The main objective of this research is to develop methods and strategies for the task of cooperative search by swarm mobile robots. In this dissertation, some non-conventional methods are investigated for overcoming the restrictions from traditional methods while the Festo Robotino robot is taken as the specific research object. First, the biologically inspired Particle Swarm Optimization (PSO) algorithm which is used as an optimization tool in tradition is extended to mechanical PSO by including the physical features of the robots. The mechanical PSO is then used for guiding and advancing the search progress for the whole swarm in which the robots are assumed to cooperate under the view of a virtually linked multibody system. An independent obstacle avoidance module is activated when robots encounter any conflicts during the search, whereas the involved mathematical constraints are effectively treated by augmented Lagrangian multiplier method. Following, a detailed investigation on the concerned robot is performed including a complete derivation of the kinematics, dynamics and equation of motion. Based on these, a linear quadratic regulator and a gain-scheduled H infinity controller are designed which correspond to the fact of taking the robot as a linear system and a linear parameter-varying system (a nonlinear description), respectively. Since the update of mechanical PSO requires the pose information of each member in the swarm, the robot localization and position control are researched consequently. The localization is realized by fusing the measurements from two relatively independent sensors, the corrected internal odometer and the external North Star system, within the extended Kalman filters. Afterwards, the filtered information is feed back to the computed poses yielded by mechanical PSO for further accelerating and precising the position control. The whole work is finally verified by plenty of tests by both simulation and experiments which has shown the feasibility, the fault tolerance, the swarmability and flexibility, the correct interactions between the simulated robot and real robot, and the effects from mathematical constraints. That is to say, the cooperative search by a robot swarm is achieved encouragingly. The mixed simulated robots and real robots cooperative search in a swarm has shown the novelty of this investigation. The synthesized methodology thus has generated a versatile robot swarm with comprehensive capabilities.