To solve the problem that the velocity of Reciprocal Velocity Obstacle (RVO) avoidance is optional in theory but is inaccessible in practical application, after comprehensively considering the advantages and disadvantages of RVO algorithm and Dynamic Window Approach (DWA), this paper fuses the two algorithms, so as to add kinematic constraints to the optional velocity of obstacle avoidance. Furthermore, when multiple mobile robots are encountering in complex environment, the conventional expansion radius may lead to failure in obstacle avoidance. To solve the problem, an Adaptive Reciprocal Velocity Obstacle (ARVO) algorithm with variable expansion radius is proposed. ARVO algorithm can adjust the expansion radius according to the complexity of the environment, which not only ensures a safe distance between the mobile robot and the surrounding obstacles, but also increases the range of optional speed. Finally, the performance of the algorithm is verified by Gazebo simulation platform in Robot Operating System (ROS).