Policy Smoothness Evaluation on Backflip
0.031Action Smoothnessreward 0.1
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| reward 0.1penalty_weight=0.12026.02 | 0.031 | 2.8 | 111.2 | |
| FF+Jac Penregularization=Action Jacobian Penalty2026.02 | 0.042 | 4 | 109.6 | |
| LPNtype=Linear Policy Net2026.02 | 0.061 | 8.7 | 140.8 | |
| reward 0.01penalty_weight=0.012026.02 | 0.068 | 11.8 | 132.9 | |
| No Regregularization=None2026.02 | 0.148 | 26.6 | 168.5 | |
| Lipschitzregularization=Lipschitz constraint2026.02 | 0.195 | 33.5 | 170.8 |