Inspiration
Modern warehouses are incredibly complex systems where space optimization directly impacts efficiency and cost. Inspired by real-world logistics challenges and the Mecalux ecosystem, we wanted to tackle a key problem: how to automatically design the optimal layout of storage bays inside a warehouse while respecting physical and operational constraints.
What it does
Our solution automatically generates optimal warehouse layouts by placing storage bays in the most efficient way possible.
It takes into account:
- Warehouse geometry
- Obstacles and restricted zones
- Ceiling height limitations
- Different bay types and capacities
Using advanced optimization techniques.
How we built it
We built a hybrid optimization engine combining:
- A Genetic Algorithm (GA) to explore thousands of layout possibilities
- Simulated Annealing (SA) to refine and escape local optima
- Computational geometry (via Shapely) to validate collisions and constraints in real time
- A Flask backend to run and monitor optimizations
- Parallel processing to significantly speed up computation
- A visual UI for real-time layout interaction
The system continuously evolves candidate solutions until it converges to a highly optimized warehouse design.
Challenges we ran into
- Handling complex geometric collisions efficiently
- Keeping performance fast despite huge search spaces
- Avoiding local optima in optimization
- Managing multiprocessing overhead and synchronization
Accomplishments that we're proud of
- Built a fully working optimization engine from scratch
- Achieved high-performance computation using parallelism
- Designed a scalable architecture adaptable to different warehouse scenarios
- Successfully combined multiple optimization techniques into one system
What we learned
- How powerful evolutionary algorithms are for real-world problems
- The importance of balancing exploration vs exploitation in optimization
- Practical use of computational geometry in logistics
- How to design systems that are both efficient and extensible
What's next for HACKUPC 2026 - Mecalux Challenge
- Integrate machine learning to predict better initial solutions
- Support dynamic warehouses (real-time changes and updates)
- Improve scalability for industrial-scale datasets
- Explore integration with real warehouse management systems (WMS)
Log in or sign up for Devpost to join the conversation.