Agent-Based Automated Algorithm Generator

Abstract : The variability of vehicles poses a great challenge on the diagnostics and prognostics for the whole fleet with a vast number of Army ground vehicle platforms. A general diagnostics/prognostics model does not exist and it is difficult to select the best algorithm from a large amount of candidate algorithms for each specific component/subsystem/system application. Therefore, it is necessary to develop a unified framework to evaluate and select the best algorithms, and further maintain the on-vehicle algorithms by updating algorithm parameters and integrating new fleet-wide vehicle data statistics and trends. To address this problem, we propose an agent-based automated algorithm generator for fleet-wide diagnostics/prognostics, which can automatically generate the most suitable algorithm(s) for each vehicle or component in the fleet from a library of light-weight diagnostic/prognostic algorithms. When sufficient fleet-wide statistics and trending information are available, the automated algorithm generator server will automatically determine whether it is necessary to update the current vehicle algorithm configuration or select a better algorithm for on-vehicle diagnostics/prognostics. To prove the concept, we used battery diagnostics as an example to demonstrate the algorithm selection & generation process, and updating capabilities in a networked agent environment.