TABLE 3.
Aspects | Analytical modeling | Supervised learning* | Reinforcement learning | |
---|---|---|---|---|
Human Intervention (Single robot) | Model derivation | ✓ | ✗ | ✗ |
Parameter tuning | ✓ | ✗ | ✗ | |
Data collection | ✗ | ✓ (offline) | ✓ (online) | |
Training | ✗ | ✓ | ✓ | |
Generalization (Same prototype) | Model derivation | ✗ | ✗ | ✗ |
Parameter tuning | ✓ | ✗ | ✗ | |
Data collection | ✗ | ✓ (offline) | ✓ (online) | |
Training | ✗ | ✓ | ✓ | |
Dependence on data | Low | high | high | |
Online refinement | ✗ | ✓ | ✓ | |
Adaptability to un-modeled disturbances | ✗ | ✓ | ✓ |
Cells marked with gray shading indicate advantages.