Using Coordination Costs to Adapt Communications

The coordination cost measures facilitates identifying which communication method is most suitable given the environment. We model every robot’s coordination cost Ci, as a factor that impacts the group productivity. We analyzed two cost categories: firstly, cost relating to communication and secondly, reactive and/or proactive collision resolution behaviors. It focuses in energy and time spent communicating and in the consequent resolutions behaviors. We then combine these factors to create the function of multi-attribute cost based on the Simple Additive Weighting (SAW) method often used for multi attribute utility functions.

Other communication issues, such as bandwidth limitations, can similarly categorize as additional cost factors as they impact any specific robot. For instance, if a robot needed to retransmit a message due to limited bandwidth, cost in terms of additional time latency and energy used in retransmission are likely to result.

There are two types of adaptive methods: firstly, uniform communication adaptation; secondly, adaptive neighborhoods of communication.

Uniform switching between methods
The first method, all robots simultaneously switch between mutually exclusive communication methods as needed. In order to facilitate this adaptation form, each robot autonomously maintains a cost estimate, V used to decide which communication method to use. Because a robot detects no resources conflict, it decreases an estimate of this cost, V, by an amount Wdown. When a robot senses a conflict is occurring the value of V is increased by an amount Wup. The values for V are then mapped to a set communication scheme methods ranging from those with little cost overhead such as with no communication, to more robust methods with higher overheads such as the centralized and localized methods.

Adaptive neighborhoods of communication
The advantage in the first adaptive approach lies in its simplicity. The uniform adaptive approach switches between existing coordination methods based on cost estimation. Assuming one analysis a new domain with different communication methods completely, and can order the communication methods based on their communication costs, this approach will be equally valid as it implements existing methods and reaches the highest level productivity.