Warehouse Optimizer

Optimizing warehouse layouts in real-time using intelligent packing, shared access zones, and multi-threaded search.

Comment

What it does

OptiRack is an intelligent warehouse optimization system that automatically generates high-efficiency layouts for storage racks (bays) under real-world constraints.

Given a warehouse geometry, obstacles, ceiling constraints, and rack types, our system finds an optimal placement strategy that maximizes storage capacity while minimizing cost.

Unlike naive solutions, OptiRack handles:

  • Complex warehouse geometries
  • Obstacle avoidance
  • Directional access zones (shared gaps between racks)
  • Arbitrary rack rotations
  • Real-time manual editing and validation

How we built it

We built a hybrid system combining:

  • C++ high-performance optimizer

    • Multi-threaded execution (parallel seeds)
    • Simulated annealing with custom move operators
    • Ruin-and-recreate strategy
    • Adaptive heuristics
    • Gap-aware packing model

  • Custom geometry engine

    • Rotated rectangle collision detection (SAT)
    • Polygon containment checks
    • Spatial acceleration structures

  • Interactive web visualizer

    • SVG-based rendering
    • Real-time validation
    • Manual editing of rack placements
    • Live recomputation of optimization score (Q)

Key innovation

Shared Access Zones (Gap Optimization)

Instead of treating gaps as fixed margins, we model them as shared directional access zones, allowing opposing racks to share the same corridor.

This dramatically improves packing density and overall score.

Arbitrary Rotation Optimization

Unlike traditional approaches limited to 90° rotations, our system supports arbitrary angles, enabling:

  • Better fit in irregular spaces
  • Higher density layouts
  • Improved score optimization

Parallel Multi-Seed Optimization

We leverage multi-threading to explore multiple solutions in parallel:

  • Each thread runs an independent optimization
  • Best solution is selected in real-time
  • Maximizes performance within strict time limits

Challenges we ran into

  • Handling rotated geometry efficiently
  • Designing a correct gap model without double-counting space
  • Balancing optimization quality with strict runtime limits (20–30 seconds)
  • Avoiding fragmentation in irregular spaces
  • Keeping the system interactive while maintaining correctness

Accomplishments that we're proud of

  • Built a full end-to-end system (optimizer + UI)
  • Implemented a correct and efficient shared-gap model
  • Achieved strong packing density under constraints
  • Enabled manual optimization on top of AI results
  • Built a system robust enough for real-time demo

What we learned

  • Packing problems are much harder than they look
  • Small modeling mistakes (like gap duplication) destroy performance
  • Multi-start + parallelization is extremely powerful
  • Visualization is critical for understanding optimization

What's next

  • GPU acceleration for even faster optimization
  • Machine learning-based heuristics
  • Real-world warehouse integration
  • 3D optimization (height-aware stacking)
  • Automated layout suggestions based on usage patterns
Share this project:

Updates