UIST 2018 was just held in Germany a week ago. I am curious that how it is like this year so I was planning to go through it quickly at the beginning.
First, I found out there was still quite a lot of work related to handmade new devices and robotics. The wearable shape-proxy interface, PuPoP, looks really fun. When I was working on haptic feedback, I was so concerned about how to provide the feeling of different shapes. And they solved that, to some extent. Second, VR and collaboration both have an individual section for technical papers. That is really important. Although we don’t do research because of conferences, it is relevant to see the academic peers have similar interest. And lastly, I found a paper talking about synchronization for multiple SLAM devices. That is something in my mind for a while.
As I mentioned before, we did quite a few projects based on co-located experience. We tried external markers, like OptiTrack, however, the markers attached to the device are possible to be occluded once the amount of the users increases. Or we tried to use AR phone, like Google Tango or Lenovo, and started the application from the same spot for the rough synchronization. They are just not ideal enough.
This paper talks about doing coordinate synchronization for mobile AR, without using any markers or external attached components. Before reaching the detail of the approach, I really like the way how it referred to the previous work. To do the synchronization for multiple coordinate systems from various devices, shared map, anchor-based, pre-scanning for the whole environment, vision-based tracking and learning-based pose estimation. I felt I can go through some of them too because I am looking for a not very cumbersome approach for the co-located experience. It does not have to be really smart, but it needs to be easy for configuration and registration. Like putting a marker on the ground is really a good start to me. Or design a device holder and put the device at the exact same location to launch the application sounds doable too.
This paper explained very clearly on the core equation part. See from the featured image, it pictured two users working randomly at their time zones. One Ultra-Wide Bandwidth (UWB) based
distance measurements module was attached to the phone so the distance between the two devices are able to be measured. Based on the transformation information in the local coordinate system at different timestamp and the distance between them. It is solvable to calculate the transformation between the two coordinate systems. What’s more, to minimize the freedom, more constraints are added here for this optimization problem, including the start height, the range of y during the synchronization and so on. The only pity in this paper is that they did not evaluate it with precise tracking system like OptiTrack or Vicon, which is their future plan. Still, looking forward to it.