Adaptive driving controls improve comfort and safety in simulations
Comfort by Construction: Adaptive, Comfort-Bounded Action Spaces for Learned Driving Policies
RoboticsArtificial Intelligence
Summary
Driving simulators often let cars make sudden moves that real drivers wouldn’t do, making the safety results unreliable. The authors found that simply limiting how much the car can turn or speed up doesn’t work well because some limits get smaller as the car goes faster, causing jerky control. They created a way to adjust the controls at every moment so the car stays within comfortable and realistic limits. Their method reduces uncomfortable driving errors and works better than older ways on real driving data and test scenarios.
What this means in practice
- •For autonomous vehicle engineers: Generate driving policies that respect occupant comfort limits, improving realistic behavior in simulators for safer autonomous driving development.
- •For simulation platform developers: Incorporate adaptive action discretization methods into driving simulators to better represent feasible and comfortable vehicle controls at varying speeds.
Authors
Anna Rothenhäusler, Daniel Jost, Raghu Rajan, Faris Janjos, Oliver Scheel, Andreas Look, Joschka Boedecker
Abstract
Data-driven driving simulators command accelerations and steering rates from a fixed grid without constraining the realized accelerations and jerks. As a result, reinforcement-learning policies inflate safety metrics through abrupt, last-second maneuvers that lie far outside the range of human driving and would be unacceptable to occupants of a real vehicle, so the metrics measure simulator permissiveness rather than policy quality. Enforcing comfort bounds naively is not enough: lateral limits shrink quadratically with speed, so clamping a static grid saturates it and destroys fine-grained control ("grid collapse"). We propose an adaptive action parameterization that rediscretizes the grid at every step to span exactly the per-step feasible control set, via closed-form inversion of the lateral-jerk constraint. We further present PufferDrive-Editor, a browser-based tool to audit realized kinematics and author kinematically challenging scenes. On the Waymo Open Motion Dataset and a hand-authored slalom, our adaptive model holds comfort violations below 1% while outperforming clipped-grid and direct-jerk baselines in navigability.