Author: Yu Hang He, Email: hey6@unlv.nevada.edu
Date Last Modified: 11/18/2019
Last week, I started working on a HoloLens application to mark a way-point in user's physical space and direct Spot to move toward that waypoint. The challenge in this task is the transformation of coordinate system between HoloLens and Spot. My initial attempt at frame transformation was not successful. However, after some help from Blake, I was able to formulate the required matrix transformation.
In summary, the HoloLens and Spot have separate coordinate system. The goal is to transform the position and orientation of waypoint in HoloLens to the Spot's coordinate system and control Spot to move to that position. The HoloLens/Unity uses a left-handed coordinate system with y-axis point upward while Spot uses a right-handed system with z-axis upward. The transformation matrix of waypoint w/ to HoloLens, T_hm, was corrected after multiply by a left-to-right hand matrix and a rotation matrix to point axes in the correct orientation. The resulting matrix was inversed to obtain a transformation of HoloLens w/ to Spot, T_sh. The subsequent waypoint matrices are post-multiplied to the T_sh to obtain the transformation matrix of waypoint in Spot's coordinate frame.
Currently, an initialization process is required where a waypoint needs to be placed on the Spot's coordinate origin to successfully compute the T_sh matrix. The matrix transformation was successful tested. However, after a few waypoint commands, the drifting in position became apparent. This can be attributed to the limitation in the accuracy of sensor from HoloLens and Spot. Spot uses dead reckoning to track its current position while HoloLens uses internal IMU to track its movement. Overtime, the errors compound to create large drifting.
In addition, I updated my programs to accommodate for the new Spot update. The update opened some more features to the users. Unfortunately, it does not like that the autonomous navigation system is exposed through their API. The only way to setup AutoWalk mission is through their controller App. Due to some mistake from BD, I did not gain access to their App until late Friday afternoon, I will leave the testing to next week.
Finally, Matt organized a meeting between BD and electrical engineers at Fremont for a scanning session at a substation. The BD reported that they have tested Spot at a substation before and there was no problem with its operation. The scanning session will take place next Thursday.
Next week, I will test the capability of Spot's autonomous walking feature. In addition, I want explore some of the new feature as well, such as specify the position and weight of additional payloads. Finally, I need to prepare for the scanning session in Fremont next week.