vulp
vulp copied to clipboard
Robot/simulation switch for the mjbots stack
Should BulletInterface also log the base state?
In our current setup, workflows take too long to finish. We should check what can be cached across builds to speed up the process.
Currently we only log sources (NB: will be renamed to "sensors") when a controller is started. The rationale for it is: https://github.com/tasts-robots/vulp/blob/ab9e555d7937a4990378ec83ce689a9a298226b7/vulp/spine/Spine.cpp#L180 Sensors don't need configuration, so we should run...
As we talked in tasts-robots/upkie#104, could you give more context on how to obtain ground truth information from the simulator to extend Upkie's observation dictionary? Particularly the height of a...
Position commands are currently forwarded unchecked to the actuators, but we can forward them e.g. at spine configuration time. A safer behavior would be to: - Store and enforce position...
Currently Bullet position tracking is implemented as $\tau = \tau_f + K_p (p^* - p) + K_v (v^* - v),$ while moteus [position control](https://github.com/mjbots/moteus/blob/main/docs/reference.md#theory-of-operation) is implemented as: ``` acceleration =...
When the spine is in neither of the following three states: - Stop - Shutdown - Act Then ``Spine::cycle_actuation`` will send the previous command: ``` if (state_machine_.state() == State::kSendStops ||...