CASOP Framework Optimizes Warehouse Operations with AI Pipelines

Warehouse order fulfillment involves complex, interconnected decisions such as item assignment, order batching, and picker routing. While integrated models exist, practical constraints often necessitate decomposed approaches. Current methods typically evaluate algorithms for isolated subproblems or fixed combinations, lacking a general way to determine applicable configurations, compose valid pipelines, and assess their performance across diverse warehouse settings. To address this, the Context-Aware Synthesis of Optimization Pipelines (CASOP) framework has been developed. CASOP offers a modular repository of algorithms for common order fulfillment problems, semantic data and algorithm cards to describe warehouse context and requirements, and a taxonomy to structure problems. A key component is the pipeline synthesizer, which identifies applicable algorithms for a given context and composes all valid optimization pipelines. Finally, a pipeline evaluator assesses these resulting pipelines. Demonstrated on 7 benchmark instance sets, generating over a million valid pipelines, CASOP empowers researchers and practitioners to automatically synthesize and select high-performing algorithmic solutions for warehouse operations. The software is open-source.