Abstract: To address insufficient joint drive torque and the tendency of motors to overload in hexapod mine rescue robots under heavy loads, a rope-pulling lead-screw dynamic torque compensation mechanism was proposed. The mechanism employed an adaptive torque compensation control strategy and a hybrid motor–extension-spring drive mode. First, a single-leg static model was established to describe the mapping between joint torque and foot-end force. A load estimator was then introduced to estimate the external load in real time using joint encoder data, body attitude, and motor torque feedback and to generate the desired compensation torque. Finally, the desired value was fed into a bi-level controller, which drove the lead-screw motor to track the torque. The extension-spring preload was adjusted in real time through a half-lead-screw and gear transmission system so that the compensation torque precisely offset the static joint torque caused by the robot's own mass and external load, reducing the burden on the joint motors and improving the robot's load-carrying capacity. Simulation results showed that the mechanism reduced peak hip- and knee-joint torques by 23.8% and 39.5%, respectively, under a 20 kg load. The reductions remained within 17.1%-41.8% during crawling and obstacle-crossing gaits, demonstrating strong dynamic adaptability. The maximum payload increased from 31.8 kg to 52.4 kg, a gain of 64.8%, indicating that the mechanism can substantially enhance the robot's heavy-load operating capability.