Background
A fulfillment technology company’s warehouse routing system, built on a standard shortest-path solver, worked well for smaller facilities but produced increasingly inefficient pick paths as clients expanded into larger, more complex multi-zone warehouses.
The Challenge
The team needed a routing approach that could compute near-optimal pick paths within a strict time budget for warehouses with thousands of SKUs and dynamic order batches, a scale at which the standard solver’s computation time became impractical.
Technological Uncertainty
It was not known in advance whether any combination of algorithmic approximation and heuristic batching could achieve near-optimal routing within the required computation time at this warehouse scale, or whether the problem’s complexity made real-time optimization infeasible at this size.
Experimental Development
The team systematically tested multiple heuristic and approximation algorithms, order-batching strategies, and zone-partitioning approaches, measuring path efficiency, computation time, and solution quality against the standard solver’s baseline across representative warehouse layouts.
What Failed?
An initial heuristic approach met the computation time budget but produced pick paths meaningfully less efficient than the baseline on the largest warehouse configurations, requiring the team to investigate a hybrid zone-partitioning strategy.
Technological Advancement
The team developed a hybrid algorithm combining zone-based partitioning with a refined heuristic solver, achieving near-optimal routing efficiency within the required computation budget across warehouse sizes the original solver couldn’t handle, generating new technical knowledge about real-time routing optimization at this scale.
Potentially Relevant SR&ED Activities
● Systematic testing of heuristic and approximation algorithms against computation time and efficiency targets
● Order-batching and zone-partitioning strategy experiments
● Path efficiency benchmarking against the standard solver baseline across warehouse configurations
What Would Generally Not Qualify
Applying the finalized routing algorithm to additional warehouses of similar size and layout using the established configuration would be routine deployment and would not itself qualify.
Documentation
Computation time and path efficiency benchmarks across algorithm configurations, technical notes on why the standard solver became impractical at scale, and comparative analysis against the baseline would support this claim.
