Abstract: The research,development and application of the driverless system of the underground locomotive in coal mine is useful to reduce the transportation accidents’ probability caused by scheduling and operation errors and reduce the number of miners used. Different from the mature technology of self-driving cars and trains,there are many new challenges in implementing driverless locomotive in underground environment,such as the sharing of people and vehi- cles in the transportation roadway,narrow roadway,uneven lighting conditions,inability to use the satellite positioning system,the lack of effective means of communication and so on. Starting from the analysis of the key problems in the realization of the underground driverless locomotive system,this paper summarized some technical research progresses. Firstly,the system architecture of underground driverless locomotive based on the seamless combination of information network and control network is proposed to ensure the intellectualization of train dispatching,the automation of locomo- tive control and state collection,the integration of transportation monitoring center and dispatching center,and the com- patibility of remote control,autonomous operation and other driverless modes. Secondly,the concept of intelligent dis- patching is brought about,that is to say,the driverless system should be realized on the basis of intelligent transporta- tion scheduling,which can effectively promote the resource sharing and function coordination between the two systems. Thirdly,the positioning technology of closed underground environment is discussed,and a clear conclusion is drawn that the UWB ( ultra-wide band) positioning can effectively cope with the high-precision wireless positioning of sub meter level locomotives with excellent performance in robustness and stability,which can meet the positioning accuracy of underground locomotive unmanned driving. Fourthly,the data communication network coverage suitable for under- ground is described,the performance indicators required to access WLAN network according to actual engineering ex- perience are listed,and the breakthrough of underground driverless application that the 5G new communication net- work will bring about is also forecasted. Fifthly,with the aspect of using machine vision for road condition analysis,the track detection algorithm based on track model and image feature,the target detection algorithm based on deep learn- ing neural network,the target distance estimation algorithm based on binocular measurement and monocular measure- ment,and the lightweight technology of deep learning network are also discussed respectively. Finally,the development and application prospect of this field are prospected.