Robot Camera Pose Compensation Inside Machine Tools

The patent transitions from using a visual target on the outer surface (2D plane) to using an identification figure on an internal structure component inside the machining area (3D space within the machine tool). This dimensional change allows the camera to capture the identification figure from within the machining area, enabling accurate pose compensation when the robot hand is inserted into the machine tool, rather than only when outside.

The patent introduces an identification figure attached to an internal structure component (such as the chuck or spindle) as an intermediary reference object. This intermediary allows the camera to measure the robot's pose relative to a fixed point inside the machining area, serving as a mediator between the robot and the machining environment, enabling accurate compensation during insertion operations.

The patent implements a feedback mechanism where the camera continuously captures images of the identification figure on the internal structure component, and the controller calculates the robot's actual pose based on these images. This real-time feedback allows the system to compensate for tilt and overhang effects dynamically, maintaining measurement accuracy regardless of the robot arm's position or the AGV's stability.

The patent replaces the mechanical assumption of a fixed, level robot-mounted surface with an optical measurement system. Instead of relying on mechanical stability, the system uses the camera to detect the identification figure and computationally determine the robot's actual pose, substituting mechanical rigidity requirements with optical feedback and computational compensation.

The patent implements a self-service positioning system where the robot-mounted camera captures images of the identification figure on the machine tool's internal structure, and the robot's controller autonomously calculates the positioning error and compensates its own pose. The system uses itself (the robot's own camera and controller) to perform the positioning correction, eliminating the need for external high-precision positioning infrastructure.

The patent changes the reference parameter from the AGV's wheel-based position (which has inherent errors) to the relative position between the camera and the identification figure on the internal structure. By using image processing to determine coordinates and calculating transformations between coordinate systems, the system transforms the positioning problem from a mechanical wheel-based reference to an optical reference that can be accurately measured regardless of AGV positioning errors.