Aiming at the path planning problem of military mobile robots，an improved algorithm based on RRT-Connect is proposed，which avoids the problems of low efficiency，large randomness，long search time，excessive number of iterations and nodes，and lengthy planned path in the algorithm planning process. Firstly，the algorithm takes the midpoint of the line between the starting point and the target point as the extension point，so that the algorithm changes from two-tree expansion to four-tree simultaneous expansion，and a range limiting function is introduced at the same time to optimize the path expansion and sampling process. Secondly，the dynamic step-size adjustment function is introduced into the improved algorithm to make the expansion process more target-oriented. Finally，the generated path is post-optimized to remove redundant nodes and shorten the path length. Simulation experiments are conducted for a comparison between the improved RRT-Connect algorithm，RRT algorithm and RRT-Connect algorithm in three different environments. Compared with the RRT-Connect algorithm，the improved algorithm has the number of overage iterations reduced by 83.8%，the number of average nodes reduced 59.3%，the average planning speed increased by 26.9%，and the average path length reduced by 12%.