorter completion times in lane-merging and intersection simulations versus baselines
Why it matters
Enables safer, more efficient autonomous driving in complex traffic by actively managing human-vehicle interaction instead of passively predicting it.
Abstract
Autonomous driving in complex traffic requires reasoning under uncertainty. Common approaches rely on prediction-based planning or risk-aware control, but these are typically treated in isolation, limiting their ability to capture the coupled nature of action and inference in interactive settings. This gap becomes especially critical in uncertain scenarios, where simply reacting to predictions can lead to unsafe ma- neuvers or overly conservative behavior. Our central insight is that safe interaction requires not only estimating human behavior but also shaping it when ambiguity poses risks. To this end, we introduce a hierarchical belief model that structures human behavior across coarse discrete intents and fine motion modes, updated via Bayesian inference for interpretable multi- resolution reasoning. On top of this, we develop an active probing strategy that identifies when multimodal ambiguity in human predictions may compromise safety and plans dis- ambiguating actions that both reveal intent and gently steer human decisions toward safer outcomes. Finally, a runtime risk- evaluation layer based on Conditional Value-at-Risk (CVaR) ensures that all probing actions remain within human risk tolerance during influence. Our simulations in lane-merging and unsignaled intersection scenarios demonstrate that our approach achieves higher success rates and shorter completion times compared to existing methods. These results highlight the benefit of coupling belief inference, probing, and risk monitoring, yielding a principled and interpretable framework for planning under uncertainty.