Multi-Target Tracking—
Naval Air Systems Command (NAVAIR)
Navy surveillance radars generate the data needed to detect and assess targets in their operational space and engage them if they prove to be threats. Translating this data into useful information, connecting just the relevant dots into useful tracks, however, is a complex challenge, one that’s exacerbated when networked systems are following multiple targets.
In addition to accurate tracks, operators want to know what these potential threats are doing, where they are going, and who they are. They want to know their capabilities and their intentions—and they want to know as much as possible as quickly as possible.
Minimizing the Trade-Offs
The challenge for creators of tracking algorithms is minimizing the trade-off between accuracy and the computational speed. They need to develop algorithms capable of analyzing thousands of data points generated by single or networked platforms showing a multitude of potential targets and produce accurate, smooth tracks in real time.
To meet the needs for efficiency, Navy tracking systems like the E-2C Hawkeye use one model when a target is maneuvering and another when it is not. Determining when a target is maneuvering is a difficult problem—but it’s a crucial one to solve. When the wrong model is used, tracking systems lose continuity, tracks begin crossing, and operators’ situational awareness is impaired.
Beyond Maneuver Detection
CCRi employed explorative data analysis techniques, constructed models that used the data in novel ways, and built an improved tracking application. In addition to traditional parameters like range and azimuth, the models incorporated the radar's Doppler speed measurement to increase tracking accuracy.
We began by devising algorithms that significantly improve maneuver detection—enabling the tracking system to switch more appropriately between models—and then went on to develop a tracking algorithm that successfully uses a single motion model on a variety of targets, whether they are maneuvering or not. We developed simulations that show that both approaches lead to significant improvement in the performance of the E-2C Hawkeye.

