Research Article
End-to-End Autonomous Exploration with Deep Reinforcement Learning and Intrinsic Motivation
Table 3
Experiment results of learning exploration from scratch.
| Environment | Method | Reward | MER (%) | IQRE |
| Maze-1 | TRPO | 327.36 | 55.29 | N/A | VIME | 321.14 | 53.58 | N/A | EX2 | 489.27 | 82.43 | N/A | ICM | 584.59 | 100.00 | 7.93 | Ours | 586.32 | 100.00 | 4.72 | Maze-2 | TRPO | 232.47 | 41.02 | N/A | VIME | 228.34 | 39.98 | N/A | EX2 | 425.73 | 74.56 | N/A | ICM | 567.28 | 100.00 | 8.07 | Ours | 571.87 | 100.00 | 5.15 | Maze-3 | TRPO | 243.49 | 41.73 | N/A | VIME | 276.54 | 47.82 | N/A | EX2 | 339.62 | 58.35 | N/A | ICM | 532.27 | 91.64 | N/A | Ours | 579.65 | 100.00 | 6.54 |
|
|