List your company
ResearchFormerly WizWand

Multi-agent Collision Avoidance on 50-map out-of-distribution benchmark

1.083Normalized collisionsIPPO + ARMS
0.815522.621014.42656.23199May 22, 2026
Updated 5mo ago

Evaluation results

MethodLinks
IPPO + ARMSAlgorithm=IPPO, Agent Count=8, alpha=0.35, reward sparsity=20, Reward Shaping=ARMS2026.051.083
IPPO + PBRSAlgorithm=IPPO, Agent Count=8, alpha=0.35, reward sparsity=20, Reward Shaping=PBRS2026.051.36
IPPO (No reward shaping)Algorithm=IPPO, Agent Count=8, alpha=0.35, reward sparsity=20, Reward Shaping=None2026.051.371
IPPO + ARMSAlgorithm=IPPO, Agent Count=16, alpha=0.35, reward sparsity=20, Reward Shaping=ARMS2026.051.955
IPPO (No reward shaping)Algorithm=IPPO, Agent Count=16, alpha=0.35, reward sparsity=20, Reward Shaping=None2026.052.875
IPPO + PBRSAlgorithm=IPPO, Agent Count=16, alpha=0.35, reward sparsity=20, Reward Shaping=PBRS2026.052.958
IPPO + ARMSAlgorithm=IPPO, Agent Count=32, alpha=0.35, reward sparsity=20, Reward Shaping=ARMS2026.054.273
IPPO + PBRSAlgorithm=IPPO, Agent Count=32, alpha=0.35, reward sparsity=20, Reward Shaping=PBRS2026.056.619
IPPO (No reward shaping)Algorithm=IPPO, Agent Count=32, alpha=0.35, reward sparsity=20, Reward Shaping=None2026.057.77