Job opportunity: looking for someone to help with a research project
 

Hey everyone, I'm working on a new company building aftermarket autonomous driving features, starting with adaptive cruise control. More on this in a different thread in the coming weeks/months.

I'm looking for someone who can help compile a list of all the vehicles built in the last 10 years with their production numbers (this part is pretty straightforward), then link them by common cruise control parts, and then look up wiring diagrams on All Data (which is accessible from many public libraries). It requires a bit of creativity and general car knowledge (so, basically everyone on this forum), but we can work together on it for the first few so it's clear what to do. I figure this will be a 10-20 hour job, but there could be more after that. $20/hour.

Anyone interested?

Not interested in the job. But very interested in the aftermarket adaptive cruise control :D

I'll make you a deal - if I haven't started a thread with a full run down on what we are doing by the end of the month, you can choose any image for my signature which I'll use until said thread has begun.

So the goal is to look for the most common and simplest to wire/integrate?

Yeah, that's a big part of it. The other part is to figure out the % of cars sold in the last 10 years we can integrate with by the method we've designed, and to identify cars we might need to develop a different integration method for.

Along the way, I want to try and get an estimate for the total % of cars on the road that have regular cruise control. However, I haven't found a decent method to put together that estimate yet.

Does adaptive cruise control need to tie into the brakes, or is that supposed to be on the driver? (I have never used/been in a car with adaptive cruise control on).

The job sounds fun, but I am currently working full time and doing two 200 level Mech Tech classes. Oh well.

Also, expressing interest in the final product for use in my '90 Miata and '11 Fit.

I use Alldata at work. Their wiring diagrams for a number of brands are next to useless. BMW has much better diagrams else where on the internet, and apparently Mercedes is next to impossible to find any info on.

Adaptive CC would have to be tied into brakes, meaning you'd need a car that has a servo motor on the brake system. Otherwise all it could do is let off the gas when someone cuts in front of you. Kinda dangerous territory to be getting in to.

The first version won't tie into the brakes. We're using computer vision and deep learning, which allows for more sophisticated predictions (as in, it starts slowing down as soon as someone starts moving into your lane). It'll also have a braking warning to help the driver take over quickly when braking is required. This allows us to have a generalized solution that's still far better than regular cruise control, and then we can add braking (and steering) for cars that have the appropriate equipment from the factory.

Why vision instead of sensors? This sounds like it's (slippery sloping) to become Open Pilot and will face the same NHTSA issues.

A big of a longer discussion, but:
1. I wouldn't call what Comma experienced to be NHTSA "issues". It was a pretty reasonable request for information.
2. Those won't come into play until we are level 3+ autonomous. Regular cruise control is level 0 and adaptive cruise control is level 1.
3. The vision/sensors question doesn't really influence the NHTSA question. If you built a level 3+ system using radar/lidar you'd still be under the same obligation to respond to such requests, and be governed by whatever regulations are forthcoming.

To answer the question about vision vs. sensors, vision has far more potential and is limited only by software (this is an oversimplification, but true enough). With software updates we can add significantly to the capabilities of the system, rather than being hardware limited with a particular sensor.

Sign me up for beta testing.

Would love to install this on the Miata for those crowded track days.

Last call before this goes out to Craigslist...

I don't mean this as thread crapping, as your project sounds interesting, but as someone who has worked on a number of serious R&D projects:

:bowrofl:

I assume you mean 10-20 hours a day, for a year.

Joe, it's super easy and profitable. He's just the first one to realize it.

:rofl::rofl:

I can see how you'd have that reaction. What I didn't mention originally is that I bought databases with every make/model/trim combination for cars built since 2000 and whether they have cruise control, and vehicle production numbers for the same period.

Ok, 6 months then.

I'm just having a hard time getting my head wrapped around something...

I'm trying to imagine myself as the owner of a company and chief architect of a product which, if it fails, will almost certainly result in both criminal charges of negligent homicide and a whole slew of of civil tort claims. And I'm imagining that there's a part of the project which is crucial to the design, and which will only take 2-3 days to complete. And I'm trying to rationalize paying some random dude on Craigslist $20 an hour to do it, rather than just putting in a long weekend and cranking through it myself.

It could be business case type research, rather than actually attempting to map out the technical parts of integration?

Like Joe said, the liability exposure seems like it would be a huge hurdle.

This