We illustrate the design and development of amodular hardware/software system for multiple unicycle-likemobile robots localized via a set of camera modules. We describethe architecture and calibration of the hardware/software setupand then discuss two continuous-discrete observation laws forthe distributed estimation of the robot positions. We show thata suitable model exploiting the onboard IMU measurementsof the robots, enables obtaining an estimation error that is acascade of two linear systems, for which we can show globalexponential convergence to zero. The results are illustrated byour experimental tests.