Horizontal Collaboration and AI for better urban logistics
Philippe Rapin, Advisor and Contributor at Urban AI, is the CEO of Urban Radar. Urban Radar is a cloud platform that revolutionises the way cities understand, manage and regulate their response to the exponentially growing urban logistics demand.
The modern challenge of Urban Logistics
The ongoing logistics revolution is changing how cities behave and organise. The accelerated growth in e-commerce and Just-in-Time logistics is increasing the demand of public space needed to accommodate the surge in logistics activity. Before the COVID-19 pandemic, the World Economic Forum projected an annual growth of 17% in e-commerce, while the pandemic has shown signs to have accelerated this trend by at least 5 years. This trend raises concerns over the impacts that increased vehicle trips can cause on quality of life, pollution, congestion, and safety if not properly managed. While many players in the logistics industry have embraced sophisticated AI driven solutions to optimise their operations and minimise their impact, the common approach is for each company to individually optimise their operations. This often explains why you see several half-empty trucks parked next to each other when a single vehicle could have sufficed. Advances in AI are powering a paradigm shift to address this problem, with collaboration and global efficiency at its core.
What is horizontal collaboration in logistics?
Modern urban logistics have the potential to operate in an integrated way. When several logistics companies collaborate to improve their operations, they can unlock operational efficiencies that benefit both cities and businesses.
Horizontal collaboration, known as collaboration between similar (and sometimes competing) players, has been widely used in some industries (e.g. airline industry where “alliances” are common, or in maritime transport). In urban logistics, horizontal collaboration can take place on multiple levels: mainly by sharing routes, vehicles, and warehouses/hubs. Sharing of routes and vehicles optimizes the supply chain by reducing half-empty trips. Shared warehouses allow higher utilization rates of storage space, lowering expenditures and reducing land use requirements.
Horizontal collaboration brings benefits to all city stakeholders. From a city perspective it can lead to less traffic, less pollution and noise, and improved road safety with the same or even better logistics productivity. Customer satisfaction also improves since order bundling can reduce the frequency and number of deliveries, delivering the same amount of goods with fewer trips. Moreover, increased storage and transport options can improve flexibility, reliability, and redundancy of the supply chain. Horizontal Collaboration has been applied through individual partnerships with positive results. A french report (1) from 2009 cites three collaboration projects. One project involved three food industry companies, in which they managed to raise the frequency of their deliveries by 34%, doubled the numbers of pallets by truck, decreased the stock covering by 16% and reduced by 56% the number of kilometers travelled. The second project also involving food industry companies, raised the service rate by half a point, reduced stocks up to 38%, reduced deliveries by 35% and saved 93 tons of CO2 emissions per year. The last project involving consumer goods companies, resulted in increases between 75% and 100% in the loading rate of their trucks, reduced by 20% the number vehicles used and kilometers travelled, and cut down 3150 tons of CO2 emissions per year.
Despite their potential benefits, horizontal collaboration can have its downsides. Logistics operators don’t have complete control of the supply chain anymore, and instead it’s required reliable coordination with other businesses, which generally translates into new investments. Some practices can help build these collaborations, such as physical proximity, relationship management, and incentives alignment, but first and foremost collaboration requires constant sharing of operations information.
Information sharing is primordial since supply chains become entangled and compatible. Nevertheless, at the scale of a metropolis, ad-hoc coordination between multiple logistics players can quickly become unsustainable. Hundreds of hubs, thousands of vehicles, and millions of parcels can render large scale collaboration efforts impossible without a “coordinator” entity that manages operations at a city scale. Benefits obtained as a product of horizontal collaboration can be distributed among the players by the coordinator applying concepts of cooperative games from game theory.
Horizontal collaboration needs strong coordination
As shown in the figure below, collaboration can create new possibilities for operations, but collaboration without effective coordination at city scale can do more harm than good. Only when all elements in a logistics network are effectively coordinated can the benefits of horizontal collaboration be fully realised.
Above are represented different stages of the supply chain. Figure 2.1 represents a non-collaborative supply chain where each company operates individually. Figure 2.2 represents a collaborative but uncoordinated supply chain: warehouses and routes are inefficiently shared. It increases the cost of operation and potential externalities such as added congestion and pollution. Figure 2.3 is what a collaborative supply chain is aiming for: complementarity. In this case, routes and use of warehouses are globally optimised to pursue a city wide optimal allocation of logistics resources.
But at a city scale, coordinating operations by computing efficient routes and optimal use of resources can be computationally infeasible using classic approaches of optimization. Solving this challenge can be quite difficult, but recent advances in AI applied to logistics can offer previously unavailable solutions, possible.
AI as an enabler of Horizontal Collaboration
In logistics, AI applications have proven effective to find optimal sets of routes for vehicles to satisfy customers demands; but this activity can become quite difficult at city scale (2).
Thanks to advances in computational capacity and data collection, multiple AI techniques can be now applied in real scenarios, showing promising close to the optimal solutions with a really fast computational time.
To apply horizontal collaboration at city scale, several AI approaches are available such as route optimization with multiple actors. For instance, one approach is the use of simple heuristics. Many heuristics exist and have been already explored for this problem. Ant colony optimisation (3) and simulated annealing offer near-optimal results, but can be slow for real-life applications. Reinforcement learning (4) has also been explored recently under simplified scenarios, the strength of this method is that it scales well with increasing problem size and has a competitive solution-time with a close to optimal solution. Digging further, neural networks (5) are also quite effective to propose near to optimal solutions, especially when there is abundance of operational data. Another important benefit is that specialised neural networks can accept many inputs and constraints common in a city, such as circulation regulations, congestion, geographic factors, periodic patterns, etc. Application of these AI methods can be the missing piece to scale coordinated horizontal collaboration for an entire city. Urban Radar has also worked on implementing a solution based on a reinforcement learning algorithm and is currently making progress on applying more specialised AI techniques for real life applications.
The moment of change
It is unquestionable that city logistics are getting more relevant for cities to achieve sustainability goals, and minimize congestion and emissions. Increasing city logistics activity demands taking a citywide approach to coordination regardless of having multiple competing operators. Just as other industries have identified the benefits of horizontal collaboration, it is time for city logistics to take this approach more seriously. The advances of AI algorithms in resource allocation and optimisation makes it now possible to have a role of citywide logistics coordinator, capable of minimising half-empty vehicles and warehouses and distributing the benefits of collaboration systematically. It is only a matter of time and political will to see this win-win approach for cities and industry to take over all around the world.
By Philippe Rapin, with the help of Geoffrey Bir, Joseph Desquaires, Hernàn Mejia and Raphael Collin.
This article is also available on our Medium Publication : https://medium.com/urban-ai/horizontal-collaboration-and-ai-for-better-urban-logistics-d3fc6a04404
(1) : PIPAME et CNAM, 2009 Logistique mutualisée : la filière « fruits et légumes » du Marché d’Intérêt National de Rungis.Ministère de l’économie. Paris, Pôle interministériel de prospective et d’anticipation des mutations économiques et Conservatoire National des Arts et Métiers. Page 47–48
(2) : Ben Ticha, Hamza & Absi, Nabil & Feillet, Dominique & Quilliot, Alain. (2018). Vehicle routing problems with road-network information: State of the art. Networks. 72. 10.1002/net.21808
(3) : John E. Bell, Patrick R. McMullen, Ant colony optimization techniques for the vehicle routing problem, Advanced Engineering Informatics, Volume 18, Issue 1, 2004, Pages 41–48, ISSN 1474–0346
(4) : Nazari, Mohammadreza & Oroojlooy jadid, Afshin & Snyder, Lawrence & Takáč, Martin. (2018). Deep Reinforcement Learning for Solving the Vehicle Routing Problem
(5) : Nazari, Mohammadreza & Oroojlooy jadid, Afshin & Snyder, Lawrence & Takáč, Martin. (2018). Deep Reinforcement Learning for Solving the Vehicle Routing Problem