A multi-radar data fusion method for small scale features

There has been a recent push in the unmanned aircraft community to integrate unmanned aircraft systems (UAS) into the National Airspace System (NAS). Recent work at the University of North Dakota (UND) has developed a concept called the Ganged Phased Array Radar – Risk Mitigation System (GPAR-RMS) that utilizes three phased array radars to monitor a given airspace, providing a potential basis for assessing collision risk when flying an unmanned aircraft (UA) within the NAS. Under GPAR-RMS, multiple radars would have overlapping coverage of an observational airspace and provide updated information on a time scale of seconds. This increase in overall update rate for a given volume decreases the time between target detections. Fusion of the data from the individual radars in the GPAR-RMS both decreases uncertainty in target location and tracking, and also reduces the amount of information the user has to interpret. The GPAR-RMS will help to ensure a safer NAS for that observed airspace.

While radars report positions of range resolution volumes (RRVs) based on the location of the center of the radar beam, the RRVs actually possess a complex geometry. This research provides a method for fusing multiple RRVs by using spherical trigonometry to identify horizontal overlapping regions of individual RRVs. The sampling area of the RRV increases with range from the radar, creating a larger uncertainty in the position of the objects sampled (especially in cross-beam directions). In the presented fusion technique, the beam power density from an individual radar is assumed to follow a Gaussian distribution across the RRV. The overlapping region encompasses a portion of each RRV's Gaussian curve which allows weights to be calculated that can be applied to the respective RRV data value during fusion.

Identifying the overlapping region of multiple RRVs decreases the overall target positional uncertainty (especially when the individual radars possess disparate viewing angles), and identifies weights to be used in fusing the meteorological data. Simulated results indicate an average decrease in horizontal positional uncertainty around 550 to 600 meters using the assumption of one degree beam resolution in both horizontal and vertical directions. The vertical uncertainty of the fused RRV was determined by finding the vertical overlapping region of the RRV's vertical boundaries and was shown to decrease by 250 to 350 meters.

In addition to decreasing target location uncertainties, such an approach has potential for meteorological applications by providing the degree and parameters of the spatial overlap which enables one to identify both similarities and differences between fields observed with multiple radars. Such an approach could be used as an alternative first step in producing radar mosaics wherein data from multiple-radars are first combined and then spatially analyzed to a grid. Further, information regarding differences and the orientation of RRVs could be used to help recover (deconvolve) the underlying field. Finally, this technique could be used to track small scale atmospheric phenomena such as microbursts, tornadoes, etc.