To test crash barriers, we crash cars into them at a specific speed and angle. The car must hit a designated impact point within plus or minus X inches.

I'm wondering if it is possible to build a system to automate the steering process, using sensors (or whatever else) without access to the test car's on-board ECUs. The first idea which came to mind was a computer vision system which controls the vehicle's steering wheel mechanically - essentially a mechatronics system.

Are there any existing systems that do this that I can take a look at? I think the hardest part may be the accuracy, since the car must be steered into a specific spot on the test barrier. The speed is taken care of by another system.

  • 1
    $\begingroup$ system to automate the steering process ... a rail $\endgroup$
    – jsotola
    Sep 25, 2023 at 15:09
  • $\begingroup$ Sounds like a job for a crash test dummy :) $\endgroup$
    – Solar Mike
    Sep 25, 2023 at 19:23

2 Answers 2


Yes, we have steering robots that attach a motor to the steering wheel. Trouble is that interferes with the airbag, so an important part of the test is invalid. The proper way around this is to correlate your FEA to the ACTUAL test path, and then run it again at the perfect test path. Crash FEA is good enough these days that we can use it to calibrate the airbag sensors. Rails can't be used as they provide an unrealistic boundary condition.

  • $\begingroup$ Thanks for the input. We are actually testing the barriers themselves, not the cars, so the lack of airbags is not an issue. Question about the steering robots: is it a custom solution? Are you guys using any off the shelf parts (and/or software) you can point me to? Are you hooking into the car's ECUs? Additionally, how accurate would you say the system is? $\endgroup$ Sep 26, 2023 at 0:11
  • $\begingroup$ Sure, off the shelf, anthony best abdynamics.com/en/products/track-testing/driving-robots/…. For straight line work you scarcely need any steering wheel torque, 5 Nm would be wildly excessive, 2 would be marginal. Talk to them about accuracy, I've only used them open loop, ie defined steering wheel input. So you need to hook them into a path following computer vision/driver system, the software for which is available in Matlab, which would be overkill. I'm guessing you are talking about offset frontal crash? $\endgroup$ Sep 26, 2023 at 3:32
  • $\begingroup$ Sorry I forgot, no this is standalone, no access to ECU needed. Someone who can code the EPAS could use that directly, rather than a seperate robot, but that is not routinely done. $\endgroup$ Sep 26, 2023 at 3:35

if it is possible to build a system to automate the steering process, using sensors (or whatever else) without access to the test car's on-board ECUs.

Clearly it's possible, several systems exist to self-steer Automated Guided Vehicles. Wire guided, GPS-based, Lasers, inertia systems.

The better question is why you would want to. FMEA would probably lead you away from complicated systems when the effects of failure are so extreme. AGV's travel about 4 miles per hour, which limits the damage they can do.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.