feat(mit): per-model torque coefficients for effort and command#86
Open
JafarAbdi wants to merge 5 commits into
Open
feat(mit): per-model torque coefficients for effort and command#86JafarAbdi wants to merge 5 commits into
JafarAbdi wants to merge 5 commits into
Conversation
Add joint_torque_coefficients(arm_type) with per-model torque constant k and gear ratio b. get_joint_efforts decodes true joint torque (current*k*b) instead of the SDK base-piper current*k. direct_scaling_factors takes arm_type and scales commanded torque per model relative to base piper (identity for base, b_base/b on >= S-V1.8-3, 1/b on <= S-V1.8-2).
Greptile SummaryThis PR adds per-model torque handling for Piper arms. The main changes are:
Confidence Score: 5/5This looks safe to merge.
Important Files Changed
Reviews (5): Last reviewed commit: "fix(mit): scale commands by per-model k*..." | Re-trigger Greptile |
![greptile-apps[bot]](https://avatars.githubusercontent.com/in/867647?s=60&v=4){kind=small}
run_gravity_compensation.py defaulted GravityCompensationModel and PiperInterface to base piper, so a piper_h on >= S-V1.8-3 mis-scaled J5 (1.0 instead of 1/1.7). Add --arm-type and thread it into both.
Comment on lines
+643
to
+653
| k, b = joint_torque_coefficients(self._piper_arm_type) | ||
| arm_msgs = self.piper.GetArmHighSpdInfoMsgs() | ||
| return [ | ||
| arm_msgs.motor_1.effort / 1e3, | ||
| arm_msgs.motor_2.effort / 1e3, | ||
| arm_msgs.motor_3.effort / 1e3, | ||
| arm_msgs.motor_4.effort / 1e3, | ||
| arm_msgs.motor_5.effort / 1e3, | ||
| arm_msgs.motor_6.effort / 1e3, | ||
| currents = [ | ||
| arm_msgs.motor_1.current, | ||
| arm_msgs.motor_2.current, | ||
| arm_msgs.motor_3.current, | ||
| arm_msgs.motor_4.current, | ||
| arm_msgs.motor_5.current, | ||
| arm_msgs.motor_6.current, | ||
| ] | ||
| return [currents[i] * k[i] * b[i] / 1e3 for i in range(6)] |
There was a problem hiding this comment.
get_joint_efforts() now depends on the configured arm type, but scripts/generate_samples.py still creates PiperInterface(args.can_port) with the implicit PIPER value and has no arm-type option. Recording samples from a Piper H/L/X therefore applies base-Piper coefficients—for example, Piper H J2 uses 1.18125 * 4 instead of 1.65375 * 2.5. The saved effort targets are then wrong for gravity-compensation training, so a correctly configured runner can still deploy a model trained on incorrect torques. Propagate arm type through sample generation.
direct_scaling_factors used base(b)/arm(b), which over-scaled joints whose torque constant k also differs from base piper (piper_h J2: 1.6 instead of 1.143). The gravity model is calibrated in the current*k*b feedback frame, so commands must scale by the k*b ratio. J4/J5 are unchanged (equal k); base piper unchanged.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Add joint_torque_coefficients(arm_type) with per-model torque constant k and gear ratio b. get_joint_efforts decodes true joint torque (currentkb) instead of the SDK base-piper current*k. direct_scaling_factors takes arm_type and scales commanded torque per model relative to base piper (identity for base, b_base/b on >= S-V1.8-3, 1/b on <= S-V1.8-2).