Systems and methods for distributed hierarchical control in multi-agent adversarial environments

A system comprising:

• a first platform including a first platform level agent configured to direct one or more actions of the first platform based on at least one of a selected target or a selected goal, wherein the first platform includes an aircraft;
• and
• a computer system comprising at least one processor in communication with at least one memory device, wherein the computer system is in communication with the first platform level agent, wherein the memory device includes computer-executable instructions that when executed by the at least one processor cause the at least one processor to:
• train the first platform level agent and a supervisor level agent on a plurality of training scenarios in an adversarial environment including a plurality of allied agents and a plurality of opponent agents, wherein the plurality of allied agents includes the first platform level agent, wherein the training comprises:
• training the first platform level agent using a dense reward quantifying a goal of the first platform level agent;
• and
• training the supervisor level agent using a global reward comprising a sparse reward and a weighted sum of a plurality of dense rewards, wherein the sparse reward describes a result of a training scenario of the plurality of training scenarios, and wherein the plurality of dense rewards includes the dense reward of the first platform level agent;
• execute the trained supervisor level agent configured to select at least one of a target or a goal for one or more platforms including the first platform;
• receive targeting information including one or more targets;
• receive platform information for the one or more platforms;
• select, by the trained supervisor level agent, a target of the one or more targets based on the target information and the platform information;
• and
• transmit, to the trained first platform level agent, the selected target, wherein the trained first platform level agent controls the first platform based on the selected target.