Table 4 Ablation experiments.
From: Efficient crowd simulation in complex environment using deep reinforcement learning
| Â | Level | Calculation cost(s)\(\downarrow\) | Collision(%)\(\downarrow\) | Timeout(%) \(\downarrow\) | Completion time(s)\(\downarrow\) | Velocity variance\(\downarrow\) |
|---|---|---|---|---|---|---|
w/o AFs & expected speed | Easy | 2.12 | 4.46 | 4.82 | 282.8 | 0.000472 |
Middle | 2.79 | 9.14 | 10.1 | 376.1 | 0.000454 | |
Hard | 3.68 | 14.0 | 15.7 | 488.6 | 0.000462 | |
w/o AFs | Easy | 2.16 | 4.52 | 3.86 | 280.8 | 0.000196 |
Middle | 2.88 | 8.72 | 7.94 | 361.7 | 0.000187 | |
Hard | 3.96 | 12.4 | 12.5 | 500.8 | 0.000188 | |
w/o expected speed | Easy | 2.11 | 2.22 | 2.20 | 244.9 | 0.000484 |
Middle | 2.63 | 5.34 | 5.12 | 316.9 | 0.000431 | |
Hard | 3.48 | 9.66 | 10.2 | 420.0 | 0.000455 | |
Ours | Easy | 2.09 | 1.84 | 1.90 | 231.7 | 0.000171 |
Middle | 2.43 | 4.38 | 4.62 | 293.7 | 0.000181 | |
Hard | 3.22 | 7.72 | 8.10 | 386.5 | 0.000188 |
