The Effect of Autonomous Trucks on Network Design
By Bruce Baring
While the jury is still out on how soon we will see driverless trucks on our interstates, there is little doubt that they are coming. Big money is being invested by well-known companies like Uber (Otto) and Google (Waymo) with a focus on semi-autonomous technologies – and smaller startups like Starsky are betting that the real opportunity is eliminating the driver all together from the long-haul equation.
We will likely see an evolution starting with semi-autonomous trucks with additional legs becoming fully automated as the technology progresses. Unfortunately, the political challenges to reach greater levels of autonomy will most likely outweigh the technical challenges.
But whatever the future holds, autonomous and driverless trucks will have a significant impact to the design of distribution networks. These new technologies will enable further distances to be travelled without the constraints of service hours. Figure 1 below demonstrates a typical logistics network today that covers most of the country’s population within one day based on a 500 mile per day reach.
Assuming autonomous vehicles stretch the reach within a day to 1000 miles, the majority of the population could be reached with 2 DCs as seen in Figure 2.
The interesting thing to note, is not the miles that can be covered in a day nor the number of DCs required, but that the locations of the DCs are different in each scenario modeled. If you were to build a typical one day network today, it is most likely that none of the locations would be best for the long term.
If we focus on cost rather than service, we realize a very similar dynamic. Driver pay can represent approximately 35-40% of the total cost per mile. If the task is fully automated and these costs go away, the balance between transportation costs and warehousing costs shifts resulting in fewer warehouses in an optimal network.
While we do not need to be too concerned about an immediate impact to network design due to autonomous vehicles, we cannot ignore their potential. When planning your network for today, anticipate future changes and make sure to run robust sensitivity models. Network planning is a long term strategic endeavor and you do not want to make long term commitments to facilities that may not be optimal long term.