Hocus: A 3D Puzzle Solver

Authors

• Maheen Ahmed (429551)
• Taha Shah (408351)
• Zain Ali (405704)

---

Table of Contents

1. Introduction
2. Problem Statement
3. Methodology
4. Technologies Used
5. Results
6. Conclusion
7. References

---

Introduction

Hocus is a 3D puzzle-solving project aimed at automating the solution for puzzles in the Hocus app. The game features a maze-like structure with a movable box that must be inserted into a target goal.
The solver uses a static image of the puzzle, extracts a graph representation, and generates a sequence of moves to automatically solve the puzzle in real-time.

---

Problem Statement

The challenge involves:

• Interfacing with a mobile device to capture puzzle screenshots and send swipe signals.
• Extracting the 3D graph structure from the screenshot using image processing techniques.
• Finding the optimal path using A* or BFS algorithms, depending on the graph's nature.
• Translating the solution into swipe actions to solve the puzzle automatically.

---

Methodology

Steps:

1. Screenshot Capturing:
   Use ADB (Android Debug Bridge) to capture and crop the puzzle image.

2. Image Processing:

   • Create a mask to filter grayscale pixels.
   • Apply contours to outline red objects (box and goal).
   • Segment the maze into a hexagonal grid.

3. Graph Construction:

   • Build a six-dimensional graph based on hexagon connections and intersections.
   • Define special nodes (e.g., portals) for multidimensional transitions.

4. Pathfinding:

   • Use A* or BFS to traverse the graph and identify the optimal path.

5. Execution:

   • Map the solution path to swipe actions.
   • Send swipe commands to the device in sequence to solve the puzzle.

---

Technologies Used

• Programming Languages: Python
• Libraries: OpenCV, ADB
• Algorithms: A*, BFS, Image Segmentation, Graph Construction
• Tools: Excalidraw (for visualization)

---

Results

• Accuracy: 100% (solved all 40 puzzles in the Hocus app).
• Speed: 1-2 seconds to extract and solve each puzzle.

---

Conclusion

The Hocus Puzzle Solver demonstrates the potential for real-time automation in gaming by integrating AI, image processing, and graph theory. Its generalized design enables applicability to other puzzle-solving scenarios where access to source code or training datasets is unavailable.

---

References

1. Holzinger, A., et al. (2014). Graph Extraction from Image Data. Brain Informatics and Health. DOI: 10.1007/978-3-319-09891-3_50
2. Valderhaug, V. D., et al. (2019). Deep Learning Toolbox for Neuronal Graphs. arXiv preprint.
3. Hart, P. E., et al. (1968). A* Search Algorithm. IEEE Transactions on Systems Science and Cybernetics.
4. Likhachev, M., et al. (2005). Anytime Dynamic A*. ICAPS Proceedings.