For solving the problem of tactical communication support, a task allocation model is established, and a twice search particle swarm optimization(TSPSO) algorithm based on gradient descent is proposed. There are four parts added into the TSPSO algorithm, including determination of the extremum trap, particle twices search, set forbidden area, particle elimination and generated. The TSPSO algorithm and other four improved algorithms are applied to the optimization problem of four typical test functions. The results show that, the convergence accuracy of the TSPSO algorithm is higher, the convergence speed is faster. In the solution of the task allocation model, the probability distribution of the support units to communication places is used to encode and decode the particle swarm. The simulation results show that the proposed algorithm can quickly find the optimal allocation scheme of tactical communication support tasks.