AI Paper Atlas

Research map / Multi-agent systems and games / Cooperative multi-agent RL

MetaDrive: Composing Diverse Driving Scenarios for Generalizable Reinforcement Learning

Quanyi Li, Zhenghao Peng, Lan Feng, Qihang Zhang, Zhenghai Xue, Bolei Zhou

TPAMI 2022 · Multi-agent systems and games › Cooperative multi-agent RL · approach: MARL benchmarks and grouping · 316 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
train on more procedurally generated scenes (larger N)SAC and PPO agents · MetaDrive PG scenes (test)test success rateImproved
train on more procedurally generated scenes (larger N)SAC and PPO agents · MetaDrive PG scenes (test)rule violation and crash rateImproved
increase number of real Waymo training casesSAC · Waymo real scenariostest success rateImproved
increase PG training set diversity, tested on real Waymo scenesPPO · Waymo real test settest success rateNo clear effect
raise the ratio of real scenes in the training mix (0% to 100%)SAC · Waymo real test settest success rateImproved
use Lagrangian with PID multiplier instead of reward shapingSAC · MetaDrive safe explorationcumulative cost: 3.38 -> 1.90Improved
use Lagrangian instead of reward shapingPPO · MetaDrive safe explorationcumulative cost: 3.33 -> 1.82Improved
use off-policy SAC instead of on-policy PPOreward-shaping RL · MetaDrive safe explorationsuccess rate: 0.207 -> 0.801Improved
offline RL (CQL) instead of behavior cloninglimited human demonstrations · MetaDrive safe explorationsuccess rate: 0.01 -> 0.11Improved
train safe RL on more scenesPPO-RS and PPO-Lag · MetaDrive safety generalizationtest episode costImproved
concatenate K=4 nearest vehicle states into centralized critic instead of mean-field averagingCCPPO · MetaDrive MARL Tollgatesuccess rate (%): 48.62 -> 31.92Hurt
use CoPO instead of IPPOmulti-agent PPO · MetaDrive MARL Intersectionsuccess rate (%): 73.93 -> 77.79Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy