The artificial search swarm algorithm (ASSA) is a kind of bionic intelligent algorithm, which is inspired by the human communicating with each other in search. The fixed collaborative parameter and global weight in the standard artificial search swarm algorithm easily lead to local convergence and influence the convergence speed. Therefore, this paper proposes an improved artificial search swarm algorithm (IASSA) by introducing global weight, linearly dependent synergistic coefficient and dynamic step. The global weight improves the historical experience of the search group in reconnaissance behavior, so as to strengthen the guiding role of global optimal individual for all population. The dynamic collaborative parameters are used to improve the search probability of the soldier's cooperative behavior, so as to enhance the local information exchange between the populations. The dynamic step strategy is used to improve the convergence speed and convergence accuracy. The performance of the algorithm is verified on 15 benchmark functions. The results show that the improved algorithm can effectively avoid the local convergence and significantly improve the solution accuracy and convergence speed compared with the standard ASSA algorithm and the other three kinds of optimization algorithms.