Hyundai: Car Port for Kona 2022

[sunnyhaibin]

Checklist

• added entry to CarInfo in selfdrive/car/*/values.py and ran selfdrive/car/docs.py to generate new docs
• test route added to routes.py
• route with stock system: 3bd70ed1a93771dc|2023-09-19--19-16-45
• route with openpilot: 0099bdb24d82951b/00000005--c38d940b04
• harness Type: Hyundai O
• architecture: CAN

Link to all physical parts during installation

• comma 3X
• car harness kit
  • Hyundai O

Thanks to the community 2022 Hyundai Kona owner big_bull_ (Discord) and johnnathanwood.

Requires

• Hyundai: Second alternative steering limits panda#2038

Supersedes commaai/openpilot#30025

[github-actions]

Thanks for contributing to opendbc! In order for us to review your PR as quickly as possible, check the following:

• Convert your PR to a draft unless it's ready to review
• Read the contributing docs
• Before marking as "ready for review", ensure:
  • the goal is clearly stated in the description
  • all the tests are passing
  • include a route or your device' dongle ID if relevant

[sunnyhaibin]

More Ecu.transmission firmware: 0099bdb24d82951b/00000001--582f1c0a5d

[jyoung8607]

At a glance it looks okay, presuming the lateral accel magnitude tweak was correct, but that's hard to judge with the data available. The test routes are present, but I think they were probably driven on the higher limits, plus torqued wasn't running to judge how the car was performing, plus some actuator output structures have been refactored.

Can you procure a new test route for this one? Have them drive on this branch, with the proposed limits, and take some curves to see where it maxes out.

[sunnyhaibin]

At a glance it looks okay, presuming the lateral accel magnitude tweak was correct, but that's hard to judge with the data available. The test routes are present, but I think they were probably driven on the higher limits, plus torqued wasn't running to judge how the car was performing, plus some actuator output structures have been refactored.

Can you procure a new test route for this one? Have them drive on this branch, with the proposed limits, and take some curves to see where it maxes out.

@johnnathanwood can you install and update here when you get a chance?

[sunnyhaibin]

New test route: 0099bdb24d82951b/00000005--c38d940b04

[github-actions]

This PR has had no activity for 9 days. It will be automatically closed in 2 days if there is no activity.

[adeebshihadeh]

So this ALT_LIMITS scheme has a unique failure mode, where a simple bug can silently allow more torque than expected. If the wrong safety mode is selected for any other supported car, the worst case is the RX checks are invalid (so no actuation) and any TX is blocked (since there's overlap).

I already wasn't super happy about this with the original ALT_LIMITS, but this car seems to have even more torque.

Net, I think it's better to get this merged as dashcam than not merge at all, such that it's detected and disabled. We'll have to revisit this at some point to un-dashcam it.

[sshane]

@adeebshihadeh these cars are not dashcammed due to the line below...

ret.dashcamOnly = candidate in {CAR.KIA_OPTIMA_H, }

[johnnathanwood]

I run one of these — a gas Kona 2022 (the ALT_LIMITS_2 / 170 Nm car, dashcamOnly today) — and I've been logging it against the current learned tune. Toward the un-dashcam revisit noted here, and jyoung8607's ask for a route that takes curves near the limit, here's what I have so far, stated with its limits up front.

• ~67,376 actively-steering frames on one learned seed (latAccelFactor 2.84 / friction 0.087). Post-clamp commanded torque peaked at 170/170 Nm; 0 frames over 170 across all 67,376, and at the 170 ceiling only 0.3% of the time. The 170 is held independently by both the car-side limiter (apply_driver_steer_torque_limits, STEER_MAX 170, up 2 / down 3) and the panda TX check (HYUNDAI_LIMITS(170, 2, 3)).
• Live torque learner converged: latAccelFactor 2.747 ±0.033 (3.3% off seed), friction 0.083 ±0.002 (4.9% off seed), over 21/38 liveValid drives. Lateral-accel tracking error 0.043 m/s² mean; actual/desired gain 1.006.
• Steer faults: 0 permanent, 166 transient (stating both).

Honest gaps: this is only ~2.0 engaged hours, and envelope coverage is 14/16 cells — I don't yet have hard cornering (>3 m/s² lateral accel) at 40–60 mph or 60+ mph. So it's data toward the question, not a claim the envelope is fully exercised. I'm happy to go collect the missing cells — curves near the limit, the kind of route jyoung8607 asked for — if that would help the revisit.

One thing I wanted to check before doing anything beyond that: I take it the intended direction (per #3026 / #2072) is to move torque cars off the ALT_LIMITS-style tier flags rather than keep tuning them, so the flag becomes a transition detail. I'd rather pull in that direction than propose anything of my own. As a small interim guard, I opened #3523 to add ALT_LIMITS_2 coverage to test_alternate_limits, so the conservative tier can't silently regress in the meantime.

Separately, and unrelated to the safety limits: I'm also planning a normal tuning PR for this car's native lateral torque params (it currently borrows the EV Kona's). That's a tuning change only.

Mostly I want to make sure I'm helping in the direction you already intend. Glad to pitch in on the Hyundai safety cleanup in #1942 if that's useful.