Human-like autonomous car-following model with deep reinforcement learning
Abstract
A car-following model was proposed based on deep reinforcement learning. It uses speed deviations as reward function and considers a reaction delay of 1 s. Deep deterministic policy gradient algorithm was used to optimize the model. The model outperformed traditional and recent data-driven car-following models. The model demonstrated good capability of generalization.
- Publication:
-
Transportation Research Part C: Emerging Technologies
- Pub Date:
- December 2018
- DOI:
- 10.1016/j.trc.2018.10.024
- Bibcode:
- 2018TRPC...97..348Z
- Keywords:
-
- Autonomous car following;
- Human-like driving planning;
- Deep reinforcement learning;
- Naturalistic driving study;
- Deep deterministic policy gradient