This dissertation presents a solution to the problem of determining and maintaining the visibility of a moving target by using one or more robots. Within the framework of surveillance systems and mobile sensors, our work contributes to the relatively less addressed research on visibility in the case of a moving target and moving cameras. The design goal is to attain views of a moving target from multiple angles while an unpredictable target performs a variety of motions. One or more robots follow the target and approach the target at a pre-determined distance and angle to obtain images. In this way, more informative and task specific results can be obtained from the sensors, rather than a general view that may not be always useful. Although our main motivation is to obtain visual information about a target by positioning mobile cameras at task specific locations, the method can also be used with other sensors to accomplish a wide variety of tasks such as docking, transportation, automated repairs, etc.

We present three algorithms for the solution of multi-angle view problem. The first one addresses the general solution in the case of a single-target single-pursuer (STSP) setting and with a fixed order of approach angles. We demonstrate the method by simulations and on a robot. The second algorithm improves the first one by enabling the pursuer to adaptively decide which approach angle to aim rather than following a fixed order. Performance in each case is described by the task completion time. The method is generalized for the case of multiple pursuers and a pareto optimal solution for task allocation among the robots is presented and evaluated by simulations.

An unpredictable target motion presents many challenges on the design of the pursuer motion strategies and the initial placement of the pursuers. For the STSP case, we present a mathematical analysis on the pursuer step size such that the target can always escape. For multiple pursuers, an analysis is presented on the initial placement of the pursuers and their step sizes to complete the task in a pareto optimal manner.