Abstract:  In actual welding scenarios, uncertain factors such as workpiece position errors and shape changes result in inaccurate grid maps, leading to significant deviations in welding trajectories. Under the combined effect of this deviation and offline deviation, the complex coupling characteristics result in the inability to obtain error compensation for each joint, leading to a high degree of dispersion in the final welding task. Therefore, a method for automatic compensation of offline deviations in complex curve trajectories of welding robots is proposed. Map the centerline coordinates of the weld seam to a grid map, divide the grid units, assign numbers and passage rights, and consider the cost of passage to construct a weighted graph model to complete the grid map construction, providing a foundation for subsequent ideal welding curve trajectory planning. Set start and end points and initialization parameters in the grid map, use ant colony algorithm to select nodes based on pheromones, update pheromones, and filter out ideal welding trajectories. The end pose of the welding gun generates errors due to joint rotation, and a compensation value is selected to obtain a specific compensation amount for the joint. The resulting perturbation is eliminated from the original offline deviation. The pose is represented by a homogeneous transformation matrix, and the error transformation matrix is derived based on the coupling relationship of joint rotation angles. After ignoring high-order terms, the joint error compensation value is determined by decoupling calculation, and the compensation value is obtained to adjust the pose, achieving offline deviation automatic compensation. The experimental results show that the proposed method significantly reduces the deviation between the compensated welding trajectory and the expected trajectory, and tends to overlap; The trajectory tracking dispersion of the welding robot remains stable below 0.2, and the compensated actual welding trajectory coincides with the expected trajectory height. It can achieve automatic correction of welding trajectory and improve the accuracy of welding tasks.