3D pose estimation solution based on time of flight (TOF)

[Mali.1]

This topic aims at the implementation of a self-localization solution for a probe moving through hollow spaces of parts or machines of interest. Here, a small TOF-sensor (time of flight) is to be used and its raw output data to be processed and analyzed. The sensor is given, but the entire pipeline starting from acquiring measurement data, followed by pre-processing steps, 3D model information matching und real time results representation as well as providing interfaces for further pose information utilization need to be developed. The concept shall address certain given requirements. Achievable accuracies of the system are to be evaluated.

can anyone suggest me the starting point for this ? recommended hardware and software with which i can make a prototype ? Thanks

[John E KVAM]

Oh boy, that's a tall order. The problem is that the ToF from ST sends out a cone of light and the distance you get back is an averaged distance to the target. In an complex enclosed space, I fear that your interior will be so complex that you won't be happy with the answer.

let me know how long the ranges are likely to be and I can comment further.

• john

[Mali.1]

thanks for your reply and support. The maximum range will be 500mm. The shape of the object will be rectangular box or the circular shape of a part for which i will obtain i point cloud.

[John E KVAM]

our best short-distance sensor is still the VL6180. It claims 60cm, but in real life situations 40cm is realiable. The trouble is that dull targets might not return enough photons to get a reliable answer beyond 40cm.

It's main advantage is that it's linear down to about a cm.

I think I'd buy the P-nucleo-VL6180A1 evaluation kit and really put the sensor through every test you can think of.

Get the data and see if it works for you.

Try it on everything you can think of.

and if you buy the VL6180-Satel to go with it, you can attach some lead wires and fit the small sensor into any tight space you can think of.

Doing this way you get a lot of test data for only a few dollars and a few hours.

Good luck,

• john