| A wide variety of tasks, from household chores to manufacturing and space exploration, have been accomplished successfully using single mobile robot systems. Many of these tasks could be executed faster, cheaper and more reliably utilizing groups of robots working in a cooperative fashion. Formation control algorithms are a key component to enable cooperative capabilities in multi-robot autonomous systems.;This work presents a novel formation control framework called Cluster Space Control. This framework allows the user to define, control and monitor formation parameters relevant to the specific task to be executed. This technique enhances usability by offering a level of control abstraction.;Within the proposed cluster space framework, two alternative control approaches are introduced. The first is an inverse Jacobian kinematic control where compensation commands have the form of cluster velocities to be executed by the formation. The second control implementation is a Jacobian transpose dynamic control, where compensation commands are forces and torques to be applied to the formation. A novel cluster space dynamic model is introduced where formation dynamics can be derived from the dynamics of the individual robots in the group.;Collision avoidance functionality is added to the architecture in two different ways: robot-level obstacle avoidance that allows robots to break ranks to negotiate obstacles individually and a cluster-level avoidance feature that allows the group to navigate around obstacles while keeping the desired formation. Depending on the nature and the requirements of the task, one method may be more appropriate than the other.;Singular cluster configurations, analogous to those found in robotic serial manipulators, are addressed and quantitatively assessed analytically and experimentally. Stability of the closed-loop system implementing the cluster space control is studied making use of Lyapunov theory.;To address the generality of the framework, some of the most common formation control techniques found in the literature are proposed as particular cases of cluster space control.;Different experimental testbeds composed of non-holonomic rovers and marine robots are used to implement different missions in order to validate the approach. These applications are used to illustrate the advantages of the framework.;The cluster space framework for the control of robot formations is a general approach that is formally defined, analyzed and experimentally validated on different robotic platforms, providing measurable benefits for the operator and allowing simplified specification and supervision of real-world scientific and engineering tasks. |
