The invention discloses a pulse neural network based method for controlling object tracking of a robot. The method is characterized in that after being preprocessed, multi-sensor information sensed by a robot is subjected to integration processing in a pulse neural network in a pulse string manner, and further, the robot is controlled to perform collisionless object tracking; through integrating visual, coded disc and ultrasonic data, the robot generates object information and obstacle information which are preprocessed and then sent to an input layer of the pulse neural network to generate corresponding pulse signals, wherein the input layer comprises object neurons, sensor neurons and obstacle avoidance neurons, the pulse signals generated by the obstacle avoidance neurons are processed by neurons of a hidden layer; the generated pulse signals are sent into two pairs of motor neurons of the input layer after being weighted, and pulse strings output by the two pairs of neurons control left and right motors after being subjected to competing, filtering and amplitude limiting. According to the invention, the robot make real time decisions according to information sensed by the robot so as to provide the technical support for the application of the robot in security, safeguard and the like.