SMARTS Zoo

This repository contains reinforcement learning (RL) environments and models built using SMARTS.

Agent Zoo

• Interaction-Aware Motion Prediction

  • used at: SMARTS/zoo/policies/interaction_aware_motion_prediction

• Control and Supervised Learning

  • used at: SMARTS/zoo/policies/control_and_supervised_learning

• Discrete Soft Actor Critic

  • used at: SMARTS/zoo/policies/discrete_soft_actor_critic

Environment zoo

• ULTRA provides a gym-based environment built upon SMARTS to tackle intersection navigation, specifically the unprotected left turn.

NeurIPS2022

• The directory neurips2022 contains additional example solutions for Track2 of the Driving SMARTS competition.