Multi-sensor optimal allocation is significant for mobile robot navigation, and its perception robustness is an important performance standard for multi-sensor system. Thus this paper designs sensors optimal allocation model and defines a robustness evaluation fitness function. Meanwhile, an improved DPSO algorithm is proposed to solve the sensors coalition allocation with the best robustness. Firstly, a multi-sensor optimal allocation model is devised with single sensor level constraint and sensor coalition level constraint. Secondly, an improved DPSO with new updating mechanism is developed for sensors coalition with the best robustness. Furthermore, the particles updating mechanism is redefined, which composed of 0 and 1 binaries of DPSO position element. Finally, heterogeneous sensors allocation experiments are analyzed in order to show the feasibility and effectiveness of improved DPSO. Those experiments show the sensor coalition work smoothly and efficiently, even though several sensors are damaged. Additionally, the advantages of our model and algorithm is proved by comparison experiments.