Significant Risks
This week we made progress on creating smoother localization calculations, improving the stability of our location estimates by using distance averages, 4 anchors instead of 3, and using gradient descent (decreasing the volatility of a “bad” distance measurement).
Through our testing this week, we have stumbled upon an issue regarding the configurations of our anchors. Currently, each anchor is configured as an initiator. From our understanding, this means that if a set of anchors is sufficiently far from another, they may choose to create a separate network of anchors. The problem arises in our tag device, which can only read distance measurements from a single network at a time. However, we believe it is possible to resolve this issue by reconfiguring our network of anchors to only have one initiator.
Additionally, the IMU we are using seems to have a few systematic inaccuracies when we are reading from it. Ifeanyi is currently working on resolving these issues with calibrating the IMU.
Changes to System Design
This week we made improvements to both the webserver as well as the localization system. We have found an issue with our network of anchors; however, we believe that reconfiguring the anchors will allow them to function the same way we originally envisioned. Because of this, there was no need to make any updates to the system design.
Schedule
We would like to continue improving the localization system for the following week. Firstly, we would like to make improvements to our network of anchors and possibly try decreasing some of the latency present in our localization system getting positional updates. Additionally, we would like to finally incorporate the IMU’s orientation data. Finally, the navigation system needs to be worked on to display the path between the user to rooms.
An updated schedule with our current progress can be found here.
Progress
Here is a video of our localization system posting updates to the user’s phone as we walk down the HH A-level main corridor. We see a significantly more stable localization than the previous week. Please ignore the orientation of the arrow; the tag is generating random angles (they are there for a visible indication of the update frequency).
Here is a video showing our browser with more clarity (using randomly generated data). It showcases the higher update frequency and decreased latency as a result of using the websockets instead. Additionally, the video displays the indicators present for when the tag is not connected.