This paper addresses the distributed and coordinated target tracking and collision avoidance using multiple nonholonomic mobile agents whose dynamics are subject to uncertainties and disturbances. A cascaded control architecture is proposed, which consists of a virtual "simulator" and a local tracking controller. The simulators exchange their uncertainty-free states over the network and achieve flocking, while avoiding collisions, using the information from their neighbors. The local tracking law, which consists of an outer-loop guidance law and an inner-loop L1 adaptive control law, solves the tracking problem with respect to the state of the virtual simulator in the presence of plant uncertainties. The guaranteed transient performance of the L1 adaptive controller is essential towards resolving the unavoidable coupling between the communication topology and the system dynamics. Extensive simulation results demonstrate the capability of the proposed algorithms to recover the desired flocking behavior. The proposed algorithms are implemented on an indoor multi-robot platform.