MPC—Multi Platform Controller
ATCorp has developed an agent-based framework for adhoc wireless networks in support of the Multi-Platform Controller (MPC) program in ARDEC, Picatinney Arsenal, New Jersey. As an Army Phase II SBIR, its goal was to provide a communications and operational framework to facilitate interaction with remote mobile platforms from a central control station. The MPC program here at ATCorp yielded the communications framework that included 1) guaranteed messaging with all connected platforms, 2) agents that maintained network connectivity, and 3) tele-operation of remote platforms.
This research effort developed a layered infrastructure of wired/wireless networking services, proxy and distributed processing, agent-based behaviors, remote tele-operative services, and mixed-initiative planning and scheduling technologies that would support the planning, control, coordination, and reconfiguration across multiple robotic platforms in a distributed set of Future Combat System (FCS) platforms. The design maximized commonality, reuse, and adaptability across platform type and configurations. The design includes approaches for controlling platforms within tele-operational, semi-autonomous/ supervisory, and fully autonomous modes. This provides the remote platform operator with maximum control and flexibility over multiple platforms and vehicles.
The development effort concentrated on reusable components for Army/DOD applications across multiple robotic platforms. The MPC system concept approach starts with the MPC system supporting re-supply, logistics, and other various material-handling missions. These missions are input to the MPC planning component and, with optional user interaction, are decomposed into a set of partially ordered tasks. These tasks are sent to the MPC scheduling component and merged into a schedule of multiple missions, also with optional user interaction.
During execution, the progress of planned tasks must be monitored for possible deviation from the tasks' goal. Also, any coordination between the tasks must be supported by the MPC environment. If any deviation is detected, the tasks' progression/course must be dynamically redirected back on track, and if that fails, then the task needs to be halted and a new plan/task needs to be planned and scheduled.
The process employed by the MPC system is to assign intelligent software agents (called Task Agents) to each task on the schedule. The MPC agents are launched or instantiated onto a processor within the network. Each agent has the purpose of controlling the execution of its assigned task, and is responsible for a successful task completion. Other agents are also instantiated and launched to monitor the Task Agents, and to aid in the coordination between tasks as guided by the schedule.
The execution component consists of the active MPC agents accomplishing their assigned tasks. Some agents are assigned a specific task from the schedule (Task Agents). Some are supporting a task by collecting and supplying input data to a Task Agent. Others are created to control the coordination between tasks (Coordinated-Behavior Agents), which is also as specified in the schedule.
Finally, additional agents are created to monitor the schedule execution for: valid execution, successful completion, and dynamic recovery from deviations to the expected results. As execution progresses, there may be points in a task where the robotic platform is required to be tele-operated by the human MPC operators. This could be due to defined points of tele- operation in the task that were designated by the human operator during the planning or scheduling phases. They also could be due to the MPC agents deciding that the platform's execution of the task has deviated too far from expected operation, and thus the agents will summon the human operator to intervene and take control. In either case, there is a tele-operation handoff from the agent, and then when the human operator has accomplished her part, there is a release of the tele-operation control back to the controlling task agent. Such handoff/release mechanisms provide a smooth transition between controlling entities, and are critical to effective and errorless control of the platforms.