Benchmark assesses driving decisions by changing road intentions
PlanContext: Benchmarking In-Context Driving Decision-Making
RoboticsComputer Vision and Pattern Recognition
Summary
Driving a car safely means making the right decisions based on the current situation. The researchers created a new testing method called PlanContext that looks at how well driving systems adjust their decisions when the driver’s high-level goals change but the road scene stays the same. They tested many existing driving models and found that a car driving well by usual measures doesn't always make good decisions when the route changes. This means it’s important to judge driving systems not just on smoothness or safety, but also on how well they understand and react to the driving context.
What this means in practice
- •For autonomous vehicle developers: Evaluate how driving models adapt to different driving goals in the same scene for safer decision-making.
- •For automotive simulation teams: Test decision-making robustness in simulated driving scenarios varying only the driver’s intentions.
Tested on simulated data.
Authors
Ruikai Li, Naiyu Gao, Hao Shan, Yizhe Fan, Yanhao Wu, Han Jiang, Yilong Ren, Zhenning Li, Haiyang Yu, Yanhu Shan, Rui Wu, Hao Zhao, Hongyang Li, Zhiyong Cui
Abstract
Real-world driving requires context-aware decision-making. Existing end-to-end driving benchmarks primarily assess trajectory quality through safety, compliance, comfort, and driving progress, leaving context-dependent decision-making insufficiently characterized. To fill this gap, we present PlanContext, a benchmark for evaluating in-context driving decision-making. PlanContext constructs paired evaluations by varying high-level driving intentions while keeping the observed scene fixed, enabling a natural and controllable assessment of context-dependent decision-making. We further define six representative driving contexts and jointly evaluate contextual decision correctness, road compliance, and dynamic safety with class-balanced metrics. Experiments on 9,561 scene-condition instances from NAVSIM and 23 public end-to-end driving models show that high conventional driving-quality scores do not consistently correspond to strong route-level in-context driving decision-making. Further analysis shows that model behavior varies with decision timing, road structure, and traffic interaction depending on the decision required by the current context. Our results highlight in-context driving decision-making as an important complementary dimension for evaluating end-to-end driving systems.