Real-Time Density Tracking Distributed Kernel Method for Realizing Collaborative AI at Extended Ranges

Modern sensing systems are increasingly more diverse (e.g., type, function, design, etc.) and decentralized. These systems require new methods to efficiently merge sensor data across distributed networks when centralized data fusion centers are impractical due to communications constraints, methods such as the novel Density Tracking Distributed Kernal Fusion (DTDKF) approach provided by MTRI. While MTRI’s approach has shown promising results, additional research is needed to discover the optimal strategy to develop it in C/C++ and for it to function in real-time on a low-SWaP embedded platform. Therefore, this effort will focus on leveraging MTRI’s DTDKF method to build a low-SWaP embedded solution. Specifically, DornerWorks will use the Python source code developed by MTRI (that combines a consensus and innovation framework with kernel density estimation to track complex non-observable features of targets over a high dimensional space), translate it to C/C++, optimize it in a C/C++ build environment, and create an executable for the Qualcomm QCS6490/QCS5430 chipset. Once an application is developed, the ESCAPE II dataset, which was used in MTRI’s research, will be used for validation. At the end of phase I, DornerWorks will deliver the validation results and a C/C++ form of the DTDKF method in a package manager that is targeted to initially run on a Qualcomm QCS6490/QCS5430 chipset. Once in this C/C++ deployable form, additional embedded platforms can be targeted in the future.