Robots have been widely used in daily life. Path planning is one of the main technologies of robots. An outstanding path planning algorithm can improve the work efficiency of robots, reduce cost, and lay a good foundation for the research of robot navigation. The RRT (rapidly-exploring random trees) algorithm, which is well known for its strong scalability, has some shortcomings, such as suboptimal path length and poor smoothness. An improved algorithm of reverse optimization and cubic spline interpolation is proposed, and then simulated in different scenarios in MATLAB. Taking the restaurant environment as an example in ROS (robot operating system), the experimental results show that the improved RRT algorithm can decrease the path length and the number of nodes, and improve the smoothness, verifying the effectiveness of the algorithm.