Multi-Agent Coordinated Strategic Team Behavior

Author(s): Michael Franklin

Many researchers and industrial partners have worked over the years to produce multi-agent artificial intelligence and teams of agents working together. Most of the previous work to date has fallen short of this goal, resulting in multiple agent AI, where each agent has their
own policy but there is no larger group policy. This results in each agent doing the best they can to achieve the goal but falling woefully short because they have no awareness of the other agents around them. In multi-robot scenarios, this means that the robots move inefficiently and often collide with one another or must wait for clear space to operate. This is far from the desired goal of synergistic team behavior.
Our latest research has seen the emergence of multi-agent AI, where each agent has an individualized policy, but they are also compliant to a larger group policy. This means that our newer systems are functioning as a team rather than as a set of individuals with similar goals