In this project we want to assess the convenience of temporary storage called mini hubs in the last mile delivery. If mini hubs prove to be more convenient than van2door deliveries a competitive price can help unroll the business model to the market.
This work will as well suggest the best spots for these mini hubs and the operational advantage in handling deliveries.

Mini hubs are short retention parcel (indicated as p ) stores located in strategic places in densely populated area.
mini hubs can be of different types:

• rented areas: 3k€/month capacity 2k p , 1k€/month capacity 500 p
• minimarket: 0€5/ p capacity 50 p
• lockers: ?€/ p capacity 30 p

The mini hubs allow :

• to avoid van delivery door to door (traffic, parking, bad timing)
• to match the delivery time
• to re-iterate delivery
• to redistribute workforce
• to use less expensive vehicles

The mini hubs allow the redistribution of shifts and workforce where we consider few main categories:

• van drivers for bulk delivery
• cart drivers for door to door delivery
• merchants storing parcels
• lockers

When a customer purchases a good the parcel gets ready in one of the big store of the city. In this project we want to compare the costs of different scenarios :

• van delivery warehouse to door
• same day in the evening
• van delivery warehouse to hub, hub to door
• hub2door in the evening
• hub2door a day after

For a precise calculation we need to define few variables to calculate the total cost of the operation.

We model the transportation options

Service time is the time per parcel for:

• find a parking
• load/unload/sort the parcels
• reach the customer/co-worker
• start driving again

charge time is the time the vehicle is unavailable, swappable batteries reduce the time to zero.

total cost includes:

• driver salary
• vehicle cost (maintenance, renting)
• fuel/electricity cost

Deliveries per day is the estimated number of orders per day which will be simulated: 8k p /day

The objectives of this work are to compute and compare the costs of different operation schemes. The target is to find the most convenient location and size of the mini hubs.

The orders will be simulated randomly across the city, the delivery time is within the day at early stages. Vans and carts have different speed and operation times, they will require separated routing and time estimates.

We will model 50 pickup stores which are collecting the orders and preparing the parcels. We will locate at first the stores where big shopping malls sit.

• identify the pickup stores: using overpass-turbo with node["shop"="appliance"]({{bbox}});
• pick the minimarkets as possible locations for minihub (whitelist)
• build a graph for vans
• build a graph for carts
• divide the city in geohashes (resolution 8) and pre compute the routes from geohash to geohash
• create a batch of simulated deliveries (8k p per batch) across the city (uniform on the graph nodes)
• create the cost function for van and for cart
• cluster orders and select the closest minihub from the whitelist
• compare van2door with van2hub + cart2door

Successively additional information should be added to the project:

• isochrones for hubs

We are using a geomadi library called graph_create to download and subset a graph for Berlin. We than calculate all the geohashes with precision 8 and assign a unique node to each geohash.

We use routing_node2node to precompute all the routed distances from geohash to geohash, both for vans and for cycles.

We create a batch of different orders, for each batch we find the mini hub clustering a specific distance radius and calculate all the routes.

The simulation considers the following steps:

• generate orders
• assign orders to pickup stores
• assign batch mini hub to order
• route pickup to hub
• route hub to door
• compute averages per ride
• calculate costs for different scenarios

The simulation will allow the following outputs and analytics

• shift length
• mini hub size and position
• cost scenarios

The main output table displays the averages values per ride

The simulation provided the following results (time and distance are approssimated):
Data available under intertino

At a first sight the kilometers run between pickup stores and hubs are really large. Considering the driving time between any pickup store to any point in the city it might not be too unrealistic.

Too many hubs cause a long driving time for vans and don't increase much the service time around the hub.

The hubs are closing to a crossing in dense populated areas.