Crowd Robot Navigation on Crowd Robot Navigation 500 cases (test)
95.4Success RateSPLC-CQL
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| SPLC-CQLReward Function=Social Preference Learning (SPLC), Offline RL Algorithm=CQL2026.07 | 95.4 | 4.6 | 0 | 11.18 | |
| SPLC-IQLReward Function=Social Preference Learning (SPLC), Offline RL Algorithm=IQL2026.07 | 94.6 | 5.4 | 0 | 11.37 | |
| HR-IQLReward Function=Handcrafted (HR), Offline RL Algorithm=IQL2026.07 | 93.4 | 6.2 | 0.4 | 11.95 | |
| HPR-IQLReward Function=Human-labeled Preference (HPR), Offline RL Algorithm=IQL2026.07 | 93.2 | 6.8 | 0 | 11.1 | |
| RPR-IQLReward Function=Robust Preference-based (RPR), Offline RL Algorithm=IQL2026.07 | 92.2 | 7.8 | 0 | 11.06 | |
| SPLC-TD3BCReward Function=Social Preference Learning (SPLC), Offline RL Algorithm=TD3BC2026.07 | 90.6 | 7.4 | 2 | 11.9 | |
| RPR-CQLReward Function=Robust Preference-based (RPR), Offline RL Algorithm=CQL2026.07 | 86.2 | 13.8 | 0 | 10.75 | |
| HPR-CQLReward Function=Human-labeled Preference (HPR), Offline RL Algorithm=CQL2026.07 | 82.6 | 17.4 | 0 | 10.86 | |
| HR-TD3BCReward Function=Handcrafted (HR), Offline RL Algorithm=TD3BC2026.07 | 82.4 | 13 | 4.6 | 13.7 | |
| RPR-TD3BCReward Function=Robust Preference-based (RPR), Offline RL Algorithm=TD3BC2026.07 | 79 | 12.6 | 8.4 | 14.63 | |
| HPR-TD3BCReward Function=Human-labeled Preference (HPR), Offline RL Algorithm=TD3BC2026.07 | 62.2 | 7 | 30.8 | 16.74 | |
| HR-CQLReward Function=Handcrafted (HR), Offline RL Algorithm=CQL2026.07 | 36.4 | 63.6 | 0 | 11.05 |