In order to increase the feasibility of using microrobots to perform microscale tasks and widen the range of potential applications, the use of autonomous control algorithms such as obstacle avoidance is essential. In this study, aiming at the requirement of microrobots for automatic obstacle avoidance in the simulated blood vessel environment, an automatic obstacle avoidance algorithm based on the combination of improved Rapidly-exploring Random Trees (RRT) algorithm and improved artificial potential field (APF) algorithm is proposed. The improved RRT algorithm is used to plan a global path first, and the redundant nodes on the global path are selected by using conditional constraints and key points, which is prepared to optimize the security and length of the path. Then the global path is segmented according to the key nodes, and each path is optimized with the improved APF algorithm to enhance the real time performance. Comparative simulations and experiments show that the fusion algorithm realizes the optimization of path length, safety, and local minimum problem, and can automatically avoid static and dynamic obstacles in the simulated vascular environment.