This paper proposes a firm tying method by a dual-arm robot that actively exploits the corners of a target object as geometric constraints. The method employs a four-step sequence of linear arm movements and gripper actions to place the knot at a corner, where the rope digs into the edge and generates large normal force and rotational resistance that prevent loosening. Tensile experiments with a six-axis force/torque sensor confirmed that corner-tying significantly outperforms center-tying in both tensile force and torque, with geometric locking being the dominant factor in binding strength. Material comparison revealed that rope flexibility is more critical than surface friction. A linear-interpolation based trajectory correction using the object’s external dimensions also achieved high success rates across differently sized objects without detailed physical modeling.