Effective cooperation among machines typically requires some form of communication or mutual understanding. One method for doing this is to explicitly build a communication protocol into the machines as is done in computer networks. This can be overly time consuming for complex systems and limits the adaptability of the machines. A second strategy is to design the machines with motivation driven by a cooperation-quality metric. This allows the machines to perform with little to no communication in a highly dynamic environment. However, this technique depends upon the machines having similar goals, experience, capabilities, and design. The focus of this research is to develop a method for adaptive machines with different goals, experiences, capabilities, and design, to communicate in a dynamic environment. This will expand the cooperative capabilities of current and future machines utilizing learning-based algorithms.
