Driving Behavior Modeling Using Naturalistic Human Driving Data With Inverse Reinforcement Learning

Driving behavior modeling is of great importance for designing safe, smart, and personalized autonomous driving systems. In this paper, an internal reward function-based model that emulates the human’s decision-making mechanism utilized. To infer function parameters from naturalistic human data, we propose a structural assumption about focuses on discrete latent intentions. It converts continuous problem to setting thus makes maximum entropy inverse reinforcement learning (IRL) tractable learn functions. Specifically, polynomial trajectory sampler adopted generate candidate trajectories considering high-level intentions approximate partition in IRL framework. An environment interactive behaviors among ego surrounding vehicles built better estimate generated trajectories. The proposed method applied functions individual drivers NGSIM highway dataset. qualitative results demonstrate learned are able explicitly express preferences different interpret their decisions. quantitative reveal robust, which manifested by only marginal decline proximity when applying testing conditions. For performance, outperforms general approach, significantly reducing errors likeness (a custom metric gauge accuracy), these two methods deliver compared other baseline methods. Moreover, it found predicting response actions incorporating potential decelerations caused vehicle critical estimating trajectories, accuracy planning using relies forecasting model.