An improved rapidly-exploring random tree（ I-RRT） algorithm with target probability offset and step size control is proposed to solve the problem that RRT path planning algorithm lacks guidance and has high time complexity when the planning space increases. I-RRT combines with target probability offset

takes a certain probability to offset the sampling point as the target point

improves the directivity of path planning

and introduces step-size control optimization algorithm to improve the operation efficiency and optimize the path. The simulation experiment of the algorithm is established on the platform of MATLAB

and the results show that the directivity and time complexity of I-RRT are better than those of classical RRT algorithm. The obstacle avoidance planning and control experiment of 6-degree of freedom（ DOF） manipulator is set up on the platform of ROS

and the validity of the algorithm is verified by the experiment.