The task of visual surveillance involves pervasively observing multiple targets as they move through a field of sensor nodes. Mutational analysis and shape based control have been proposed to overcome the limitations of current feature (point) based visual servoing and tracking techniques generally employed to provide an optimal solution for the surveillance task. Hausdorff tracking paradigm for visual tracking of multiple targets using a single sensor has been proposed for accomplishing the surveillance task. However, Hausdorff tracking incorporates some redundancy in the actuation mechanism. This paper exploits this redundancy in the camera motion in order to accomplish various sub-tasks which can be assigned to the system, such as minimization of consumed energy maintaining manipulability etc. The complete task can then be expressed in a multi-objective constrained optimization framework and can be solved, i.e., the input to the camera can be derived, using various methods such as physical programming, nonlinear programming, weighted sum method, etc. In this paper, we use the physical programming method based on the various advantages such as ease of expressing multiple objectives in a physically significant manner. Experimental results are provided which show the advantages of using the physical programming approach over the weighted sum method for constructing the task criterion for multi-objective optimization problems.
