Research project investigating search and tracking algorithms for movement through dynamic environments, under the guidance of Dr. Triet Le. The project focused on implementing the Fast Sweeping Method for efficient path discovery while incorporating terrain and elevation data into movement-cost models.
The resulting simulation framework combined geospatial data processing, terrain-aware cost modeling, and GPU-accelerated computation to evaluate movement through complex, 3D environments.
- Technologies UsedPython (RichDem, Elevation, GDAL, Matplotlib), CUDA, Google Colab
The core of the project was an implementation of the Fast Sweeping Method, a numerical approach for efficiently solving the Eikonal equation and computing cost-to-go information across an environment. This provided a foundation for determining efficient movement paths through dynamic terrain.
The simulation also incorporated elevation and terrain characteristics into the pathfinding model. Tobler's hiking function was applied to estimate movement speed based on terrain slope. This allowed the cost model to adapt to the difficulty of traversing different areas, producing paths that account for realistic movement constraints.
Elevation and terrain data were processed through an automated geospatial pipeline using GDAL and RichDEM. The pipeline prepared terrain information for use by the simulation and transformed elevation data into inputs for terrain-cost calculations. Matplotlib was then used to visualise the resulting terrain, movement costs, and pathfinding behaviour.
Because the search algorithms repeatedly evaluate large numbers of positions, the project also focused on computational optimisation. Parallel processing, NumPy-based optimisations, and CUDA acceleration were used to move computationally intensive operations onto the GPU. This reduced simulation runtime by approximately 10%.
The project was organised as a scalable, object-oriented simulation framework so that different search and tracking approaches could be tested within a common environment.
