This repo exemplifies the response to the Amazon Last Mile Routing Research Challenge, which strives to incorporate actual Amazon driver experience into the best possible route planning system. The answer may be used as an Amazon SageMaker Processing Job and is based on our paper, Learning from Drivers to Address the Amazon Last Mile Routing Research Challenge. Two sample sequences produced from the same Amazon route carrying 100 parcels are displayed in the diagram below. Our approach computed the sequence on the left (driver-friendly), whereas a traditional Traveling Salesperson Problem (TSP) solver produced the sequence on the right (cost-optimal).
Refer to lastmile_usage_directions.txt for implementation