Collaborative Optimization via Apprenticeship Scheduling (COVAS)

Organization: MIT Lincoln Laboratory
Co-Developer(s): Sponsor: Scott Orosz, Deputy Program Manager for Electronic Warfare Programs, Office of Naval Research
Year: 2018

The Collaborative Optimization via Apprenticeship Scheduling (COVAS) algorithm is a highly customizable, adaptive, machinelearning optimization pipeline, developed for Navy self-defense. In addition to the current military applications, COVAS has already been employed successfully in the fields of hospital logistics and triaging, vehicle routing, and supply chain management. This algorithm is the first and only algorithm to perform real-time ship defense by learning rules of thumb and heuristics from human experts. Rather than interrupting these experts with prompts during real or simulated battles, COVAS learns by observing them as they demonstrate the desired task. These strategies are leveraged within a mathematical optimization framework to provide solutions that are better and computed more rapidly than human experts could produce. COVAS efficiently reasons how to best allocate defenses within the computational search space. Then, COVAS scales the power of the human expert to provide real-time solutions for problems larger than a single human expert or team could manage.