The use of autonomous robots in complex exploration tasks is rapidly increasing. Indeed, robots can provide speed and cost effectiveness in many tasks, as well as allow operation in environments that are hostile to humans. In this dissertation we: (1) provide two adaptive navigation algorithms; (2) develop a coordination mechanism; (3) develop a dynamic partnership formation mechanism; and (4) demonstrate the use of algorithms in a hardware implementation.;The two adaptive navigation algorithms are neuro-evolution and policy gradient, where the results show that effective, adaptive navigation techniques can be developed for mobile robots in an exploration domain when the robots have limited capabilities. In addition, we show that policy gradient approaches thrive on short-term objective values, whereas neuro-evolutionary approaches provide more robust results with a time-extended objective value. Finally, we show that summing short-term values to generate a time-extended value does not capture the complexities of some real world exploration tasks.;Coordinating multi-robot systems to maximize global information collection in these exploration domains presents additional challenges. In particular, in many multi-robot domains where communication is expensive, the coordination must be achieved in a passive manner. This is done in this dissertation via objective design on a hierarchical control scheme where both a navigation algorithm and coordination algorithm are operating simultaneously.;We then extend results on such multi-robot coordination algorithms to domains where the robots cannot achieve the required tasks without forming teams. We investigate team formation where: (i) robots must perform a task together; (ii) there is an optimal number of robots; and (iii) individuals vary, forming heterogeneous teams. The results show that using neuro-evolutionary robot teams with objective functions that are aligned with the global objective and locally computable significantly improve over robots using the global objective directly, particularly in dynamic environments.;Finally, we develop a path to implementation of all of the coordination research done to date into robot hardware. The design represents a stable, robust robotic platform on which navigation and coordination algorithms can be run in the fashion they were developed and intricacies of real-world operation can be analyzed. Functional experiments show that the platform operates as expected and performs similarly to algorithm work done in simulation.
展开▼
