Motion capture results and a new direction…

The anvil prepared for the day’s work!

As cold and crisp morning dawned in Gothenburg the crew gathered at the RISE Interactive Studio ready and eager for a productive day of testing. Today (or well, when writing this it was actually three weeks ago…) was a big day, since we were going to test our tracking system together with a real blacksmith’s oven in order to see if that would disturb the measurements. The tracking itself had also been improved by our Augmented Reality-sages Gunnar and Mike, and should now posess such features as automatic calibration, making the initial setup a breeze. Or so we thought…

At first we pondered the possibilities of setting up the equipment inside the studio in our VR-area, where we had performed the testing the previous time. But since we figured that maybe the rest of the studio would not appreciate listening to the “DING”-sound of a hammer hitting a vice for several hours we decided to move all of our equipment outside. This included two computers, the whole HTC-vive tracker system, the anvil and tools for blacksmithing, and the heater. It might be worth pointing out that the heater is not what most people might concieve when they think of a heater in a forge (a huge oven with an open fire), but a modern counterpart based on induction. So no HE-MEN were required in the moving of the heater, a strong blacksmith master named Gustav proved to be enough.

With all of our equipment outside Gunnar, Mike and Niels went about the business to calibrate the trackers so that we could get ready for testing. This however turned out not to be quite the walk in the park that we had anticipated. Instead, initially nothing really went our way. When doing the calibration in a static position everything seemed to be fine. But when we started to move around with the camera it seemed like that the tracking of the camera started to drift, resulting in the markers on the anvil moving when we were moving around with the camera. We kept trying until we got a calibration result that was just good enough so that we could actually start with the testing. And just when everything seemed fine and dandy and Gustav was about to start the induction heater, well the heater didn’t work anymore.

Time to get creative! So Gustav brought forth a portable propane heater and built a small oven from a couple of bricks and was able to heat the glorious piece of metal to at least somewhere near the optimal temperature. So finally we were able to test our tracking system and generate traces! Too bad they really didn’t act as we wanted them to. Instead they pretty much went haywire and as soon as Gustav struck his hammer onto the anvil they completely went wild. So Gustav tried to mount the tracker unit on a saw instead, but we still experienced the same problems.

So what went wrong? Regarding the problems with calibration we believe that it might have been an issue with the stability of the tripods that we used for the HTC light house laser trackers (the laser sending tracking boxes that the HTC-system uses to track the controllers using reflected laser pulses and trigonometry), that when they swayed even just a little bit the calibration was ruined.

Regarding the problem with the traces, our analysis was that the built in IMUs (Inertial Measurement Units) that got a very high acceleration value when the hammer hit the anvil, and since they will always contain a small error in the measurement due to the accuracy of the accelerometers this caused the problem. When the acceleration gets very high, this error becomes proportionally larger, and since that acceleration is then integrated twice to achieve the position, this error in measurement gets even bigger resulting in the position of the tracker marker being completely off. This remains until the laser from the HTC light house towers manage to bring the tracker point back to its correct place.

So where do we go from here? It might be possible to turn off, or change the weighting function to make the IMU listen less to the signal from the accelerometers. We have also pondered the possibility of using Intel’s “Real Sense” depth cameras to create a 3-dimensional depiction of the work, but without traces this time.

But that is something for next time!