Coordinated Exploration With A Shared Goal In Costly Environments

The paper studies distributed cooperative multi-agent exploration methods in settings where the overall benefit of an opportunity is the minimum of individual findings and the exploration is costly. The primary motivation for the model is the multi-channel cooperative sensing problem which draws from the inter-vehicular cognitive offload paradigm. Here, vehicles try to coordinate an offload channel through a dedicated common control channel, and the resulting quality of the channel eventually selected is constrained by the individual qualities. Similar settings may arise in other multi-agent settings where the exploration needs to be coordinated. The goal in such problems concerns the optimization of the process as a whole, considering the tradeoff between the quality of the solution obtained for the shared goal and the cost associated with the exploration and coordination process. The methods considered in this paper make use of parallel and sequential exploration. The first approach is more latency-efficient, and the latter is shown to be more cost-effective. The strategy structure in both schemes is threshold-based, and the thresholds which are analytically derived in this paper can be calculated offline, resulting in a very low online computational load. A comparative illustration of the methods' performance is given using a synthetic environment, emphasizing the cost-latency tradeoff.