Inverted Double Pendulum on Inverted Double Pendulum
100Convergence (%)DAgger
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DAggerInitial expert dataset size (M)=1000, Training steps=15000, K=52025.11 | 100 | 19,031,379 | 15,545,119 | |
| DAggerInitial expert dataset size (M)=2000, Training steps=15000, K=52025.11 | 100 | 23,241,211 | 15,606,251 | |
| DAggerInitial expert dataset size (M)=3000, Training steps=15000, K=52025.11 | 100 | 34,332,497 | 15,650,151 | |
| DAggerInitial expert dataset size (M)=5000, Training steps=15000, K=52025.11 | 100 | 33,062,662 | 15,461,143 | |
| DAggerInitial expert dataset size (M)=10000, Training steps=15000, K=52025.11 | 100 | 642,878 | 1,550,868 | |
| EnsembleDAggerInitial expert dataset size (M)=1000, Training steps=15000, K=52025.11 | 100 | 1,161,874 | 19,601,425 | |
| EnsembleDAggerInitial expert dataset size (M)=3000, Training steps=15000, K=52025.11 | 100 | 910,735 | 1,400,740 | |
| ThriftyDAggerInitial expert dataset size (M)=1000, Training steps=15000, K=52025.11 | 100 | 17,577 | 1,843,678 | |
| ThriftyDAggerInitial expert dataset size (M)=2000, Training steps=15000, K=52025.11 | 100 | 303,211 | 856,323 | |
| CRSAILInitial expert dataset size (M)=1000, Training steps=15000, K=52025.11 | 100 | 16,442 | 25,469 | |
| CRSAILInitial expert dataset size (M)=2000, Training steps=15000, K=52025.11 | 100 | 20,630 | 33,873 | |
| CRSAILInitial expert dataset size (M)=3000, Training steps=15000, K=52025.11 | 100 | 19,030 | 484,247 | |
| CRSAILInitial expert dataset size (M)=5000, Training steps=15000, K=52025.11 | 100 | 23,345 | 866,355 | |
| CRSAILInitial expert dataset size (M)=10000, Training steps=15000, K=52025.11 | 100 | 466,233 | 784,172 | |
| EnsembleDAggerInitial expert dataset size (M)=2000, Training steps=15000, K=52025.11 | 80 | 583,504 | 1,537,954 | |
| EnsembleDAggerInitial expert dataset size (M)=5000, Training steps=15000, K=52025.11 | 80 | 815,612 | 1,571,744 | |
| EnsembleDAggerInitial expert dataset size (M)=10000, Training steps=15000, K=52025.11 | 80 | 512,109 | 1,318,717 | |
| ThriftyDAggerInitial expert dataset size (M)=3000, Training steps=15000, K=52025.11 | 80 | 24,658 | 1,975,656 | |
| ThriftyDAggerInitial expert dataset size (M)=5000, Training steps=15000, K=52025.11 | 80 | 813,988 | 1,342,748 | |
| ThriftyDAggerInitial expert dataset size (M)=10000, Training steps=15000, K=52025.11 | 80 | 282,155 | 1,581,781 |