Abstract: An improved RRT(rapidly-exploring random tree) algorithm is presented from the definition of nonholonomic constraints to solve the motion planning problem for mobile robots with dynamics constraints.The algorithm combines the nonholonomic constraints of mobile robot with RRTs.For RRTs,in order to reduce unnecessary costs caused by uniformly random searching in the global state space,the idea of goal bias is introduced and the distance parameter with low computational complexity is also selected to increase the calculation speed.Simulation experiments with several typical robot motion planning with nonholonomic constraints verify the validity and high efficiency of the algorithm.