Loci: A Step Towards Optimizing Last-Mile Delivery and Parcel Automation
Since COVID-19 outbreak, U.S. retail e-commerce revenue grew by 19.9% from $360.1 million to $431.7 million in 2020. Given such unprecedented growth in e-commerce along with global supply chain issues, the increasing demand for efficient last-mile delivery has led to an exponential growth in the global smart parcel locker market; the market size ($644.8 million in 2020) is expected to reach $1.63 billion by 2028, exhibiting a CAGR of 12.4%.
Smart parcel lockers automate the process of parcel deposit, storage, and dissemination while reducing cost on last-mile delivery (which accounts for 50% of total supply chain costs), improving parcel management, with an elegant process for parcel retrieval that are secure and available 24/7. With the rise of smart parcel lockers and the potential high-stake loss in hand, choosing the right locations for Pick-up/Drop-off (PUDO) points is crucial as it will determine the volume of demand a facility can serve and the costs of choosing wrong are often high and can lead to further disruption in last-mile capabilities.
Created out of industry need for efficient location management for PUDO points, Loci is committed to making life simpler by optimizing smart parcel locker locations to cover the maximum amount of demand points and solving courier’s problems. Whether it’s last-mile delivery at home and office, or online orders in retail, Loci is dedicated to finding the best solution for all courier and retail business customers. Loci sets out to develop a system that is designed to solve a customer’s specific needs, allowing customizable data input that leads to optimal PUDO location outputs whilst providing their users with an easy-to-use and state-of-the-art visualization tool to help them assess the optimal smart parcel locker placements. All that the users need to do is to input their parcel demand data and desired number of PUDO locations in a given area, and they will handle the rest from there.
The key elements of Loci algorithm are illustrated in the image above. First, Loci takes in user’s parcel package demand location data, convert it into dataframe and extract all sets of longitude and latitude which will then be turned into a numpy array, clean it so that it can be used later. Second, their UI takes in user’s desired number of PUDO locations along with the cleaned numpy array of demand coordinates to run a state-of-the-art Maximum Coverage Location algorithm (MCLP) which will output coordinates for the optimal locations that covers as many demand points as possible in a given area. Loci’s MCLP algorithm is powered by Gurobi, one of the most powerful and efficient mathematical optimization package that can quickly computes optimal PUDO points from the mathematical formulation of Maximum Coverage Location Problem in the context of smart parcel locker management. Lastly, Loci uses Google Map integrated with Anvil to display an interactive map that shows all optimal locations while recording all user feedback and inputs in the Anvil data table. If time permits, Loci plans to address issues regarding crime and security of packages with local crime data, weighing the importance of optimal location output by density of demand, along with taking in user’s budget constraints as input while displaying outputs like expected costs to set up a given PUDO point and expected revenue to be generated from that point.
In the future, Loci hopes to work on incorporating more custom features within the user interface where user can select different filters, sortings, and modifications, expanding their area of service beyond Berkeley, CA, incorporating environmental and contextual data, such as crime rate, household income, and willingness-to-travel, that can better help their users with PUDO location optimization and last-mile delivery issues.
Project By: Rachel Huang, Dilain Saparamadu, Charles Fan, Sofia Sayyah
Industry Mentor: Ann Snitko
Anvil Link: https://anvil.works/build#clone:SGVHMBQIEUNPNDMC=RU7HMDAMUCIWNIPWNYGJN4SP
Google Colab Link: https://colab.research.google.com/drive/1TicMdnjq2OMOqtm1PAGi_IcNHlyeaQl7?usp=sharing