Benchmarking Behavior Prediction Models in Gap Acceptance Scenarios

Autonomous vehicles currently suffer from a time-inefficient driving style caused by uncertainty about human behavior in traffic interactions. Accurate and reliable prediction models enabling more efficient trajectory planning could make autonomous assertive such However, the evaluation of is commonly oversimplistic, ignoring asymmetric importance errors heterogeneity datasets used for testing. We examine potential recasting interactions between as gap acceptance scenarios evaluating this structured environment. To that end, we develop framework aiming to facilitate any model, metric, scenario. then apply state-of-the-art models, which all show themselves be unreliable most safety-critical situations.