DeepTraffic: Crowdsourced Hyperparameter Tuning of Deep Reinforcement Learning Systems for Multi-Agent Dense Traffic Navigation
Lex Fridman, Jack Terwilliger, Benedikt Jenik
Code Available — Be the first to reproduce this paper.
ReproduceCode
- github.com/NeekhilD/Learning-Competitionnone★ 0
- github.com/lexfridman/deeptrafficnone★ 0
- github.com/asarav/MIT-Deep-Traffic-Solutionnone★ 0
- github.com/ashtawy/deeptrafficnone★ 0
- github.com/Bhaney44/MIT_DeepTrafficnone★ 0
- github.com/xiexiexiaoxiexie/Udacity-self-driving-car-engineer-P7-Highway-Drivingnone★ 0
Abstract
We present a traffic simulation named DeepTraffic where the planning systems for a subset of the vehicles are handled by a 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 deep Q-learning network 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.