Particle swarm optimization using average information of swarm is presented based on Kennedy and Eberhart’s binary particle swarm optimization. The probability of the position vector depends on the values of the average individual best position and the individual best position . The new algorithm can makes each particle use the whole swarm information effectively. The test of function optimization and 0-1 knapsack problem show that this new algorithm has a good speed of convergence, a high stability and performs better than BPSO and some modified algorithms.