Decision Support for Anomaly Detection and Recovery for Unmanned System (ADRUS)

Deployment of unmanned systems continues to expand across a wide range of missions; for example, logistics and resupply missions, force application and protection, and improving battlespace awareness. Presently these unmanned systems run at the lowest of four possible levels of autonomy, that is, in a teleoperated mode, and each system typically requires multiple operators. To address this problem of operator scalability, investigations into approaches for human supervised autonomy were called for in this SBIR. In Phase I, we took a step toward addressing this capability gap by developing a decision support system for Anomaly Detection and Recovery of Unmanned Systems (ADRUS). In particular, we demonstrated that ADRUS could successfully handle unexpected events or anomalies and replan to recover from them. Our demonstration included a proof-of-concept prototype implementation and its performance in simulated logistics and resupply missions. In Phase II, we will continue algorithmic development of anomaly detection, mission risk analysis, and replanning reasoning services to meet the performance requirements of our target transition environments. Our approaches and extensions will include methods for improving reasoning accuracies, model coverage and fidelity, as well as the ability to learn and improve knowledge models by exploiting operator interactions and decisions data. We will implement and evaluate progressively mature versions of ADRUS throughout the performance period. We will conduct repeated tests and evaluations (T&E) in simulation using realistic models of target unmanned platforms. Based on T&E, we will characterize the robustness, scalability, and coverage of ADRUS. In addition, we will evaluate the operational effectiveness resulting from human supervisory control enabled by ADRUS. For these evaluations, we will engage application subject matter experts (SME) and operators from candidate transition programs. We have initiated discussions with prime performers from target programs developing these unmanned platforms and we will develop our transition requirements accordingly.