Building the Next Generation of Trackers
I previously worked on an airborne ground surveillance radar tracker for one of the large defence manufacturers. I found the technology very interesting. It combines clever algorithms, real world experience, working within the limits of the computing platform resources available, and knowing how to hold up under heavy system load.
Many of the design decisions made on that particular project were conservative: aimed at increasing the odds of delivering within time and budget, at the price of delivering a less capable tracker than might have been possible. This was the right approach for that project, but I was left wondering what could be achieved if the other unexplored paths were taken.
In particular, the following were of interest:
- Using global coordinate systems, to build a unified picture incorporating multiple sensor feeds.
- How to expand the total number of tracks that the system can cope with.
- Look at the use of modern geospatial databases and data sources to improve tracking accuracy.
- Apply modern AI and machine learning approaches to improve the tracking performance.
- How to use the parallel processing capabilities of modern CPUs and GPUs.
Since completing the original design, I pondered a few ideas, and sketched out approaches for the above, but for a long time never committed to the effort of actually building it. That has changed, and a prototype is nearing completion. Once done, I’ll make it available online for people to try.