Assistive driving agents improve safety by understanding human intent
Dream to Assist: Intent-Conditioned World Models for Shared Vehicle Control
RoboticsHuman-Computer InteractionMachine Learning
Summary
Driving aids that help control cars need to both act in ways humans want and be clear enough for humans to understand. The researchers created a system, Dream2Assist, that learns to predict what a human driver intends to do and helps in a way that doesn't take control away. They tested it with people in racing and city driving scenarios and found it improved how well the car's assistance matched driver actions and reduced crashes. The work also shows how to make such systems safe and practical for real-world use.
What this means in practice
- •For automotive engineers: Develop assistive driving systems that predict driver intent to improve safety and control alignment in both racing and urban environments.
- •For robotics software developers: Implement modular world models with intent conditioning for tightly coupled human-robot shared control tasks requiring real-time interaction.
Authors
Jonathan DeCastro, Thomas Balch, Emily Sumner, Deepak Gopinath, Laporsha Dees, Patricio Reyes Gomez, Elliot Weiss, Megan Applegate-Kenton, Todd Rowell, John Leonard, Avinash Balachandran, Guy Rosman
Abstract
Assistive agents in shared-control settings face a dual challenge: acting proactively to align with the human's intentions while remaining legible enough for the human to comprehend and fluently respond in real time. We present Dream2Assist, a model-based reinforcement learning framework for tightly-coupled shared decision making that addresses both sides of this challenge. Our approach extends a recurrent state-space world model with intent inference, enabling an assistive policy trained via fictitious co-play to anticipate and support discrete human objectives without overriding human agency. Beyond the computational framework, we describe the full deployment pipeline needed to translate a world-model-based policy into a safe, human-centered system in a resource-efficient way. Crucially, we address the sim-to-real-human gap via policy down-selection and reward restructuring within an iterative human-machine interface design. We validate the framework through three human-subjects studies and a preliminary on-vehicle deployment spanning two qualitatively distinct domains: high-speed racing (N=30) and urban driving with richer intent structure (N=10). Our results demonstrate that intent conditioning yields measurable improvements in torque alignment and collision rate. We surface key design insights and practices for deploying learning-based assistive agents in safety-critical, real-time human-robot teaming.