The inefficiency, exponential amount of communication and computational time are high undesirable for realistic applications under utilizing the existing distributed task allocation approaches, especially for the dynamical tasks that move randomly before assigned robots to execute them, and the condition of these tasks could vary over time. For such dynamical task assignment problem, we propose a dynamical task allocation and reallocation method for multiple robots coordination system based on multi-round negotiation and body expansion behaviour. For the first time round negotiation, robots sequentially negotiate and select tasks to perform according the proposed algorithm, and declare the information to other robots. When all robots have finished first time selecting, then the remaining un-selection robots choose the rest un-assigned tasks again sequentially. We set two distance thresholds for decision making so as to implement body expansion behavior. Based on the body expansion behavior, one robot can request, accept and refuse other robots to execute tasks by intention communication under the order of two distance thresholds. The advantages of dynamical task allocation and reallocation approach is demonstrated by comparing with existing task allocation algorithm in this paper. The simulation results show that the efficiency for whole multi-robot coordination system to accomplish all tasks is improved by utilizing our approach. Moreover, it is more conducive to reduce the numerous computational time and communication compare to the existing investigated task assignment methods.