Last Mile Logistics

Ira Winder, MIT Media Lab
Matthias Winkenbach, MIT Center for Transportation and Logistics

(originally published on our blog)

MIT Media Lab Changing Places Group and MIT Center for Transportation and Logistics are developing a decision support tool for calculating delivery service areas. Logistics experts can use the platform to present parametric models of logistics in a real-time, changeable environment. Researchers expect the tool to improve collaboration and consensus when optimizing distribution networks for last mile logistics.

The tool uses the tangible interactive matrix (TIM) developed at MIT Changing Places Group. TIM uses an array of optically tagged Lego objects, computer vision, and 3D projection mapping.

Users operate the tool by manipulating tangible objects that represent distribution centers. All together, the objects represent a distribution network. Meanwhile, algorithms provide real-time performance evaluation of the users’ configuration. Key performance metrics in a demonstration include average delivery cost and customer demand saturation.

The use of both tangible bricks and geospatial models led us to adopt a voxel-based method for data abstraction.  (Note: a voxel is a multi-dimensional pixel).  The result is a mathematical model uniquely structured to be compatible with TIM (Fig. 2).

Figure 2. A typical GIS polygon construct (left) is translated into a TIM-compatible voxel and Lego construct (right).

Figure 2. A typical GIS polygon construct (left) is translated into a TIM-compatible voxel and Lego construct (right).

GIS data such as US Census parcels are processed and cleaned to be compatible with the system at three scales: 2km, 1km, and 500m per pixel.  In this scenario, we use population as a proxy for demand (Fig. 3).

Figure 3. Voxel-ized data can represent different regions at different scales.

Figure 3. Voxel-ized data can represent different regions at different scales.

Average delivery cost is a function of both distance traveled from distribution centers and the density of deliveries made at the “last mile”. Average delivery cost “C” is proportional to customer’s distance from a distribution center “D” divided by density of customers at last mile,  “ρ” (source: MIT Center for Transportation and Logistics).

C α D / ρ

The “last mile” refers to the short but most difficult last leg of a journey, such as a walk from a subway station to home. In the case of delivery logistics, the last mile can refer to the difficulty of handing off packages to customers at home or finding short-term parking. Cost is reduced when many drop-offs can occur within a small area. 

Figure 5. Service area solutions for 3 different placements of a single distribution center.

Figure 5. Service area solutions for 3 different placements of a single distribution center.

Customer demand is saturated when a distribution center has capacity to serve a given area.  Service areas are automatically allocated in a global manner such that average cost is minimized (Fig. 5). The result is often a non-intuitive pattern of service areas (Fig. 6).

Figure 6. Complex solution with 5 distribution centers of various capacities. Green denotes cheaper areas to serve, while red areas are more expensive.

Figure 7. Delivery service areas that are more expensive to serve are colored red while areas cheaper to serve are colored green.  Areas further away from a collection of distribution centers are more expensive to serve (left).  A strategically located distribution center significantly reduces the cost of serving deliveries in the area (right). Photos: Ira Winder

SPECIAL THANKS

Daniel Merchan
Edgar Blanco
Brandon Martin-Anderson
Mike Winder
Nina Lutz
James Li