What’s the advantage of real time for route optimization?

Route planners often lack visibility on what is happening in the field. Most of the time, they build their route plans manually. However, optimizing routes is a complex issue, frequently disrupted by numerous constraints (time slots, vehicle capacity, available drivers, road traffic, etc.) that are not always easy to anticipate. An address error, an absent customer, traffic jams, a late driver… hazards can quickly arise during a tour. Planners find themselves struggling to react efficiently to these complications. This threatens their entire organization but also their quality of service: their ability to adapt and find an alternative solution is challenged so as to guarantee the best possible experience to their customers.

Different optimization softwares are available to relieve operational staff of this tedious task. However, they are often not adapted to the complexity of the activity and do not always take into account their constraints. Routes are therefore no longer viable in the field. To meet this challenge, Kardinal has developed a route optimization solution that “never stops” in order to offer even more efficient routes that are perfectly adapted to field reality.

What is continuous optimization? In which cases is it relevant to use this type of technology? What does it mean for users?

Cédric Hervet, Scientific Director and co-founder of Kardinal, answers these questions to explain the importance of real-time and continuous route optimization.

Cédric Hervet, Head of Science @Kardinal

Cédric Hervet has a PhD in Applied Mathematics and co-founded Kardinal. For over 10 years, he has been studying and designing Artificial Intelligence systems for industrial applications in the telecommunications, digital marketing and transportation industries.

His dual expertise in statistics/machine learning and algorithms/operational research allows him to combine these two sets of techniques to design the intelligent systems of tomorrow.

What is real-time route optimization exactly?

Cédric Hervet: First of all, we must make a distinction between real-time optimization and continuous optimization. “Real-time” means that the optimization is done at each moment, adapting to the events that are taking place. It is thus possible to react immediately in case of a problem: change the routes by assigning certain points to be visited to other drivers, proactively inform customers of the delay of their delivery, etc.

Continuous” optimization never stops optimizing: before, during and after operations. From the first package that enters the system to the last delivery made, the software continuously optimizes during this period and suggests recommendations. The planner never needs to proactively restart the optimization with new data. The algorithm is constantly connected to the real world and has permanent exchanges with the information system.

By combining real-time and continuous optimization, the routes will be more efficient and adapted to field reality. All data defining the day’s activity (orders/tasks to be carried out and available resources) are integrated into the global optimization plan.

Data sources can be multiple:

  • The driver notifying via an app that he is arriving at the delivery point and that he will leave,
  • The operator on the platform loading the truck reporting that the actual weight of the package does not match the declared weight,
  • New orders entering the system,
  • and so on.

All these changes occurring over time will be matched by the machine to a single schedule. At any time, the logistician can check the routes suggested by the algorithm (which are the best possible routes with the current known data) in order to make decisions when needed.

Why did you create a real-time and continuous optimization solution?

Cédric Hervet: Route optimization is not a new technology, the first research on the matter dates back to the 1970s when mathematicians tried to solve this problem with the creation of dedicated software. However, when Kardinal was created, we realized that despite the existence of these technologies, the vast majority of logisticians did not use them and relied instead on tools such as Excel.

One of the reasons for this was that digital literacy was not strong in these companies yet, as digitalization was relatively recent in these areas. The difficulty is to have a tool that blends in well with the client’s process and does not disrupt their working process. Route optimization tools are rarely able to manage all the constraints and complexity of the field, which generates frustration with regard to the routes generated and leads to projects being abandoned. Continuous and real-time optimization allows for a much smoother integration into processes. It adapts to the complexity of the activity and allows planners to react very quickly to unforeseen events.

Another important failure element of this type of project comes from the data: the company lacks data or, if not, the data is of poor quality, which distorts all the planned optimizations (for example: the delay of a tour is not reported to the planner). Thanks to continuous optimization, it is possible to solve this problem: one of the ways to manage flawed data is to be able to correct it while the optimization is in progress.

Why use this type of optimization?

Cédric Hervet: Real-time optimization is performed during the execution of tours, especially to react to delays caused by traffic jams. But it is also important, even more so, upstream of the tours because the planning process is not fixed (it is very rare that the routes all leave at the same time, so it is possible to add delivery points to the trucks leaving later) and it is necessary to be able to readjust the routes as decisions are made.

A route optimization can, for example, start 2 days before the tour, as it was the case for one of our customers, a home furniture delivery company: orders arrive 2 days before, and a process to make appointments is set up to fix the tours. The algorithm suggests the best time to place a delivery, taking into account the slots already scheduled.

It is essential to take into account traffic during the tours but also beforehand thanks to predictive traffic. The planned routes will thus be more realistic, as traffic is a difficult variable to anticipate. This is essential for parcel delivery companies where the last parcels enter the agency 1 or 2 hours before the departure of trucks. Real-time and continuous optimization updates the routes according to the traffic data that comes in.

In case of unforeseen events such as a driver’s delay, continuous optimization is essential. This makes it possible to change the tours, to better organize the visits or to assign them to other drivers, especially if some of them have time slots to be observed. Real-time and continuous optimization solutions therefore avoid operational staff having to completely change their working habits to adapt to the software. At Kardinal, we aim for the opposite: it’s the machine that adapts to humans.

What does this mean technologically?

Cédric Hervet: This type of optimization requires a software architecture that allows for a continuous optimization engine that can integrate new information at any time to update any data. This is precisely the technology that Kardinal has been developing. For logisticians, it is essential to be able to feed the algorithms with data more or less continuously. Depending on their real-time optimization needs, some export their data to Kardinal’s systems more or less frequently.

The technology developed by Kardinal relies on data from its clients’ information systems (TMS, ERP), which must have the ability to proactively send this information as soon as it is updated. Thanks to an API connection, the data is integrated into the optimization solution. The customer can make changes directly on an app connected to the Kardinal software solution.

What impact for users?

Cédric Hervet: Most route optimization software use static optimization. There are several steps:

  1. The data is added to the solution
  2. The planner launches the optimization
  3. The solution returns, after some time, a fixed route plan

This result can be applied as is or modified by hand. In contrast to continuous optimization, the software’s response is definitive and the optimization ends. With continuous optimization, the slightest change in data generates a new tour in a few seconds without having to restart it.

In the case of static optimization, planners wait for the software’s feedback before working. Continuous optimization, on the other hand, is based on their needs: if they have to make a decision, then at any time they have the best possible plan at their disposal. If drivers are waiting for their instructions, the planner can already make decisions.

However, it is not possible for the optimization to work endlessly. If actions are taken in the real world, the algorithms must be informed of this so that they do not touch what has been decided. Continuous optimization implies that we must reproduce this natural boundary in static optimization so that the human can take back control of the operations. With Kardinal, the user can, at any time, decide to lock a route plan that is being optimized, either entirely or partially. The user is able to add or remove routes from the optimization process with ease.

This new way of thinking about the relationship between the operational staff and the algorithm is what we at Kardinal consider to be the foundation of the route optimization of the future.