0

I am looking for a indoor positioning/2d motion tracking system for small robot cars (1:8 scale RC cars). We want to use the system as a ground truth for the development of autonomous driving applications, so looking for an accuracy of a few millimeters. The testing area is around 10x10m. The cars are running ROS (Robotic Operating System), so an existing implementation would be nice.

A known solution is a OptiTrack motion capture system, but with a cost of >10kEuro this is way above our budget of around 1.5kEuro. I have also been looking in using HTC Vive trackers with OpenVR, but I am not sure if this is a reliable solution.

Any idea would be very welcome!

Mbrs
  • 1
  • From a corner of the room, attach a string potentiometer to the RC car on a swivel mount point, and an angle sensor to the string. That is the cheapest, quickest, and lowest latency method, as it's purely mechanical & electrical. Otherwise, simulation would be a good option; unity if dissatisfied with gazebo. – JWCS Jun 03 '20 at 16:47
  • Else, if you have multiple good cameras, and you can trigger them in sync or publish the time of capture, you can manipulate a PnP alg [(related)](https://stackoverflow.com/questions/18637494/camera-position-in-world-coordinate-from-cvsolvepnp?rq=1), or stick a big [aruco marker](https://docs.opencv.org/master/d5/dae/tutorial_aruco_detection.html) on it. – JWCS Jun 03 '20 at 16:47
  • I don't know if it's a mm accuracy, but on the cheap side we've used (a) stereo camera(s) and large fiducials to good effect (Intel realsenses seems to be popular these days for sub 200USD, with self publishing calib mat, but perhaps with not enough res for your case). I'd recommend getting something working quick, if you have that kind of budget, and it'll help you see what the pain points are of your system. In this case, it could be either/both timing issues (esp while moving) or resolution issues (we just can't quite see where it is). – JWCS Jun 03 '20 at 16:53

0 Answers0