[LilyPad12]

Has anyone looked into this? comma.ai – Introducing openpilot

A lot of the same functionality in Lane Assist and adaptive cruise control.

 

[02fordsvt]

Car and Driver magazine did a test a couple months ago. Check it out if you can find it.

 

[wiz043]

After a Pause, Comma.ai Delivers a Driver-Assist System That Rivals Super Cruise and Autopilot

And it could be retrofitted to millions of cars.

www.caranddriver.com

 

[roxasthenobody98]

Right now, it's a difficult question to assess. Yes, it technically does work, but the steering has a lockout that stops accepting commands after 10 seconds. On my F150, it works pretty good, aside from curves.

We are working on something where it uses Active Park commands, which don't seem to lock out. Hopefully, we can get this working for Fords

 

[SledDog]

According to the website, while it’s open source, there is no direct application for the escape. That is unless you want to code it to work yourself. Not very likely. There are people here that won’t use Forscan because it’s too technical for them.

Open source is great, but producing, verifying, and taking the liability of creating a system to work with a specific vehicle is completely different.

I know I’m not willing to beta test someone’s driver assist development software. Just look at what has happened with some of the self-driving test vehicles. Tesla may be able to defend themselves against a multi-million injury dollar lawsuit, most people can’t. And I’m sure once your insurance company finds out you are using a non-OEM software interface to control your driver assist functions, they will be willing to pony up money to cover an accident.

Besides, with the way car companies partner with companies to develop this type of software, the open source development time alone will make this add-on software obsolete before it can be deployed for profit.

Plus, at a cost probably north of the 1000.00 Comma charges for vehicles they have developed for, it will be a very small market.

and since it open source, unless you completely secure the code, it’s more open to be hacked then the OEM and it’s com network.

ford also has this habit of changing module access and the commands use. Just look at the difference between a 2017 F250 and a 2020. Or a 19 escape and a 2020 escape. This will also limit your market. Active park assist? Ok. If the vehicle has it. I know in forscan, without the modules and the appropriate configuration, the commands are a dead end. There’s probably a reason the steering is being locked out after 10 seconds. Safety comes to mind.

The car buying public maybe willing to drop 1500 on new wheels and tires, but add-on driver assist software? Not very likely. That is unless you have that type of disposable income. But if that’s the case, you won’t be buying an Escape. You’ll be buying a higher end vehicle.

 

[roxasthenobody98]

I mean, it would need to be fingerprinted on my repo and have lateral tuning done, but I bet I can get it to steer.

 

[SledDog]

Sure. The park assist has set parameters in order to work. I program for a living. And with the safety protocol, and I’m just guessing but I would program it something like this in order to access the steering control for park assist. I don’t know what all requirements may be so this would be a minimum.

1. Verify the operational status of the park assist sensors. Any sensor marked bad, no park assist
2. the vehicle would have to be stopped
3. Once stopped, It must be placed in reverse
4. The front and rear sensor on the parking side must see an opening the vehicle will fit in. This would probably be at the start of the code along with all status checks.
5. Once these conditions, at a minimum are met, allow steering commands to steering into the opening detected.
6. Only allow vehicle to accelerate to 2 mph. If speed over this limit is detected terminate park assist and immediately stop.
7. if any of the conditions are false, do not allow steering commands. And immediately terminate park assist and steering commands.
8. And most importantly do not allow non-OEM identified devices to access the system while the key is on and the engine is on. If this is detected, immediately put the vehicle into limp mode and slow to stop.

I do something like this all the time when programming PLCs when they are running safety critical systems. If the code is running, no on line edits. If the code is running all non-program inputs are ignored for safety systems. Outside intrusion of the code commands detected, ignore all outside command and set a fault alarm or immediately e-stop the program. The PLC network and processor are locked down by software and hardware key, MAC address or all of these. Or other security protocols.

Forscan can not access all the modules. some modules require the IDS in order at access them.

attempting to access the park assist will require modifying the code to disable the safety features that allow it do what it does.

I sure it’s possible, but you will have to fool so many inputs and multiple on/off, and/or and safety protocol statements. So fool the computer into think the vehicle is stopped. What else will that affect? Fool the vehicle to think it’s in reverse. That could be exciting. Fool it into thinking it’s going slow enough at access the steering, that could get really fun.

 

[roxasthenobody98]

Sure. The park assist has set parameters in order to work. I program for a living. And with the safety protocol, and I’m just guessing but I would program it something like this in order to access the steering control for park assist. I don’t know what all requirements may be so this would be a minimum.

1. Verify the operational status of the park assist sensors. Any sensor marked bad, no park assist
2. the vehicle would have to be stopped
3. Once stopped, It must be placed in reverse
4. The front and rear sensor on the parking side must see an opening the vehicle will fit in. This would probably be at the start of the code along with all status checks.
5. Once these conditions, at a minimum are met, allow steering commands to steering into the opening detected.
6. Only allow vehicle to accelerate to 2 mph. If speed over this limit is detected terminate park assist and immediately stop.
7. if any of the conditions are false, do not allow steering commands. And immediately terminate park assist and steering commands.
8. And most importantly do not allow non-OEM identified devices to access the system while the key is on and the engine is on. If this is detected, immediately put the vehicle into limp mode and slow to stop.

I do something like this all the time when programming PLCs when they are running safety critical systems. If the code is running, no on line edits. If the code is running all non-program inputs are ignored for safety systems. Outside intrusion of the code commands detected, ignore all outside command and set a fault alarm or immediately e-stop the program. The PLC network and processor are locked down by software and hardware key, MAC address or all of these. Or other security protocols.

Forscan can not access all the modules. some modules require the IDS in order at access them.

attempting to access the park assist will require modifying the code to disable the safety features that allow it do what it does.

I sure it’s possible, but you will have to fool so many inputs and multiple on/off, and/or and safety protocol statements. So fool the computer into think the vehicle is stopped. What else will that affect? Fool the vehicle to think it’s in reverse. That could be exciting. Fool it into thinking it’s going slow enough at access the steering, that could get really fun.

For stock APA, yes thats the control flow. All I need to do is isolate the PSCM, spoof the speed to 0mph, handshake, and send angle commands.

my code for this is on Github.

https://github.com/roxasthenobody98...aster-apa/selfdrive/car/ford/carcontroller.py

 

[murcod]

Just look at what has happened with some of the self-driving test vehicles. Tesla may be able to defend themselves against a multi-million injury dollar lawsuit, most people can’t.

There was a recent post on the UK Kuga forums where a new 2020 model Kuga decided to emergency brake to a stop by itself on a freeway/ highway. It was rear ended by the following vehicle. There was nothing ahead of it or any reason for it to brake. Apparently the rear brake lights didn't come on either (I think it was caught on dash cam they hadn't.)

There's no way I'd be trusting any open source type developed system. Then there's the risk of hacking with vehicles having wifi and cellular connections.

 

[roxasthenobody98]

2015 Ford F150 - openpilot Active Park Assist

Watch "2015 Ford F150 - openpilot Active Park Assist" on Streamable.

streamable.com

That's with the Active Park commands. Needs to be fine tuned some more.

 

[Jay Presley]

I'm very interested in this as well. I'd be willing to test it. Can the angle commands be sent directly to the PSCM, without the PAM?

 

[roxasthenobody98]

I'm very interested in this as well. I'd be willing to test it. Can the angle commands be sent directly to the PSCM, without the PAM?

Yes. The pscm is intercepted and the hardware sends the apa commands as if it was the Pam.

I do need to point out that my code is not technically openpilot. I got a cease and desist for using their name as they do not approve of this. But I'm allowed to modify the code as I see fit.

Once I get done with some preliminary testing, along with more safety, I'll get more people on board.