Machine Reasoning and Learning Workshops III and IV Kickoff Wen Masters Office of Naval Research Code 311 Wen.Masters@navy.mil (703)696-3191 A Desired Future State ISR Information (hours-days latency) Commercial Information (minutes-hours-days) C2 Information (mins-hours latency) Information Space Real Time Fire Control Information (fractions of secs latency) 2 Mission-Focused Autonomy The control of networks of diverse sensors designed to seek, understand and shape battlespace in complex, uncertain environments with the following capabilities: – Independently understands commander’s intent regarding missions and/or objectives – Understands battlespace (events, activities, entities, networks of entities, etc) based on data it has collected or has access to via other sources – Assesses information to determine shortfalls and threats in the battlespace relative to commander’s intent – Optimally (resources, time, significance) determines/evaluates options for courses of actions and self-tasks specific components of network sensor(s) to resolve shortfalls and threats – Executes tasks as it adapts to changing conditions and is self-aware and teamaware Vision: Develop autonomous control that intelligently understands and reasons about its environment relative to its objectives and independently takes appropriate actions – Intelligently alerts proper forces or commanders to engage critical threats 3 Machine Reasoning and Learning Overarching Challenges and Goals Overarching Naval/ Marine Corps Goals • Reducing manpower while integrating and interpreting large volumes of data from sensors of multiple types with HUMINT, other –INT, and open source data, formulating hypotheses, and plans to resolve uncertainty, imprecision, incompleteness, and contradiction to achieve commander’s intent • Focusing the available manpower on cognitive tasking instead of orchestrating the collection and processing of data • Increasing the operational tempo and shaping the battlefield through increased automation while operating in a dynamic uncertain environment • Dispersing geographically while locally achieving the desired effect • Achieving disaggregated functionality while minimizing resources and providing flexibility Overarching S&T Challenges • Automation and robustness of the overall system that can also interpret the current state of knowledge in the context of the mission • Defining the information needs of the mission and creating courses of action to improve the state of knowledge • Computing with quantitative and qualitative data that are uncertain, imprecise, incomplete, and contradictory • Understanding how much error/uncertainty can be tolerated within the holistic system while achieving a correct inference/decision • Defining context and employing context • Representations for data and knowledge • Defining clutter and the background • Establishing fundamental performance limits for the ability to detect, track and identify objects; for establishing relationships, activities, and events • Aligning data sets with disparate signatures in space and time • Developing an information infrastructure that supports distributed large data sets Machine Reasoning and Learning – Questions ONR Thinks About Workshops I and II Workshop III Workshop IV address System Integration Success for these Workshops • Identify critical issues and high payoff approaches that will enable Machine Reasoning and Learning related to Action/Reaction and Integration • Community – Create awareness of the problem and issues – Formation of an interdisciplinary community that addresses the key issues • Information that supports ONR planning – Potential redirection of existing program goals – 6.1 and 6.2 Broad Agency Announcements targeted at specific issues – MURI Topics – SBIR/STTR – Other mechanisms