Applying GPU programming to obtain average AS/RS performance in optimal zoned aisles

Abstract

This paper presents a methodology to find optimal zones in automated storage/ retrieval systems (AS/RS) and derive the performance gain for different operating strategies. One central feature of the described method is the computation of precise results in very short time (less than one second in most cases on standard hardware) for any given rack shape independent of interface alignment. These short runtimes lead to new openings in warehouse planning and control: Warehouse planning is broadly driven by the usage of experience and analytical models without respect to influences through operating strategies in early design stages. Embedding the presented method in specially tailored software enables the planner to evaluate the possible system performance achieved under a class-based storage strategy. Additionally the computed results are utilized to assess performance gains using multi-shuttle storage and retrieval systems with advanced routing and can be easily extended for double-deep systems. Furthermore, warehouse control systems can adopt the introduced models to calculate optimal article to zone associations continuously and improve overall performance in existing systems. Short runtimes are realized by a new procedure to determine optimal zones in the rack and thereon a highly parallelized algorithm calculates the average performance for a given system on usual graphic processors.