Abstract: Although fire-fighting robots are important devices for underground coal mines to prevent and deal with fire incidents, current research has hardly studied the approaches of robot locating, path planning, and accurate fire detection and extinguishment for complex underground coal mine topography. To fill the gap, this paper proposes a system design for underground coal mine fire-fighting robots. By combining the technologies of ultra wide band (UWB) and laser radars, the design applies the iterative closest point (ICP) and the adaptive Monte Carlo localization (AMCL) algorithms to initialize robots' position and locate robots in a real time manner, respectively. By using inertial measurement units and odometers, the design can raise the real localization accuracy up to 5-10 centimeters. By adding a distance parameter to limit the searching area, an enhanced A^* algorithm is proposed to plan paths for robots. Experiments show that the algorithm can find the suitable paths in lower time cost. The design supports a robot to detect targets using a template mapping system based on a pre-generated feature image set. It limits the false positive rate to 10% (true positive rate of 90%) and completely meets the design requirement. The design allows a robot to be equipped with a cascade controller that uses feedbacks of velocity and position to control the elevation and yaw angels of the pan-tilt-zoom (PTZ). Fire extinguishing bombs can be further thrown to targets, via the PTZ, to support accurate fire extinguishment.