Tangent of the original thread "comma.ai would you do it.."

Cash Traylor

Cash Traylor

Well-Known Member
Greetings all,
I don't want to go over the original thread mentioned in the title regarding comma.ai and the comma two etc. I was curious what the answer would be, if the solution did not include the windshield camera and hanging hardware like the comma.ai option, currently, does (and other similar OP solutions with regard to the Clarity ADAS camera limitations and housing)?

So, if you could achieve this "driver assisted" function with hardware that is, for all practical purposes (once "finalized"), completely hidden from view with a "factory" appearance similar to a high quality car audio installation, would that be a "thing?"



The laptop etc is for development, the final system would be completely hidden and only be a couple hundred in hardware (likely under $500 for DIY).

Just wondering what the interest would be, no specific plans yet.

Cheers,

Cash
 
Personally, I am quite disappointed by the factory driver assist functions of the Clarity...
  1. Adaptive Cruise Control is my favorite, but it often delays too long for my taste when traffic slows down (even at longest following distance), and it always delays too long when returning back to the set speed.
  2. Regular Cruise Control (really ACC, but with no vehicle to 'follow')- very poor control of speed with any kind of hills and valleys. Lags well below setpoint when encountering a hill until it finally catches up. Pretty good maintaining downhill speed using regen. Works a little better in Sport Mode, but I don't typically use Sport.
  3. LKAS - I feel that it wanders too much, even under ideal conditions. I never use it.
  4. Road Departure - Very frustrated right away with false alarms. Shut it off and never looked back.
  5. Collision Avoidance - I hope I never need it, but happy it is there. A few minor false warnings, but hasn't been too annoying.
  6. Brake lock - OK, but haven't gotten in the habit of using it regularly.
Although Driver Assistance has a long way to go, and perhaps 'comma.ai' could help with some of these, I don't think I could bring myself to undertake that kind of a serious modification to a ~$30K vehicle. Too many risks without a major reward.

I was very optimistic about autonomous driving before I bought this car. I thought this would illustrate a real stepping stone toward greater autonomy. After seeing the 'current' mainstream vehicle technology in this car, I now see just how far away we are. I have become much more pessimistic about the time frame for full autonomy. It will certainly happen, but probably not even with my next vehicle (hope to keep this until 10 years old).

I would be curious as to your assessment of what your proposal could offer regarding these specific deficiencies (or other things that you may be aware of).
 
Just to be clear, I am NOT talking about autonomous driving or anything close to that. This is just a WAY better version of LKAS and ACC. I am a pilot, I use a "real" autopilot and still prefer to often hand fly, that's why I became a pilot. I also like to drive, have no intention of ever owning a "self driving car" or getting in one, anytime in the foreseeable future (until they make me, kicking and screaming). As to your points, I agree. However, most of what is wrong with the Honda Clarity (and a lot of Hondas) is the anemic EPS. This has been fixed, add to that a far more advanced algorithm for following the road (not just the lane) and you get a better driver "assist" system. This would never take you from point A to point B automatically, only make the longer commutes less fatiguing. Honestly I don't even know if I would like or use it. However, I bought this car as a driver and tinkerer - and I like making "improvements" versus waiting on the next year model. I will likely work on porting this to my Clarity, if possible. In this video he is specifically demonstrating the EPS firmware mod increasing the available torque by a factor of two (very similar to Toyota's EPS torque curve with slightly higher PWM).



This is not headless, so not what I am talking about from an implementation case (phone on windshield and wires hanging to side), but demonstrates the improved capability of this more open source developed platform. I too plan to keep my car for 10 years, and this allows me to improve the ADAS systems, as the technology improves (software) over time, versus being stuck with the 2018 version of Honda's LKAS/ACC.

Cheers,

Cash
 
Last edited:
@iHack

Welcome to the forum! I am curious if you noticed any improvement or degradation in the OEM ADAS system after flashing the EPS firmware? The author of that patch (assuming you are using wirelessnet2's work) believes it works fine without OP on just factory. I have no reason to doubt wirelessnet2 on this, as the dude is smart. However it often bears out that you need more than your own tests to avoid confirmation bias. Have you driven the car with the OEM LKAS active on the new EPS firmware?

Thank you for posting your information! I am still working on my setup but it is going more slowly since my workload changes during covid... May have to wait until summer for any real progress.

Cheers,

Cash
 
Nothing seems to have changed when I use the original camera with lkas and acc, but yeah, we'll need a bigger sample size if we want to know if anything is different.
 
I forgot to mention why I am even concerned about that. I am pursuing RP (raspberry pilot), not OP (OpenPilot or Comma.ai's implementation). No one is yet doing this... unfortunately as I am not that smart so this is an uphill curve for me. I am very fortunate as the Gernby, the primary Dev for RP is actually not far from me at all. Covid has ruined my immediate plans to meet up and talk about integrating and testing RP on the "unique" Clarity! Ugh... The big deal for me is that if I cannot get RP to work, I am not pursuing the project with Comma or OP as a "fall back plan." I certainly understand it is a vast improvement over the OEM ADAS, however it is just not something for me. So, if I decide to go the path of RP, and in the shake downs try EPS to dial the training in, and it fails - I wanted to be reasonably certain that things will still work. Since "downgrading" the firmware is not currently an option as far as I know, and Honda has not published a firmware update for the Clarity - and beyond that, no one really knows what will happen if the OEM tries to flash over the modded firmware version. If it brinks and still under warranty - well, Honda already has a firm stance on Honda Hack regarding warranty - I think we would be out the money.

Thanks again for your feedback and info!

Best wishes and welcome again!

Cash
 
Cash Traylor said:
Greetings all,
I don't want to go over the original thread mentioned in the title regarding comma.ai and the comma two etc. I was curious what the answer would be, if the solution did not include the windshield camera and hanging hardware like the comma.ai option, currently, does (and other similar OP solutions with regard to the Clarity ADAS camera limitations and housing)?

So, if you could achieve this "driver assisted" function with hardware that is, for all practical purposes (once "finalized"), completely hidden from view with a "factory" appearance similar to a high quality car audio installation, would that be a "thing?"



The laptop etc is for development, the final system would be completely hidden and only be a couple hundred in hardware (likely under $500 for DIY).

Just wondering what the interest would be, no specific plans yet.

Cheers,

Cash
Click to expand...

I would love to upgrade my adaptive driving and similar features. I came from a Tesla to a clarity and it is a hard transition. I don't currently want an option like comma ai because of the aftermarket look of a big camera on my front window and would love to explore other options.
 
JElectric said:
I would love to upgrade my adaptive driving and similar features. I came from a Tesla to a clarity and it is a hard transition. I don't currently want an option like comma ai because of the aftermarket look of a big camera on my front window and would love to explore other options.
Click to expand...
With the Clarity's run coming to an end, I doubt any adaptive driving software developers will be turning their attentions toward this low-volume car. Due to the small number of Clarity cars on the road, I was surprised that comma included the Clarity in their portfolio in the first place.
 
Actually, comma.ai does not support the Clarity. The OpenPilot code required for the Clarity is "aftermarket" and downloadable from GitHub. And the comma 2 isn't really that obtrusive. You can snug it up close to mirror like any other dashcam. See my YouTube video here, for example.
 
You must log in or register to reply here.
Back
Top