The Butterfly Effect
Most of us have heard of the “Butterfly Effect” — the phenomenon where the flapping of butterfly wings in one location can affect how weather plays out in a distant location. This is an example of nonlinear system behavior, or how a small change in conditions in one location can result in large changes in another.
Logistics transportation systems behave the same when they are running near capacity. Small increases in flow can make a system become congested, unstable, unbalanced or even oscillate between periods of predictable and highly unpredictable behavior.
In addition to frequent small disturbances, logistics systems also experience larger, external disruptions from time to time. For example, mechanical component or vehicle failure or disruptions caused by late arrivals due to extreme weather. Looking long-term, there are also technology disruptions that change the behavior of customers, e.g. e-commerce or new technologies such as autonomous transportation vehicles.
Hub & Spoke versus Point to Point
A concrete example of ongoing disruption is provided by the relative merits of point-to-point (P2P) versus hub and spoke (H&S) architectures for logistic networks. Prior to the 1970s, logistics networks were dominated by P2P topologies, which have one direct route from each origin and destination node. The H&S architecture was introduced as a way to reduce the number of routes between each node. Fewer routes means higher traffic volumes per route, resulting in additional benefits from the economy of scale for larger capacity equipment.
However, there have been several disruptions to the dominance of the H&S architecture for manufacturers and retailers. For example, e-commerce has changed customer behavior, increasing the expectation of rapid delivery and demand for expanded inventories. This includes items that aren’t easy to on- and off-load from a vehicle or pass through a hub via an automated conveyer system. The pressure of time-critical deliveries also places strong demands on sorting windows at the hub as well, and the total travel time is longer in these H&S networks due to additional processing in the hub and the longer route lengths.
As a result, many hubs are now running near capacity and becoming more sensitive to instabilities that can quickly starve or flood the process. The situation becomes even more complicated with the introduction of large multi-modal hubs (air/land/sea) and new options in last mile delivery by drones or autonomous vehicles. Further ahead there is the possibility of forms of modular containerized land transport where some aspects of sortation (smaller container exchange) could be done by small numbers of vehicles meeting temporarily at a road-side “virtual hub.”
Because of these factors, the future of logistics for these retailers will involve hybrid architectures that combine elements of the H&S, P2P and “linear” run topologies. The structures that are most efficient can vary spatially, but also by season or even day of week, so dynamic flexibility and agility of the network will become a key requirement in shipping goods.
A New Generation of Shipping
New software tools and technologies, including machine learning, Artificial Intelligence (AI), IoT sensors and new techniques of Optimization Under Uncertainty (OUU) are now available. However, it is not broadly realized that these models have no concept of causation and really only look at correlations present in vast amounts of historical data.
What is required is a holistic solution, using IoT data infrastructure, computer vision sensors and package scan data to better observe and estimate the current system state, and new kinds of optimization models that can provide rolling adjustments in response to early error signals. New hybrid models are required where OUU works as a master alongside machine learning/AI sub-models which assist in providing rolling forecasts of demand, disruption, weather, resource availability, asset failure prediction etc.
The next generation of shipping logistics will give retailers and manufacturers the ability to design and operate networks with an optimal tradeoff between cost efficiency and robustness to disruption through proactive network readjustment. Service disruption risk levels can be chosen in advance and charged to customers accordingly. The nonlinearities of chaos can finally be used to our advantage.
