Abstract: Abstract: By the visual feedback and the space position information of the target object of stereo vision camera, the construction robot can realize the autonomous task according to the kinematics analysis and trajectory planning. However, in the process of the autonomous task, because the link mass of construction robot is big, the driving force of the cylinder calculated by the pressure sensors attached at the cylinders will be divided into two parts, one is used for balancing the link gravity, and the other is used for driving the moving of the cylinder. Therefore, the construction robot will overcome the gravity to work in the process of gravity rising along with the lift of boom and swing of arm, and the links gravity of boom and arm will participate in working in the process of gravity fall along with the dropping of boom and swing of arm, this phenomena will influence the moving velocity of construction robot and the effect of the effect of trajectory tracking, especially in the lifting process, moreover, in the process of the links dropping of construction robot, it is dangerous to the construction robot because of the bigger links masses. Aiming at this problem, the dynamics models of construction robot were deduced followed by kinematics analysis, and the least squares method was used for identifying the dynamics parameters, and then online gravity offset method was purposed based on the dynamics parameters, which was used for eliminating the gravity impact from the driving force of the cylinders, and improving the trajectory tracking effect in autonomous task. Finally, experiment was finished on the construction robot test bench, and the experimental results show that the online gravity compensation algorithm could compensate the gravity of construction robot effectively, and eliminate the influence of gravity working to the trajectory planning of construction robot, and the tracking errors under online gravity compensation are smaller than no gravity compensation, therefore, the online gravity compensation will be propitious to reduce the trajectory tracking errors, and improve the trajectory tracking performance of autonomous task, at the same time, the gravity compensation algorithm has important significance to reduce the energy input and consume for the autonomous task of construction robot.