The distribution of parcels from local distribution centres to customers can be met with hiccups, as customers may be unavailable to receive packages at optimal delivery times.
Jhonny Pincay-Nieves and colleagues, have developed a framework for improving first-try success in last-mile delivery, demonstrating how complex processes and improvements can be performed using approximate, or ‘fuzzy logic’ based methods.
Read the original research: doi.org/10.1007/978-3-031-16704-1
Listen to more ‘Fuzzy Logic’ podcasts
Find out more about the FMsquare Foundation
Image Source: Adobe Stock Images / TensorSpark
Hello and welcome to ResearchPod. Thank you for listening and joining us today.
In this episode, we look at the work of Jhonny Pincay-Nieves in collaboration with Swiss Post, the national postal company of Switzerland, and Viasuisse, the traffic competence information center of Switzerland, as well as the University of Fribourg as the Faculty prize awarding institution. In his work, Pincay-Nieves demonstrates how complex processes and improvements can be performed using approximate, or ‘fuzzy’ methods that do not require large amounts of precise information, while still achieving good results. In short, how can we feasibly ‘Smartify’ our cities?
Mobility is crucial in the functioning of a city. Smart cities aim to enhance the city services to improve the quality of life and inclusion of people living or working there. There are several dimensions that make a city “smart”: Smart Economy, Smart Government, Smart Living, Smart Environment, and Smart Mobility. From these components, mobility is considered one of the most difficult to address given that it is directly related to environmental and economic aspects.
Pincay-Nieves attempts to improve, or ‘smartify’, the service area of mobility, developing a framework for improving the first-try success in the last-mile delivery. The last-mile delivery refers to the distribution of parcels from local distribution centers to the customers. It is crucial because a whole supply chain could be in its optimal state but if the person is not at home when the delivery team rings the bell, the whole process will basically fail. In conceptualizing the framework for working towards a ‘smart’ city, three artifacts were developed: an analysis tool for traffic areas addressing the uncertainty and incompleteness of geospatial data, a linguistic traffic summarizer, and a customer classifier that does not compromise their privacy.
‘Smart’ Mobility intends to optimize the traffic flow and services that use the road and street networks. When services are not related to public transportation, for instance, logistics and delivery services, then the concept of Smart Logistics is adopted. Smart Logistics solutions are frequently built with traffic-related data. The data can be difficult to examine and understand as they differ in type and scale. What’s more, transportation and traffic parameters are defined in uncertain, imprecise, ambiguous, and subjective terms.
Research efforts in the field of advanced travel information systems and Smart Logistics have, therefore, been directed toward determining where and when a traffic anomaly occurs and by how much travel time increases; most privately and publicly owned fleet vehicles are fitted with GPS devices, which allow their time-stamped positions, speed, and direction to be recorded. This is a convenient source of data that eases the implementation of smart logistic systems. However, these data come with the disadvantages of losing local context, incompleteness, sparsity, and the complexities of business- specific operations.
These are the main reasons why little to no attention is given to traffic data from logistics or delivery vehicles. Another issue is the low sample rate, since logistic companies might have unique vehicles covering specific routes. Yet, it is also possible that this vehicle circulates through the same route every day, and thus, large amounts of data are produced. If the data collected by the logistic vehicles are adequately studied, important insights can be obtained, which could be used to draw insights from enterprise supply chains for strategic planning for instance.
Computational Intelligence, what some authors refer to computational intelligence as a branch of Artificial Intelligence, implements biologically inspired models algorithmically and can be highly beneficial. The main approaches that lead computer systems to simulate nature-inspired ways of reasoning include Neural Networks, Evolutionary Algorithms, Swarm Intelligence, Fuzzy Logic, and Reinforcement Learning. Other scholars, however, have defined computational intelligence as an interplay of the approaches mentioned above.
For Pincay-Nieves, using more nature-inspired methods in the solution of problems in our cities is a coherent way of addressing the issues of currently used approaches; with Computational Intelligence methods, it is possible to capture and process the uncertainty, incompleteness, and inaccuracy of geospatial data and thus, understand and improve how traffic behaves on the roads and streets.
From the extensive list of Smart Logistics problems that are yet to be solved, the last mile delivery is one of the most challenging ones. The last mile depicts the process of delivering a parcel from a local delivery point to the doorstep of the customers; it is basically the last stage of a long chain to provide delivery service to a customer.
It is a challenging problem to solve since it takes place in a dynamic environment such as the streets with traffic and circulation problems. Moreover, even when all the parts of a delivery chain might be optimized, the simple fact that a customer is not at home at the delivery time affects the overall service. At the end of the day, the last-mile problem is an optimization one, that seeks to deliver the highest number of parcels at the first try, without compromising the users’ privacy.
Therefore, if one thinks about solving this problem with sparse geospatial data, ‘fuzzy set theory’ is a great tool to find a solution. Fuzzy set theory emerged as a way of reasoning with imprecise predicates and finding appropriate representation of “blurry boundaries” between sets; they can be considered a generalization of traditional sets and thus, they are the base of fuzzy logic which facilitates the commonsense reasoning. With fuzzy logic it is possible to perform reasoning using linguistic variables, meaning that we can use partial information to reach a conclusion. For instance, concepts such as occasionally at home or regular customer can, with fuzzy logic, be properly represented and used to perform computations. Nevertheless, we cannot find optimal delivery routes only with fuzzy set theory, and therefore, a routing algorithm is to be implemented.
Swarm intelligence algorithms have also been around for quite some time. They precisely enable the finding of optimal routes. One of them that enables to address the issues of the last is the so-called Ant System Optimization, or ACO, algorithm. This algorithm is inspired by nature, in the way that ants can always find the best routes to their food. To do so, they leave a chemical pheromone trail when exploring a certain area. The next ants tend to choose the trails with the strongest pheromone concentration, leading them to find the shortest path from the source to the nest.
With fuzzy logic, it is possible to model or approximate how pheromones are perceived. For instance, certain paths could have a strong pheromone or a weak pheromone. Thus, the concept of pheromone can be fuzzified using linguistic summaries, to verbalize information by quantified sentences. This tries to emulate how we, as humans, process information day to day; for instance, we don’t need to know the exact outside temperature to define a warm or a cold day, we just do it using our perception. Through linguistic summarization, it is possible to extract abstract knowledge from numerical and categorical data in a straightforward manner. For instance, the linguistic summary most of the ants will follow a path with a strong pheromone, can be easier to understand than providing figures.
Linguistic summarizations ease the process of designing fuzzy rules and fuzzifying variables. For the case of finding the best route for a delivery team, besides the fuzzy pheromone intensity, we can define as well linguistic variables depicting the traffic congestion, the presence of the customers at home, and the distance between locations. As a result, the traditional ACO algorithm becomes the Fuzzy Ant Colony Optimization algorithm.
As a concrete example of a fuzzy rule when designing a delivery tour and deciding which location to visit next, if a path exhibits a strong pheromone, a certain zone has low traffic, the customer is most likely at home during the tour time, and the distance until the location is short, then this is a great location for visiting next.
At the Human-IST Institute, the team have conceptualized their own framework; the Fuzzy Ant Routing framework, or FAR framework, fuzzifies probe data coming from logistic vehicles to approximate traffic status, and the history of home deliveries to approximate the best time a customer is at home and, by deploying artificial ants capable of understanding what a strong or weak pheromone is, we could achieve the goal of finding optimal delivery routes that save resources while achieving a high first-try delivery ration.
With the implementation of the FAR framework, it is possible to show how the usage of uncertain, imprecise data to perform inferences is possible. This is to be highlighted given that, nowadays, more attention is given to precise models that require large amounts of precise data and consume high amounts of energy. Thus, with this research effort, it was attempted to show how applying computational intelligence techniques also enables practical solutions that might not be precise but good enough to solve complex problems.
Further details of the implementation of this project can be found in the book Smart Urban Logistics Improving Delivery Services by Computational Intelligence. The book is based on the doctoral dissertation of Pincay-Nieves which was awarded the Faculty Price 2022 for best PhD Thesis in Theoretical Sciences at the University of Fribourg.
That’s all for this episode, thanks for listening. Links to Pincay-Nieves’ original research can be found in the shownotes for this episode. There, you can also find links to the series of podcasts on ‘Fuzzy Logic’, the Human-IST Institute, and the FMsquare Foundation. And, as always, stay subscribed to ResearchPod for more of the latest science!
See you again soon