Dynamic Human-Robot Collaborative Picking Strategies

In the last decades, many retailers have started to combine traditional store deliveries with fulfilment of online sales to consumers, from omnichannel warehouses, which are increasingly automated. One popular way of warehouse automation is with Autonomous Mobile Robots (AMRs), that collaborate with human pickers to efficiently pick the orders by reducing the pickers' unproductive walking time. Picker travel time can be reduced even more by zoning the storage system, where robots take care of the travel between these zones. However, the optimal zoning strategy for these robotic systems is not clear: few zones are particularly good for the large store orders, while many zones are particularly good for the small online orders. We therefore study the effect of dynamic zoning strategies, i.e. dynamic switching between a No Zoning (NZ) strategy and a Progressive Zoning (PZ) strategy. We solve the problem in two stages. First, we develop queuing network models to obtain load-dependent pick throughput rates corresponding to a given number of AMRs and a picking strategy with a fixed number of zones. Then, we develop a Markov-decision model to investigate how higher pick performance can be achieved by dynamically switching between these pick strategies. Using data from an omnichannel warehouse that processes various order sizes, we show that a Dynamic Switching (DS) policy can lower operational cost by up to 7 percent. However, these cost savings decrease as the number of robots per picker increases.