🚀 Optical Neuromorphic Eikonal Solver

GPU-accelerated pathfinding with 30-300× speedup!

This demo shows pathfinding with basic CPU Dijkstra for comparison. The full GPU solver is 30-300× faster using neuromorphic computing principles.

🔗 Full implementation: GitHub

⚙️ Configuration
64 512
0 50

📍 Source (Red)
0 255
0 255

🎯 Target (Blue)
0 255
0 255

🗺️ Pathfinding Visualization

Click 'Find Path' to calculate optimal route

📚 Resources & Links

🏆 Benchmark Results

Grid Size GPU Time CPU Time Speedup Accuracy Path Quality
128×128 2.3 ms 73.9 ms 32× 0.52% error 1.004× optimal
256×256 3.1 ms 275.5 ms 89× 0.55% error 1.015× optimal
512×512 4.1 ms 948.9 ms 231× 0.69% error 1.042× optimal
Average 3.3 ms 505 ms 135× 0.64% error 1.025× optimal

🧠 How It Works

The GPU solver uses neuromorphic computing principles:

  • Each grid cell maintains 4 directional memory states (N, E, S, W)
  • Information propagates like waves through the medium
  • All cells update in parallel on GPU (massively parallel)
  • System converges to optimal solution in ~2n iterations
  • No priority queues → Full parallelization

👤 Author

Francisco Angulo de Lafuente

📜 Citation

If you use this work, please cite:

@software{angulo2025optical,
  author = {Angulo de Lafuente, Francisco},
  title = {Optical Neuromorphic Eikonal Solver},
  year = {2025},
  url = {https://github.com/Agnuxo1/optical-neuromorphic-eikonal-solver}
}