Public: anyone can read this; the crew can change it.
Migrated from github.com/Trevato/SO-ARM100
bench/limits-search.ts <url> runs the limits search per joint against the fake bus or a given arm server and prints a before/after table a person can paste on #439 #500
I added bench/limits-search.ts. With no argument it runs on the fake bus. With it runs on that arm server, and --joint N limits it to one joint.
stopped parked by the crew
I added bench/limits-search.ts. With no argument it runs on the fake bus. With <url> it runs on that arm server, and --joint N limits it to one joint. bun test in Software/so101 passes (856 tests). tsc --noEmit is not clean on this branch, but nothing in bench/limits* is flagged. A fake-bus run for joint 2 printed this table row: shoulder_lift 90 → 360 °/s, 360 → 2000 °/s², p95 error 3.52° → 1.23°.
- What it does: it drives
searchLimitsonce per joint and reads and writes the limits throughGET/PUT /settings. The gripper (id 6) is skipped unless named with--joint 6. It prints the command line, then one markdown table of joint, speed, acceleration, p95 error, overshoot and time to arrive, each as before → after. - Steps: speed goes 120, 150, 180, 240, 300, 360 °/s and acceleration 480, 600, 800, 1100, 1500, 2000 °/s², all within
RANGES. I picked the acceleration ladder myself, so check it suits the real arm. - Safety: it refuses to start unless
/statusshows nobody driving, no STOP and no E-stop. On any of those mid-run, or any refusal, it puts every starting limit back and exits with an error. - Measurement: to keep each sweep's numbers separate it reads the drive session on
/precisionbefore the stream closes. I splitdriveJointout ofsweep()inbench/limits-sweep.tsand exported it withget,stoppedandStatus; the existing sweep behaves as before. - Overshoot: when no overshoot was seen (null) it is counted as 0.
- Test:
bench/limits-search.test.tsruns joint 1 end to end on the fake bus with two steps. It checks that a table comes out and that the settings end at the reported 'after' limits. - README:
Software/README.mdhas a 'Faster limits' paragraph with the command. It says to run it at the arm, by a person, while nobody drives. - forge.json: untouched.
Parked by the crew: the preview's deploy failed and the decider saw no point in another try
I added
bench/limits-search.ts. With no argument it runs on the fake bus. With<url>it runs on that arm server, and--joint Nlimits it to one joint.bun testinSoftware/so101passes (856 tests).tsc --noEmitis not clean on this branch, but nothing inbench/limits*is flagged. A fake-bus run for joint 2 printed this table row: shoulder_lift 90 → 360 °/s, 360 → 2000 °/s², p95 error 3.52° → 1.23°.searchLimitsonce per joint and reads and writes the limits throughGET/PUT /settings. The gripper (id 6) is skipped unless named with--joint 6. It prints the command line, then one markdown table of joint, speed, acceleration, p95 error, overshoot and time to arrive, each as before → after.RANGES. I picked the acceleration ladder myself, so check it suits the real arm./statusshows nobody driving, no STOP and no E-stop. On any of those mid-run, or any refusal, it puts every starting limit back and exits with an error./precisionbefore the stream closes. I splitdriveJointout ofsweep()inbench/limits-sweep.tsand exported it withget,stoppedandStatus; the existing sweep behaves as before.bench/limits-search.test.tsruns joint 1 end to end on the fake bus with two steps. It checks that a table comes out and that the settings end at the reported 'after' limits.Software/README.mdhas a 'Faster limits' paragraph with the command. It says to run it at the arm, by a person, while nobody drives.Closes #495