Progress for week 50 (2019)
From Robin
(Difference between revisions)
(→Budget) |
(→Accounting) |
||
Line 4: | Line 4: | ||
** initial approach with input - current upper joint position did not manage to get dynamic response in CTRNN | ** initial approach with input - current upper joint position did not manage to get dynamic response in CTRNN | ||
** changed input to network to sinusoid -> now capable of generating dynamic gaits, but not much successful walking gaits | ** changed input to network to sinusoid -> now capable of generating dynamic gaits, but not much successful walking gaits | ||
+ | [[Fil:ctrnn.png]] | ||
+ | |||
=== Budget === | === Budget === | ||
* tweaking fitness function | * tweaking fitness function |
Revision as of 11:59, 16 December 2019
Wonho
Accounting
- Have been running risi's multilayered HyperNEAT tweaking parameters to get better result
- initial approach with input - current upper joint position did not manage to get dynamic response in CTRNN
- changed input to network to sinusoid -> now capable of generating dynamic gaits, but not much successful walking gaits
Budget
- tweaking fitness function
- Implement simple CPG circuit
- Implement Novelty search - how to define behavior metrics?
- No ready-to-use package available in python
- Can map-elites be something relevant here?