Front-back ambiguity in azimuth estimates on the Nearby Interaction app with DWM3001CDK

Wireless Connectivity Ultra-Wideband
1

I have been struggling with the problem of front-back ambiguity in azimuth estimates on UWB-equipped iPhones when communicating with DWM3001CDK sensors and I still have not found a solution. What I mean is if a DWM3001CDK sensor is behind an iPhone, i.e., the sensor is out of the iPhone’s [-90, 90] degree field of view, there are 3 ways the Nearby Interaction app can behave:

  1. no azimuth readings because the sensor is out of the phone’s field of view, i.e., the absolute true azimuth is over 90 degrees;
  2. there are azimuth readings but they wrap around. For example, if the true azimuth is 150 degrees, the app might say the estimated azimuth is 30 degrees. The app is right that the sensor is to the right but it can’t tell that it’s behind, so it reports that the sensor is in front. I have tried using z-coordinates from qorvoDevice.uwbLocation.direction, i.e., z-coordinates should be negative when the sensor is in front and positive when the sensor is behind, but unfortunately it’s not always accurate, so the improvement wasn’t significant;
  3. there are azimuth readings and they are more or less right, i.e., the app can detect that the sensor is behind.
    The first two scenarios are unfortunately most common.
    Based on my research, usually CIR data is required to resolve the front-back ambiguity, but it’s not available on the app. Could you please help me solve this problem? I don’t mind if the app can’t provide any azimuth estimates beyond the field of view, it’s actually better than misleading wrap-around estimates. I’ve done experiments between iPhones using the Nearby Interaction app, and the app just stops reporting azimuths if the true absolute azimuth is above around 55 degrees. Also, if possible, could you explain how azimuth estimation works? Perhaps it will help me understand how to address the issue. I have noticed that if I start with the iPhone facing the DWM3001CDK sensor and then turn around by 180 degrees, azimuth estimates tend to be much better. I wonder why that is the case. Does the app use IMU to adjust the estimates? Thank you in advance, I would really appreciate your help. By the way, this is not for commercial purposes, I need this for my research experiments.