Risked Informed Hierarchical Planning for Manned-Unmanned Systems at Operational Tempo

Background: Planning in the military context is difficult due to the complexity of the scenarios for which plans need to be synthesized and the need to operate at tempo. One main feature of military scenarios that complicates planning is the time-complexity of war games due to the combinatorial state-space and action-space explosions, resulting in an exponential growth in the number possible plans to evaluate. Complicating the process further is the significant uncertainty in the adversary and environment. State of the practice: Fortunately, military planning processes, like the one developed by the Marines, provide techniques for managing this complexity and synthesizing plans. Problem: However, the inclusion of UxSs, which provide new capabilities and enable new missions, further grows the complexity of the scenarios, stressing the military planning procedures, and in some cases, resulting in the human generation of plans at tempo to become infeasible. Furthermore, UxSs consume electronic reference command signals, not an operational plan (OPLAN) or operational order (OPORD), complicating the integration of UxSs into joint human-UxS forces. Objective: We seek to leverage the existing military science of planning, along with techniques from cyber-physical systems to perform hierarchical planning for war games at operational tempo in a manner that: 1) admits the use of UxSs to perform tactical mission tasks, and 2) enables effective integration into the current Marine planning procedures. Approach: The approach is to use a hierarchy of scalable system models along and mathematical planning algorithms of known computational complexity to actively trade planning time and risk by actively changing the size of the planning problems. We aim to quantify this risk so that commanders can make informed decisions within their allocated planning times. We propose four research thrusts. --Thrust 1 - System Model Abstractions Across Planning Levels: Leveraging Markovian processes and graph theoretic methods to create contract system model abstractions that are sufficiently expressive for representing the dynamics of complex war games while remaining tractable. --Thrust 2 - Quantifying the Uncertainty Induced by Abstractions: Quantify uncertainty in the abstraction due to necessary simplifications made during the contraction of the abstraction, using probabilistic conformance theory, approximate bisimulation theory from formal methods, and uniform boundedness from nonlinear systems theory. --Thrust 3 - Planning Under Uncertainty: Using stochastic and distributed control methods, compute and evaluates plans given abstractions and their quantified uncertainty, providing operators with the operational effectiveness of the plan and the uncertainty in that effectiveness. --Thrust 4 - Controlling Complexity: Employ mathematical programming and optimization techniques to principally leverage the first three thrusts to enable planning at operational tempo by effectively scaling the problem, and the level of coordination, with the goal of simultaneously maximizing performance and minimizing uncertainty in the plans under complexity and planning time constraints. Operational Impact: If successful, the proposal enables: 1) tractable computation of plans for complex scenarios at operational tempo, 2) quantified uncertainty in plans, enabling risk-aware decision making, and 3) intelligible system behavior, interfacing, and decision making for manned-unmanned teaming.