DeepTraffic: Driving Fast through Dense Traffic with Deep Reinforcement Learning

01/09/2018

∙

We present a micro-traffic simulation (named "DeepTraffic") where the perception, control, and planning systems for one of the cars are all handled by a single neural network as part of a model-free, off-policy reinforcement learning process. The primary goal of DeepTraffic is to make the hands-on study of deep reinforcement learning accessible to thousands of students, educators, and researchers in order to inspire and fuel the exploration and evaluation of DQN variants and hyperparameter configurations through large-scale, open competition. This paper investigates the crowd-sourced hyperparameter tuning of the policy network that resulted from the first iteration of the DeepTraffic competition where thousands of participants actively searched through the hyperparameter space with the objective of their neural network submission to make it onto the top-10 leaderboard.

READ FULL TEXT

DeepTraffic: Driving Fast through Dense Traffic with Deep Reinforcement Learning

Adaptive Rollout Length for Model-Based RL Using Model-Free Deep RL

The MineRL Competition on Sample-Efficient Reinforcement Learning Using Human Priors: A Retrospective

Driving in Dense Traffic with Model-Free Reinforcement Learning

Measuring Sample Efficiency and Generalization in Reinforcement Learning Benchmarks: NeurIPS 2020 Procgen Benchmark

Non-Deterministic Policy Improvement Stabilizes Approximated Reinforcement Learning

Deep reinforcement learning for RAN optimization and control

A modular framework for stabilizing deep reinforcement learning control

DeepTraffic: Driving Fast through Dense Traffic with Deep Reinforcement Learning

Related Research

Adaptive Rollout Length for Model-Based RL Using Model-Free Deep RL

The MineRL Competition on Sample-Efficient Reinforcement Learning Using Human Priors: A Retrospective

Driving in Dense Traffic with Model-Free Reinforcement Learning

Measuring Sample Efficiency and Generalization in Reinforcement Learning Benchmarks: NeurIPS 2020 Procgen Benchmark

Non-Deterministic Policy Improvement Stabilizes Approximated Reinforcement Learning

Deep reinforcement learning for RAN optimization and control

A modular framework for stabilizing deep reinforcement learning control