Aiming at the shortcomings of the skin force sensor to detect the collision point of the robot, a method using the six-axis force/torque sensor to detect the collision point of the robot body was proposed. The sensorcollected data was used for selfconstraint, without relying on the geometric information of the collision body surface. The space collision external force vector line was firstly projected into the optimal plane for preliminary solution, and finally brought into the original equation to solve the collision point. In order to ensure that the absolute error of the calculation result was the smallest, an error factor was introduced. In terms of data preprocessing, a dynamic force compensation algorithm was proposed to ensure that the six-axis force/torque sensor at the base had a constant reading of zero when there was no external force collision during the robot movement. The robot can be considered collide with the outside world when the sensor value exceeded a certain threshold. Finally, a simulation experiment was performed on the proposed algorithm. The experimental results showed that the maximum error of the dynamic force compensation algorithm was 4.8925%, and the collision point detection algorithm had the largest error at the experimental distance of 598.61mm, which was 8.7119%. The experimental results showed that the accuracy of the proposed dynamic force compensation algorithm was not changed significantly as the distance of the collision point was increased, but the relative error of the collision point detection algorithm was increased with the increase of the collision distance when the collision force was constant.