update DH names + notes for MPC output curvatures #24701

ClockeNessMnstr · 2022-06-01T20:19:11Z

"current_curvature" is not the correct description of what the MPC is outputting in it's curvature_ego state.
The MPC is integrating it's free variable, curvature_rate, such that curvature[0] is actually the desired_curvature before any delay.
inversely: the curvature_rate_desired is the desired rate of change to the setpoint and not the actual curvature rate.

If we were to set the initial curvature = measured curvature in the MPC initiation these names would be correct.
This was possibly how it was initially set up but the nomenclature here is now confusing.

"current_curvature" is not the correct description of what the MPC is outputting in it's curvature_ego state. The MPC is integrating it's free variable, curvature_rate, such that curvature[0] is actually the desired_curvature before any delay. inversely: the curvature_rate_desired is the desired rate of change to the setpoint and not the actual curvature rate. If we were to set the initial curvature = measured curvature in the MPC initiation these names would be correct. This was possibly how it was initially set up but the nomenclature here is now confusing.

haraschax · 2022-06-03T17:41:18Z

selfdrive/controls/lib/lateral_planner.py

@@ -79,6 +79,9 @@ def update(self, sm):
                     y_pts,
                     heading_pts)
    # init state for next
+    # mpc.u_sol is the desired curvature rate from x0[3]. 
+    # we integrate here so that x_sol[:,3] is the desired curvature for lat_control
+    # if we set x0[3] = measured_curvature the output would be desired_curvature_rate


I don't understand this whole comment block. Integrate or Interpolate.

Wanted to clarify it's interpolating over the integral of rate, since rate is really the MPC's "output" state.

This is better maybe:

# mpc.u_sol is the desired curvature rate given x0 curv state. # with x0[3] = measured_curvature, this would be the actual desired rate. # instead, interpolate x_sol so that x0[3] is the desired curvature for lat_control.

Yeah that makes sense!

haraschax · 2022-06-03T17:41:31Z

LGTM besides that comment block

* update names + notes for MPC outputs "current_curvature" is not the correct description of what the MPC is outputting in it's curvature_ego state. The MPC is integrating it's free variable, curvature_rate, such that curvature[0] is actually the desired_curvature before any delay. inversely: the curvature_rate_desired is the desired rate of change to the setpoint and not the actual curvature rate. If we were to set the initial curvature = measured curvature in the MPC initiation these names would be correct. This was possibly how it was initially set up but the nomenclature here is now confusing. * more notes * match * Clarify commaai#1

* update names + notes for MPC outputs "current_curvature" is not the correct description of what the MPC is outputting in it's curvature_ego state. The MPC is integrating it's free variable, curvature_rate, such that curvature[0] is actually the desired_curvature before any delay. inversely: the curvature_rate_desired is the desired rate of change to the setpoint and not the actual curvature rate. If we were to set the initial curvature = measured curvature in the MPC initiation these names would be correct. This was possibly how it was initially set up but the nomenclature here is now confusing. * more notes * match * Clarify commaai#1 (cherry picked from commit b215d61)

* update names + notes for MPC outputs "current_curvature" is not the correct description of what the MPC is outputting in it's curvature_ego state. The MPC is integrating it's free variable, curvature_rate, such that curvature[0] is actually the desired_curvature before any delay. inversely: the curvature_rate_desired is the desired rate of change to the setpoint and not the actual curvature rate. If we were to set the initial curvature = measured curvature in the MPC initiation these names would be correct. This was possibly how it was initially set up but the nomenclature here is now confusing. * more notes * match * Clarify #1

* Laikad: cleanup fetching orbits (commaai#24759) * Seperate prediction orbits from regular observation orbits and download them efficient * Cleanup * clean and update laika * Fix test * Fix test * Fix checking pos fix * space * update DH names + notes for MPC output curvatures (commaai#24701) * update names + notes for MPC outputs "current_curvature" is not the correct description of what the MPC is outputting in it's curvature_ego state. The MPC is integrating it's free variable, curvature_rate, such that curvature[0] is actually the desired_curvature before any delay. inversely: the curvature_rate_desired is the desired rate of change to the setpoint and not the actual curvature rate. If we were to set the initial curvature = measured curvature in the MPC initiation these names would be correct. This was possibly how it was initially set up but the nomenclature here is now confusing. * more notes * match * Clarify #1 * Laikad: Use process for parsing orbits (commaai#24769) * Use Process instead of Thread to fetch orbits * small refactor * Cleanup * UI: remove DM icon (commaai#24771) * PlotJuggler: add torque control layout (commaai#24726) * add torque control PJ layout * less custom transformation * Use curvature, less noisy * remove that * Honda: rename Bosch harness to Bosch A * GM: interface cleanup (commaai#24774) * Move all defaults above models * Remove reduntant/defaults * make minEnableSpeed common * Update selfdrive/car/gm/interface.py Co-authored-by: Adeeb Shihadeh <adeebshihadeh@gmail.com> Co-authored-by: Shane Smiskol <shane@smiskol.com> Co-authored-by: Adeeb Shihadeh <adeebshihadeh@gmail.com> * IsoTpParallelQuery: handle response pending (commaai#24724) * handle response pending * match commit * remove total timeout, just keep track of individual response timeouts * fix * add back total timeout * this isn't reliable enough * keep track of pending responses to print warning * tx_addr is (addr, subaddr) * debug only query hyundai import time reponse pending no cache all cars no timeout to test before * Revert "debug" This reverts commit abe9cfc. * always print pending always debug * Only debug * Update selfdrive/car/isotp_parallel_query.py * remove variable only for debugging * Install readme (commaai#24776) README: better install on device instructions * VW MQB: Add FW for 2020 Volkswagen Tiguan (commaai#24777) * Kia Ceed: fix eps ECU type (commaai#24767) should be eps * Add missing MIRAI ESP f/w (commaai#24779) * Add missing ESP f/w `@JaySheikh#7759` 2021 Mirai DongleID/route b2c2b20082938bed|2022-06-08--05-04-03 [confirmed working] * formatting Co-authored-by: Shane Smiskol <shane@smiskol.com> * Laikad: kill process correctly. Fixes CI (commaai#24780) Kill process correctly * LaikaD: Fix offline handling (commaai#24781) * Test handling no internet correctly. * Clean * Comment * remove del * fullframe DM model (commaai#24762) * get log * simplify two nonsense * not needed * libyuv is a joke * clean up * try small * fast but not bad * working * clean up driverview * simplified * thats mirrored * smol * tweak * ref is screen * w/ ee * update camera model * no if TICI * start * update pose thresh * less cpu more dsp * new libyuv * new snpe * add files * test * should be fast * update out len * trigger test * use master snpe * add cereal * update cereal * refactor parsing * missing ; * get * wrong type * test model * use driver data * 10829278-72fe-4283-a118-2cef959ce174/1550 * no pf * adapt driverview * ; * rhd learner * update libyuv buildi x64 * ad4337e * remove blink slack * test * no * use toggle * b16 * fix for nv12 * 5b02cff5 both * update test * update cereal * update cereal * update cereal * v2 packets * revert libyuv * no / * update snpemodel * ; * memcpy * fix test * use toggle in driverview * update power * update replay * Revert "update replay" This reverts commit 1d0979c. * update model ref * halve cpu * fake 8bit onnx runner * same thresh as report * cereal master Co-authored-by: Comma Device <device@comma.ai> * bump opendbc (commaai#24790) * Honda Bosch: fix openpilot longitudinal controls mismatch (commaai#24788) * fix bosch controls mismatch * bump * bump panda * VW MQB: Add FW for 2020 Volkswagen Tiguan (commaai#24795) * build checks do not rely on each other (commaai#24783) * Add missing CHR engine f/w (commaai#24797) `@papifrio#5178` 2019 Toyota C-HR ICE ("nodsu" TSS-P) DongleID/route ade67da77ee74b16|2022-06-09--00-36-15 * Lexus: NX Hybrid 2020 support (commaai#24796) * LEXUS: Missing esp and engine FW versions in values.py DongleId: 09ae96064ed85a14 | testRoute: 09ae96064ed85a14|2022-01-10--01-57-37 | * Make a new platform for NX Hybrid TSS2 * Update releases Co-authored-by: Shane Smiskol <shane@smiskol.com> * some release notes * EyeSight standard on all 2019+ Ascent and Forester (commaai#24799) * on all 2019+ Ascent Forester * update docs Co-authored-by: Adeeb Shihadeh <adeebshihadeh@gmail.com> * fullframe DM: flip RHD yaw to use matching thresholds * Toyota: Add missing esp fw version for RAV4_TSS2 (commaai#24793) add fwVersion for RAV4_TSS2 add b'\x01F15260R302\x00\x00\x00\x00\x00\x00' to list of Ecu.esp FwVersions for RAV4_TSS2 * Revert "UI: remove DM icon (commaai#24771)" This reverts commit 4d836c6. * Hyundai: Tucson 2021 support (commaai#24791) * add support for 2021 Hyundai Tucson base configuration and fingerprints * fix fcw again and don't use mando radar for tucson * last fixes for working tucson * Apply suggestions from code review * add to cars * increase steerRatio * missing car info * one platform * rename * add min enable speed * update releases Co-authored-by: bluesforte <harry3b9@gmail.com> Co-authored-by: Adeeb Shihadeh <adeebshihadeh@gmail.com> * docs: adjust steering torque thresholds (commaai#24798) * Prius has good steering control now * 1.5 is the threshold for good torque * get back down there, hondas * half stars for 1.0-1.5 * show number of cars * try bigger * emphasize tiers balanced * Add half star * Update ref_commit * bypass HDA2 dashcam with file (commaai#24803) Co-authored-by: Comma Device <device@comma.ai> * Honda: use common imperial unit message (commaai#24786) * CAR_SPEED should be on all cars * bump opendbc * clean this up too * it's an or * clean up * comment * add PyQt (commaai#24811) * count events: add simple camera debugging * Revert fullframe DM model (commaai#24812) * Revert "fullframe DM: flip RHD yaw to use matching thresholds" This reverts commit 2ac6931. * Revert "fullframe DM model (commaai#24762)" This reverts commit d6c07a6. * revert cereal * Add missing fw versions for 2019 Sonata (commaai#24814) * put cereal on master * Car interface: set max lateral torque from table (commaai#24789) * json * better naem * Read from table * formatting and default to nan * Generate docs * Read from table * this should be the same * Prius v is full * test we always set the tunes correctly add to release files * Set for all cars Set for all cars * Revert tuning changes Revert tuning changes * remove that * fixes * update ref commit for new maxLateralAccels Co-authored-by: Shane Smiskol <shane@smiskol.com> * Fix Lexus NX Hybrid 2020 engine ecu (commaai#24817) Wrong address in engine ecu. * compatibility docs: auto-generate star descriptions (commaai#24809) * Auto-generate star descriptions * Need this for the website * And this * required changes to make the website generation work * better names * Revert "better names" This reverts commit be7dbbb. * simpler * Misc torque control fixes (commaai#24801) * Fiction compensation should be based on error * Update refs * Add deadzone * update ref * Honda carcontroller and signal cleanup (commaai#24806) * common signals * move stopping * space * clean up * bump opendbc * test for the strip_bz2_extension method (commaai#24826) * CI: fix docker push for base image * pylint: add PyQt5 support * simulator: run simulator test in ci (commaai#24691) * run simulator test in ci * block navd process * block ui * fix jenkins * build docker * remove tty * remove tty for carla * detach carla_sim * more retries * only build once * add more time for bridge * cleanup * use qt offscreen * expose to docker * block ui * use new dockerimage * fix * from ubuntu20.04 * install curl * add ssh * add locales * noninteractive * syntax * use base * smaller image * add git + git lfs * kill carla * run in parallel * fix missing agents * default agent? * little cleanup * default doesn't work * not in ci * fix path * fix path * new msg Co-authored-by: Adeeb Shihadeh <adeebshihadeh@gmail.com> * Move selfdrive/hardware/ to system/ (commaai#24725) * move hardware to system/ * fix mypy * jenkins: remove unnecessary workstation clean * bump cereal * set AGNOS from /AGNOS file * CI: add build job for latest Ubuntu (commaai#24637) * CI: add build job for latest Ubuntu * source * source env * scons cache * cache pyenv * fix key * source * Move a bunch of stuff to system/ part 3 (commaai#24829) * move swaglog.py * timezoned * logmessaged * version.py * fix linter * acados build script improvements for mac (commaai#24634) * add Darwin build w/ universal2 libs * add rust for acados rebuilds * just build script fixes Co-authored-by: Adeeb Shihadeh <adeebshihadeh@gmail.com> * Ford: disable radar for now (commaai#24832) The newer Ford vehicles require a different radar parser. * Updated CARLA to v0.9.13 (commaai#24575) * Updated CARLA to v0.9.13 * pipenv lock Co-authored-by: Adeeb Shihadeh <adeebshihadeh@gmail.com> * misc jenkins fixups (commaai#24840) * bump cereal * remove that * pull cl image * lil docker cleanup * Ford: remove Fusion DBC from release (commaai#24841) * jenkins: lock simulator * SConstruct: set AGNOS from /AGNOS * Revert "SConstruct: set AGNOS from /AGNOS" until Jenkins devices are updated This reverts commit 84c741e. * Laikad: Use filter for correcting measurements (commaai#24824) * Update laikad. * Update log error * ui: skip texture frame copy (commaai#24700) * ui_blit working * simpler and working * more believable that it's real * working on device * build on pc * use hardware pc * reduce cpu usage * yuv conversion to EGL * move everything to cameraview * some cleanup * more cleanup * init array * init images with std::map * dont destroy images * do destroy images Co-authored-by: Comma Device <device@comma.ai> * Laikad: process executor to fetch orbits (commaai#24843) * Use ProcessPoolExecutor to fetch orbits * update laika repo * Minor * camerad: remove unused SubMaster (commaai#24844) * navd: speed limits only when localizer is valid (commaai#24845) * Toyota Camry TSS2: update torque control params (commaai#24819) Use updated accel and friction values for TSS2 Camry * ui.py: update for nv12 * Revert "thermald: consider pmic in component temp management (commaai#24708)" This reverts commit c8c21ba. * Couple more cars to torque tune (commaai#24848) * try sonata on torque tune * Couple known cars to torque control * fix * more fix * Honda Bosch long: fix ACC fault (commaai#24851) Fix Honda bosch long Co-authored-by: redacid95 <jonathanc.olivier@gmail.com> Co-authored-by: redacid95 <jonathanc.olivier@gmail.com> * Move couple toyotas to torque table values (commaai#24849) * Move couple toyotas to torque table values * Dont set for all cars yet * Dont regress docs * update ref * Move RAV4 2022 out of bronze * pin protobuf and hypothesis versions (commaai#24852) pin hypothesis and protobuf * bump opendbc (commaai#24853) * process replay: clean up common code (commaai#24855) * regen and process replay clean up * test_fuzzy actually uses fingerprint hardcoding fix * revert * revert * this can be a url or path so just print full variable * remove weights fixup with new SNPE (commaai#24254) * remove weights fixup with new SNPE * Update ref Co-authored-by: Comma Device <device@comma.ai> Co-authored-by: Harald Schafer <harald.the.engineer@gmail.com> * GM: add support for vehicles with manual parking brakes (commaai#24766) Switch to general park brake signal * little zookeeper fixes (commaai#24861) * little zookeeper fixes * bump that up * bump up modeld power draw threshold * Torque control: low speed boost (commaai#24859) * Make very low speed more aggressive * Less extreme low speed boost * Update ref * regen & process replay: support no disengage on accelerator (commaai#24850) * ACC on if enabled != 0 * small regen clean up and add HONDA3 * fixes * revert unneeded changes * not used * just alt exp Co-authored-by: redacid95 <jonathanc.olivier@gmail.com> * Laikad: Cache orbit and nav data (commaai#24831) * Cache orbit and nav data * Cleanup * Cleanup * Use ProcessPoolExecutor to fetch orbits * update laika repo * Minor * Create json de/serializers Save cache only 1 minute at max * Update laika repo * Speed up json by caching json in ephemeris class * Update laika * Fix test * Use constant * UI: new set speed design, show speed limits (commaai#24736) * basic US design * place based on center position * fix typo * eu sign without rounded box * same as steering wheel icon * proper rounded bottom for eu sign * add border * proper placement/sizes * needs to be semi bold * color changes * only when engaged * move helpers into util.h * Fix MAX placement * only change color when at least 5 over * implement override state * pixel perfect spacing around us sign * add speed limits to release notes * Laikad: Remove bearingDeg from message (commaai#24864) * Remove bearingDeg from message. * Push cereal * Commit cereal * ui: add black color define (commaai#24866) * ui: advanced network settings fix button colors and sizes (commaai#24846) * ui: button pressed color add * match colors from other buttons Co-authored-by: Willem Melching <willem.melching@gmail.com> * networking.cc: remove resolved TODO * selfdrive/ui/qt/util.cc: missing include (commaai#24867) * laikad: calc_pos_fix numpy implementation (commaai#24865) * Replace posfix with gauss newton method * Cleanup * Check if glonass is in the list * Fix * also return residual * Add residuals * Update selfdrive/locationd/laikad.py Co-authored-by: Willem Melching <willem.melching@gmail.com> * Cleanup Co-authored-by: Gijs Koning <gijs-koning@live.nl> * Update model ref * Revert "Update model ref" This reverts commit efc8aa0. * Rocket league model (commaai#24869) * dd9a502d-c8e2-4831-b365-804b0ae0739d/600 80041070-d276-4fed-bdb9-0075e5442908/420 * no elementwise op * 9dabf0f-2e60-44bf-8d3a-d20a74aca072/600 ae746590-0bb5-4a16-80db-15f02d314f03/300 c4663a12-b499-4c9b-90dd-b169e3948cb1/60 * explicit slice * some copies are useful * 1456d261-d232-4654-8885-4d9fde883894/440 c06eba55-1931-4e00-9d63-acad00161be0/700 af2eb6ba-1935-4318-aaf8-868db81a4932/425 * 154f663e-d3e9-4020-ad49-0e640588ebbe/399 badb5e69-504f-4544-a99e-ba75ed204b74/800 08330327-7663-4874-af7a-dcbd2c994ba7/800 * set steer rate cost to 1.0 * smaller temporal size * Update model reg * update model ref again * This did upload somehow * Update steer rate cost Co-authored-by: Yassine Yousfi <yyousfi1@binghamton.edu> * HKG: simplify Kia K5 compatibility (commaai#24810) LKAS/LFA is standard on the K5 * thermald: fix panda dropout when we miss a pandaStates (commaai#24870) immediate fix for "panda dropout" * VW MQB: Populate stock ACC standstill flag (commaai#24877) * VW MQB: Populate stock ACC standstill flag * it really do be like that sometimes * Mazda: fix resume spam at standstill (commaai#24876) * Fix Mazda resume spam at standstill * one line * Revert "one line" This reverts commit 30c6504. * Laikad: More logging and use last_pos_fix for correcting (commaai#24868) Use last available pos_fix for correcting measurements. Improve logging measurements * ui: disable controls unresponsive alert on PC * HCA7 + ramp speed * bump openDBC * bump openDBC * HCA7 + ramp speed * bump panda * bump panda * HCA7 tuning from edgy Co-authored-by: Gijs Koning <gijs-koning@live.nl> Co-authored-by: ClockeNessMnstr <locke.dftc@gmail.com> Co-authored-by: Adeeb Shihadeh <adeebshihadeh@gmail.com> Co-authored-by: Shane Smiskol <shane@smiskol.com> Co-authored-by: Jason Shuler <jshuler@gmail.com> Co-authored-by: HaraldSchafer <harald.the.engineer@gmail.com> Co-authored-by: Jason Young <46612682+jyoung8607@users.noreply.github.com> Co-authored-by: Erich Moraga <33645296+ErichMoraga@users.noreply.github.com> Co-authored-by: ZwX1616 <zwx1616@gmail.com> Co-authored-by: Comma Device <device@comma.ai> Co-authored-by: AlexandreSato <66435071+AlexandreSato@users.noreply.github.com> Co-authored-by: George Hotz <72895+geohot@users.noreply.github.com> Co-authored-by: Griffin Christenson <36981741+griff12@users.noreply.github.com> Co-authored-by: bluesforte <harry3b9@gmail.com> Co-authored-by: Maykon Pacheco <38124066+maykonpacheco@users.noreply.github.com> Co-authored-by: Maxime Desroches <desroches.maxime@gmail.com> Co-authored-by: Andrew <andrew.tec@pm.me> Co-authored-by: Cameron Clough <cameronjclough@gmail.com> Co-authored-by: Jeroen <33349469+jeroenlammersma@users.noreply.github.com> Co-authored-by: Willem Melching <willem.melching@gmail.com> Co-authored-by: Joost Wooning <jwooning@gmail.com> Co-authored-by: redacid95 <jonathanc.olivier@gmail.com> Co-authored-by: YassineYousfi <yyousfi1@binghamton.edu> Co-authored-by: Lee Jong Mun <43285072+crwusiz@users.noreply.github.com>

* update names + notes for MPC outputs "current_curvature" is not the correct description of what the MPC is outputting in it's curvature_ego state. The MPC is integrating it's free variable, curvature_rate, such that curvature[0] is actually the desired_curvature before any delay. inversely: the curvature_rate_desired is the desired rate of change to the setpoint and not the actual curvature rate. If we were to set the initial curvature = measured curvature in the MPC initiation these names would be correct. This was possibly how it was initially set up but the nomenclature here is now confusing. * more notes * match * Clarify commaai#1

commit d066702 Author: ClockeNessMnstr <locke.dftc@gmail.com> Date: Tue Jun 7 12:41:03 2022 -0400 update DH names + notes for MPC output curvatures (commaai#24701) * update names + notes for MPC outputs "current_curvature" is not the correct description of what the MPC is outputting in it's curvature_ego state. The MPC is integrating it's free variable, curvature_rate, such that curvature[0] is actually the desired_curvature before any delay. inversely: the curvature_rate_desired is the desired rate of change to the setpoint and not the actual curvature rate. If we were to set the initial curvature = measured curvature in the MPC initiation these names would be correct. This was possibly how it was initially set up but the nomenclature here is now confusing. * more notes * match * Clarify commaai#1 commit cb82478 Author: HaraldSchafer <harald.the.engineer@gmail.com> Date: Fri May 27 10:21:02 2022 -0700 Tssp prius torque control (commaai#24669) * use llk * use steering sensor at low speed stil * Try more simple * rm prius tune * updated ref commit f1f7a0b Author: HaraldSchafer <harald.the.engineer@gmail.com> Date: Tue May 24 17:52:33 2022 -0700 Rerevert torque control (commaai#24649) * Revert "Revert torque control (commaai#24565)" This reverts commit 9f8b037. * Move tune out of car specific stuff * Update ref commit commit 041c255 Author: HaraldSchafer <harald.the.engineer@gmail.com> Date: Tue May 17 19:00:57 2022 -0700 Revert torque control (commaai#24565) * torque reversal start * Fix carmodel tests * Update ref * update ref * Elif is better than if commit 14f5dab Author: HaraldSchafer <harald.the.engineer@gmail.com> Date: Fri May 20 13:50:47 2022 -0700 Latcontrol torque: integrator need not be reset (commaai#24606) slow integrators need not be reset commit 9221f3f Author: ClockeNessMnstr <locke.dftc@gmail.com> Date: Fri May 13 19:52:20 2022 -0400 LatControlTorque: clean up class variable (commaai#24526) * move to super * no class variable * there's CP * whitespace * drop CI from latcontrol super * Revert "drop CI from latcontrol super" This reverts commit 9218273. Co-authored-by: Shane Smiskol <shane@smiskol.com> commit efb4dbd Author: HaraldSchafer <harald.the.engineer@gmail.com> Date: Fri May 6 22:17:21 2022 -0700 Latcontrol torque: fix integrator induced ping pong (commaai#24458) * Latcontrol torque: fix integrator induced ping pong * Reset on disengage since unwind resets anywayh * Might be overkill * rm whitespace * update ref commit d105ae2 Author: HaraldSchafer <harald.the.engineer@gmail.com> Date: Mon May 2 14:42:18 2022 -0700 Latcontrol torque: update tuning (commaai#24357) * Little more chill * Update ref * Update refs commit 0ce2b58 Author: Shane Smiskol <shane@smiskol.com> Date: Thu Apr 28 00:42:52 2022 -0700 Clean up controllers (commaai#24340) * clean up lat controllers * pass CP once * sort commit 47b162c Author: Shane Smiskol <shane@smiskol.com> Date: Wed Apr 27 01:18:05 2022 -0700 LatControlTorque: fix deadzone and missing steer saturated warning (commaai#24294) * fix steer saturated alert and deadzone * fix and formatting * update refs * update refs * not needed * remove commit 51a7ef2 Author: HaraldSchafer <harald.the.engineer@gmail.com> Date: Wed Apr 20 11:26:24 2022 -0700 Latcontrol torque: max_torque rename (commaai#24265) commit 33e00f6 Author: HaraldSchafer <harald.the.engineer@gmail.com> Date: Tue Apr 19 19:34:31 2022 -0700 Lateral torque-based control with roll on TSS2 corolla and TSSP rav4 (commaai#24260) * Initial commit * Fix bugs * Need more torque rate * Cleanup cray cray control * Write nicely * Chiiil * Not relevant for cray cray control * Do some logging * Seems like it has more torque than I thought * Bit more feedforward * Tune change * Retune * Retune * Little more chill * Add coroll * Add corolla * Give craycray a good name * Update to proper logging * D to the PI * Should be in radians * Add d * Start oscillations * Add D term * Only change torque rate limits for new tune * Add d logging * Should be enough * Wrong sign in D * Downtune a little * Needed to prevent faults * Add lqr rav4 to tune * Try derivative again * Data based retune * Data based retune * add friction compensation * Doesnt need too much P with friction comp * remove lqr * Remove kd * Fix tests * fix tests * Too much error * Get roll induced error under 1cm/deg * Too much jitter * Do roll comp * Add ki * Final update * Update refs * Cleanup latcontrol_torque a little more

* update names + notes for MPC outputs "current_curvature" is not the correct description of what the MPC is outputting in it's curvature_ego state. The MPC is integrating it's free variable, curvature_rate, such that curvature[0] is actually the desired_curvature before any delay. inversely: the curvature_rate_desired is the desired rate of change to the setpoint and not the actual curvature rate. If we were to set the initial curvature = measured curvature in the MPC initiation these names would be correct. This was possibly how it was initially set up but the nomenclature here is now confusing. * more notes * match * Clarify #1

adeebshihadeh assigned haraschax Jun 1, 2022

ClockeNessMnstr added 2 commits June 1, 2022 16:29

more notes

54dd33f

match

621143d

ClockeNessMnstr changed the title ~~update names + notes for MPC outputs~~ update DH names + notes for MPC output curvatures Jun 1, 2022

adeebshihadeh added the cleanup label Jun 2, 2022

ClockeNessMnstr mentioned this pull request Jun 3, 2022

latmpc-imp #24573

Closed

haraschax reviewed Jun 3, 2022

View reviewed changes

Clarify #1

bb0ffa1

haraschax merged commit b215d61 into commaai:master Jun 7, 2022

ClockeNessMnstr deleted the patch-1 branch June 7, 2022 17:52

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

update DH names + notes for MPC output curvatures #24701

update DH names + notes for MPC output curvatures #24701

ClockeNessMnstr commented Jun 1, 2022

haraschax Jun 3, 2022

ClockeNessMnstr Jun 3, 2022 •

edited

Loading

haraschax Jun 7, 2022

haraschax commented Jun 3, 2022

update DH names + notes for MPC output curvatures #24701

update DH names + notes for MPC output curvatures #24701

Conversation

ClockeNessMnstr commented Jun 1, 2022

haraschax Jun 3, 2022

Choose a reason for hiding this comment

ClockeNessMnstr Jun 3, 2022 • edited Loading

Choose a reason for hiding this comment

haraschax Jun 7, 2022

Choose a reason for hiding this comment

haraschax commented Jun 3, 2022

ClockeNessMnstr Jun 3, 2022 •

edited

Loading