Topological reconfiguration can improve the static stiffness and fault tolerance of parallel robots, but the problem of whether the robot can reach any specified pose without singular configuration has not been solved. Therefore, this paper takes the reconfigurable 3-RRR planar parallel robot as the research object, and proposes a path planning method that can realize ' point-to-point ' tracking. The path planning algorithm is suitable for reconfigurable parallel robots with arbitrary degrees of freedom. It has intuitive geometric interpretation and can eliminate all singular configurations. Firstly, the inverse kinematics model before reconfiguration is established, and six paths based on topological reconfiguration are found. After research, there are many ways to determine each path. Then, the formulas of the relevant parameters input in the path planning process are derived, and the correctness of the derivation results is verified by the prototype experiment. Further, based on the screw theory, the input and output transfer indexes under each input mode are calculated, and the optimal solution of the input mode is obtained by using the local transfer index. Finally, the ' point-to-point ' path planning algorithm is extended to continuous path planning.