Aiming at the problem that modular mechanical arm may collide with the obstacles in working space at runtime, a path planning algorithm of obstacle avoidance is proposed based on a genetic algorithm (GA). Firstly, the Denavit–Hartenberg method is applied to the modelling of the mechanical arm; then, the kinematic and kinetic analysis is conducted and the kinematic and kinetic equations of the mechanical arm are established. On this basis, the time, spatial distance and path length of the motion are regarded as optimising indices and the optimisation of path planning of obstacle avoidance is achieved using GA for mechanical arm under a working condition with single obstacle or multiple obstacles. Through simulation, the effectiveness and feasibility of a path planning algorithm of mechanical arm obstacle avoidance based on GA is verified. This algorithm improves the efficiency of a mechanical arm to avoid the obstacles in working space at runtime.