Solving a 4×4 Maze with MyCobot Pro 600

| Links: Final Report (PDF) | Demo Video

Solving a 4×4 Maze with MyCobot Pro 600

Digital twin–backed maze solving using MATLAB’s watershed transform, pixel→world conversion, IK path planning, and TCP control of a MyCobot Pro 600.

A complete pipeline to detect and solve a printed 4×4 maze with an AI kit camera, compute a collision-free path via MATLAB’s watershed transform, convert pixel coordinates to world millimeters, and execute the trajectory on a MyCobot Pro 600—while mirroring motion in a digital twin.

Stack: MATLAB (image processing, watershed, IK) · Python (TCP/IP robot control) · AI kit camera · MyCobot Pro 600 (6-DoF).

What’s in this project?

  • Vision-based solving using watershed transform to segment the maze and trace the solution path.
  • Pixel→world conversion to map image waypoints into robot workspace coordinates (mm).
  • Path planning via IK: compute joint angles for the end-effector to follow the solved path smoothly.
  • Robot execution over TCP/IP with a Python client streaming waypoints/angles to the arm.
  • Digital twin synchronization to validate and visualize motion in MATLAB while the real robot runs.

Demo Video

Details

1) Maze Capture & Pre-processing

  • Acquire a top-down image of the printed maze using the AI kit camera.
  • Threshold → binarize, isolate the largest contour (maze body), and detect wall openings to locate start/goal.

2) Watershed-Based Solving

  • Build a heightmap from the processed image and apply MATLAB’s watershed transform.
  • Extract the watershed ridge as the centerline path between start and goal.
  • Sample the ridge into ordered waypoints (pixels).

3) Pixel→World Conversion

  • Calibrate image scale (px/mm) from known references.
  • Transform each waypoint from image pixels to world millimeters in the robot’s plane.

4) IK & Trajectory Generation

  • Use MATLAB IK to convert world waypoints into joint configurations.
  • Smooth trajectories (velocity/accel limits) and export as CSV for execution.

5) Robot Execution & Twin Sync

  • A Python TCP/IP script streams joint sets to the MyCobot Pro 600.
  • MATLAB visualizes the digital twin in parallel to verify the commanded motion.

Results

  • The arm tracked the solution path with smooth end-effector motion; the digital twin mirrored execution.
  • Minor deviations attributed to camera calibration and mechanical tolerances; within expected bounds.

Future Work

  • Scale to larger/denser mazes; add dynamic obstacle handling.
  • Closed-loop visual servoing for online correction.
  • Full 3D calibration and tool-center-point (TCP) refinement.
← Back to Projects