12.5%. This is the percentage of products purchased online in retail sales. While the trend seems to be stabilizing, the constraints for last-mile delivery in city centers are indeed increasing. Urban planning policies are leaning towards densifying downtown areas and reducing the space allocated to cars. Although alternatives such as cycle logistics are to be encouraged, delivery trucks as we know them still have many good years ahead. To maintain operational efficiency, a growing number of players are turning to route optimization solutions. Optimizing routes is a way for logisticians to improve their efficiency, reduce stress within their teams, and minimize the carbon footprint of their fleet.
Use cases
Operating in an open environment, drivers must consider the constraints of their customers, traffic, and their payload, as well as weather, the availability of goods, etc. Given the complexity of possible combinations, route optimization products and algorithms are on the rise. Although the Vehicle Routing Problem has been studied for a long time in Operations Research, recent advancements in Artificial Intelligence enable optimization through two AI disciplines: Machine Learning (ML) and Reinforcement Learning (RL).
Enhancing the traditional approach with AI algorithms offers numerous advantages:
- determining the road segments that have outperformed at certain times of the day or under certain conditions thanks to ML models
- cross-referencing data to regularly update proposed routes
- assigning the most delicate routes to the most attentive employees
- reducing the constraints of home delivery with accurate time slot estimates
- adapting proposed routes to the vehicle category, from cargo bikes to trucks
- optimizing fleet utilization rates with predictive maintenance
All these AI-boosted functionalities help improve the financial and energy efficiency of last-mile logistics while also reducing uncertainties affecting the end-customer experience.
A tech giant still dominatingt?
Some turnkey solutions are indeed available on the market, but Google has developed its solution via an API. Within its Optimization AI service offering, Google Cloud has developed a specific product for route optimization: Cloud Fleet Routing. Based on these algorithms, the American firm is currently developing an integrated solution for operators, which will also allow for a turnkey installation. Currently, a certain level of technical knowledge is still required to use these algorithms within an internally developed solution. This solution allows for direct reliance on data from Google Maps and Waze, the collaborative application acquired by Google in 2013. This abundance of data enables Google to offer optimizations that take maximum account of the drivers' environment thanks to internally developed AI models.
A market with a multitude of players
In this well-occupied market, we also find various SaaS and platforms that allow for the integration of fleet specifics into the product with varying degrees of ease. Examples include:
- RouteQ, which focuses on AI use in its algorithms
- Circuit for small teams with few operational particularities
- Routific for small local delivery businesseses
- Route4Me if maximum flexibility is needed
- OptimoRoute for sales and service teams
- OnFleet for SMEs and large companies
The products available on the market, as you will have understood, are numerous. This multiplicity of actors reflects the complexity of the problem and the constraints to be managed (planning window, objective to achieve, etc.).
And Nelson in all this?
Nelson starts from a simple observation: the electrification of vehicle fleets adds yet another constraint and criterion to the list of factors to be integrated into already complex models. The start-up team specializes in route optimization integrating recharge stops. Several options are thus possible:
- Maintaining identical routes, with recharges inserted during site visits, for example,
- Redistribution of routes and simulation of saved km (or miles) / CO2,
- Sizing on-site charging stations to minimize the need for roaming charging, and many more!
and many others !
The economical solution with open source
As often in AI, Plug & Play route optimization products coexist with open-source libraries. Although these are primarily focused on solving the Vehicle Routing Problem (VRP) with the possibility of adding several constraints, the suggestions from these algorithms can also be boosted with existing fleet data. This is advisable if you have enough resources and traffic to take the time to develop this internal tool. It will be even easier if you equip your fleet with telematics boxes! In this category, you can look into Graphhopper and its jspirit library written in java, or the project VROOM. You will find others here.
If you decide to take your last-mile delivery to the next level, it seems wise to look into the various route optimization solutions available on the market. You will surely find the tool adapted to your company based on your human and financial resources, fleet, and collaborators. A number of these solutions use AI models to propose routes that take more factors into account. These new tools will help you achieve significant operational gains to benefit your customers' experience. Nelson will also be able to assist you in refining your optimization with the scaling of your electrification.
Read all the « Expert Opinion » articles on the SprintProject blog
