Abstract Abstract: Because there are a lot of emotional robot task allocation algorithm, which can not better exert the decision function of emotion in task allocation, moreover, the task allocation performance of large-scale emotional robot teams is poor. The self-organizing algorithm based on SOM self-organizing map neural network is proposed in this paper and task allocation is integrated into the training process of network. Taking advantage of the decisive effect of the competition winning function value in network training (task allocation) and combing with the real-time of the task allocation, the emotional robot provides the competitive victory function value, which adjusted by reinforcement-learning, according to the emotional factors. In this way, the emotion directly and effectively participate in task allocation decisions and the performance of the algorithm is optimized. Finally, simulation experiments are carried out to verify the effectiveness of the proposed algorithm.Especially, with the scale of emotional robot team growing, pursuit time can shorten more than double than the existing algorithm.