Microsoft Technology Licensing, LLC

Kence Anderson

Kingsuk Maitra

Kinshumann Kinshumann

Garrett Patrick Prendiville

Edilmo Daniel Palencia

The techniques disclosed herein enable systems to enhance the resilience of autonomous control systems through a fault-tolerant machine learning architecture. To achieve this, a fault-tolerant machine learning agent is constructed with a selector agent, a nominal agent, and a redundancy agent which is a multidimensional lookup table. The fault-tolerant machine learning agent extracts state data from an environment containing a control system and various components. The nominal agent and the redundancy agent generate actions for application to the control system based on the state data which are provided to the selector agent. Based on an analysis of the state data, the selector agent can detect a failure condition. In the event of a failure condition, the selector agent deploys the action generated by the redundancy agent lookup table to resolve the failure condition and restore normal operations.

Redundant machine learning architecture for high-risk environments

The techniques disclosed herein enable systems to integrate aging awareness into machine learning agents for management of control systems. To achieve this, the machine learning agent extracts a set of states and an aging model from the control system and derives an aging term from the aging model. Based on the states and the aging term, the machine learning agent determines a set of actions that are applied to the control system. Subsequently, the machine learning agent can extract a new set of states to analyze the efficacy of the set of actions. This is achieved through an optimality function that quantifies the success of the set of actions by calculating an optimality score. The calculation is based on both a current performance of the system and a future drift or degradation of system components. Through many iterations of the action set, the machine learning agent can optimize system operations.

Aging aware reward construct for machine teaching

The techniques disclosed herein enable systems to measure the long-term reliability of machine learning agents prior to deployment at a control system. This is achieved through analysis of control system component specifications to determine a useful lifespan of the components such as projected failure rate, hours continuous operation, and so forth. The system can derive parameters for the machine learning agent to interact with the components such as action frequency and action range. From the component lifespan, action frequency, and action range, an accelerated test procedure is constructed to evaluate the reliability of the machine learning agent. From executing the accelerated test procedure, a reliability score can be calculated for the machine learning agent.

Growing community of inventors

Berkeley, CA, United States of America

Kence Anderson