🤖

Jigsaw: The Ultimate Puzzle Solver Robot

Jul 16, 2024

Jigsaw: The Ultimate Puzzle Solver Robot

Introduction

  • Jigsaw: A robot designed to solve jigsaw puzzles extremely fast.
  • Development: Took 3 years to develop; 200x faster than fastest human puzzler.
  • Objective: To prepare for a robot vs. human puzzle-solving face-off.
  • Bonus: Sharing helpful tips for human puzzlers.

Human Puzzle Solving Skills

  1. Picking Up a Piece
    • Human hands have 27 bones, 34 muscles, high nerve concentration.
    • Opposable thumbs aid in grasping and manipulation.
  2. Rotating the Piece
    • Leveraged by human hand abilities.
  3. Moving Piece into Position
    • Requires whole arm; humans have the most capable arm configuration.
    • Precise movement within spatial boundaries is challenging for robots.
  4. Determining Piece Placement
    • Visual perception, pattern recognition, spatial reasoning, executive function.
    • Humans' complex brains make task seem obvious.

Jigsaw Robot Design

Step 1: Picking Up a Piece

  • Suction Cup: Tiny, specialized, used in assembly lines.
  • Solenoid Control: Connects to vacuum pump for precise pick-up/drop.

Step 2: Rotating the Piece

  • Donut Motor: Accurate to 0.005 degrees.
  • Precision Cut Analogy: Could slice a birthday cake into 65,000 pieces.

Step 3: Moving a Piece

  • Avid CNC Router: Upgraded with ClearPath industrial servo motors.
  • Accuracy: Can place pieces within 0.0005 inches.
  • Demonstration: Can move a pencil lead around a table precisely.

Step 4: Determining Piece Placement

  • Challenge: Replicating human brain’s complex pattern recognition.
  • Solution: Ignore printed images, focus on edges.
  • Edge Analysis: Use a cell phone camera to take pictures, convert edges into splines.
  • Spline Matching: Compare splines based on overlap area.

Key Developments and Challenges

  • Competitor: Inspiration and rivalry with Shane from “Stuff Made Here”.
  • Breakthrough: Ryan from Zipline coded a solution focusing on edge analysis.
  • Scaling Up: From a 12-piece to a 1000-piece all-white puzzle.
  • Error Handling: Addressing minor errors like puzzle slop and puzzle shifts.
  • Final Upgrade: Added wiggle routine for placing pieces with high precision using a z-height encoder.

Final Test and Completion

  • First Attempt: Took pictures and solved 1000-piece puzzle in under 1 minute.
  • Challenges: Tiny errors accumulated over the puzzle leading to misfits.
  • Solution: Implemented wiggle routine using a spring-loaded linear slider to ensure pieces fit snugly.
  • Successful Completion: Placing all pieces without errors, robot was ready for real-world face-off.

Human vs Robot Face-off

  • Human Competitor: Tammy McLeod, world record holder, and jigsaw puzzle champion.
  • Initial Tests: Tammy vs. Kristen Bell; Tammy dominated.
  • Final Match: Tammy vs. Jigsaw; Jigsaw won, solving a 1000-piece puzzle in 4 hours.
  • Realization: Recognized human skills and shared tips for faster puzzling.

Tips for Human Puzzle Solvers

  1. Organize Pieces: Turn all pieces over at the start.
  2. Edges Strategy: Save edges if they have distinct patterns, solve interior pieces first.
  3. Group by Features: Sort pieces by colors, textures, patterns, shapes.
  4. Shape Sorting: For similar pieces, sort by the number of ins and outs, then match.

Conclusion

  • Jigsaw’s Success: Demonstrated superior puzzle-solving abilities.
  • Human Skills Acknowledged: Practical tips for humans to improve their puzzle-solving speed.
  • Invitation to Learn: Encouraged viewers to build and program robots via CrunchLabs HackPack.

Full transcript