Spinal cord injury and stroke are the main diseases that lead to loss or decline of hand function, and hand rehabilitation exoskeletons can be used to help restore hand function. Due to the problems of traditional hand exoskeletons such as heavier weight, higher cost and limited mobility, this project aims to improve an index finger exoskeleton robot with an underactuated mechanism, improve its bending and extension angles, and optimize the robot to a certain extent. flexibility. At the same time, the genetic algorithm is combined in the control system to make all the control indexes reach excellent. This project establishes a 3D model through Solidworks, and at the same time proves the feasibility of improvement by establishing a kinematic model. In terms of design, the mechanism is adjusted according to the range of activities of MCP(metacarpophalangian joint) and PIP(proximal interphalangeal joint) to optimize the flexibility of the design model. In the aspect of control system, the genetic algorithm is combined with system identification and PID(Proportion Integration Differentiation)control, in order to reach the objective of establishing a more precise training model and finding appropriate PID parameters. This project has a positive effect on the medical efficiency of the underactuated hand exoskeleton.