The last mile of the logistics chain poses a significant challenge for delivery companies. Alone, it accounts for 30% of urban center traffic, 25% of CO2 emissions, and 20% of total delivery costs. These figures underscore the critical importance of effective optimization to reduce environmental impact and operational expenses. By reducing costs through better resource management and minimizing distance traveled, route optimization emerges as an indispensable solution. However, the variability of traffic complicates this optimization process. Fluctuations in traffic patterns make it challenging to adhere to commitments made by carriers, particularly regarding time slots.

The Limitations of Conventional Optimization Models

Conventional optimization algorithms rely on a “distance matrix” to determine the duration and distance between each pair of points for optimization. The goal is to design routes that minimize distance traveled and/or travel time. However, in classical route optimization models, this distance matrix does not account for temporal variations. Consequently, the estimation of the duration and/or distance between two points remains fixed, regardless of the departure time. To anticipate the potential impact of actual traffic on travel times, adjustments in the form of coefficients, varying in sophistication, are applied to the distance matrix to obtain more realistic solutions. However, even with these adjustments, the matrix remains time-independent, with the duration between each point remaining constant throughout the day.

The Integration of Traffic for Smarter Urban Optimization

By incorporating traffic forecasts into the optimization algorithm from the outset, it becomes possible to execute certain routes at strategic times, thereby avoiding periods of heavy congestion. Conversely, if traffic impact is only calculated post-hoc, traffic variations cannot be anticipated during planning. This can lead to routes encountering numerous bottlenecks, resulting in delivery delays and reduced overall performance. For a realistic estimation of traffic impact, it is necessary to know the traffic conditions in advance for each pair of points at every moment of the day. For instance, to plan for 500 points, it would be necessary to consider 250,000 possible routes over 48 time intervals, totaling 12 million routes. Such an approach would require several hours, if not days, of computation, rendering it impractical for the real-time optimization needs of the last mile.

Kardinal’s Innovative Solution

At Kardinal, we have developed a solution distinguished by its Artificial Intelligence system capable of collecting and aggregating large-scale traffic prediction data. This allows us to estimate, in advance, the duration of each journey for every moment and route, with a computation time comparable to a traffic-free approach. Concretely, we can predict the duration of a journey with traffic between two points, A and B, at a specific time. Our predictive model is tailored to each journey, every day of the week, and each hour of the day. Moreover, our route optimization algorithms are inherently designed to work based on time-dependent distance matrices. This enables us to build routes that avoid congested routes during peak hours and capitalize on quieter periods for deliveries in highly urban areas when possible.

Tangible Benefits Witnessed by Our Clients

As a result, our users experience more efficient and logical routes from the perspective of field teams. In Paris, for example, our users confirm that Kardinal algorithms naturally avoid the ring road and the most congested routes during peak hours. They also avoid planning routes that involve multiple crossings of the River Seine bridges. This approach offers routes better suited to the realities of urban traffic, reducing costs, CO2 emissions, and improving adherence to delivery time slots. The integration of traffic into route optimization revolutionizes urban logistics, making the last mile more efficient and sustainable.