Project overview
Recent years have seen a growth in ecommerce and home delivery purchase options, partly due to fast delivery options (eg. same-day delivery) becoming increasingly available [1]. This is a large market in the UK. In the financial year 2022-23, parcel volume was 3.6 billion items, compared to 2.8 billion in 2019-20 [2]. In turn, this has increased the requirement for last-mile delivery operations, that is, the last leg of the journey products take before reaching the consumer’s home. In cities, a large proportion of these deliveries are performed by vans, which contribute significantly to traffic and its associated externalities such as congestion, air and noise pollution.
As a response to this, in recent years there has been a drive to explore more sustainable and green delivery options. These include the usage of low emission vehicles (eg. electric cars/vans) or no emission alternatives, such as delivery via porters (on foot) or riders (by bike), which typically operate from microconsolidation depots, small spaces in cities designed to store local delivery parcels [8]. These and other initiatives, such as carrier collaboration, bring new planning challenges. These include optimisation with multi-mode delivery platforms [3], coordination between carriers or sharing of collaboration benefits, which have been tackled by operational research methods such as combinatorial optimisation or game theory [4, 5]. These methods have typically translated and adapted traditional vehicle routing research to these new environments. In practice, however, the business models that sustain riders and porters tend to be grounded on the gig economy, where porters will be registered with multiple logistics platforms. During their workday, they will be presented with multiple work options and choose (or not) to perform some of them. Their choice will be driven by their preferences, and be impacted by the differences the routes have in terms of route aspects (reward it offers, length, weight it carries, locations it visits) and their own situation, typically unknown to the planner (where they live, what vehicle they have, what other activities they will be doing during the day, safety, etc.) [6, 7]. From the planner’s perspective, this carries a risk, as they might create unattractive routes for porters/riders which either require them to offer a high reward to be delivered or will not be undertaken by any worker.
To deal with this challenge, a growing body of literature in the mathematical sciences demonstrates that incorporation of discrete choice models into the operational management problem can mitigate this issue [10]. These models can exploit experimental data to understand individual choices (in this case whether to accept work or not) and construct porter/rider profiles which can inform optimisation models for planning. For large scale operations, experiments can be run by intelligent agents following principles of discrete choice experiments and dynamic pricing schemes. This project aims to perform foundational work to combine operational decisions related to vehicle routing with discrete choice models to develop intelligent agents capable of planning delivery work, making use of alternative delivery models, and doing so effectively in terms of sustainability, cost, and fairness and satisfaction of delivery workers.
As a response to this, in recent years there has been a drive to explore more sustainable and green delivery options. These include the usage of low emission vehicles (eg. electric cars/vans) or no emission alternatives, such as delivery via porters (on foot) or riders (by bike), which typically operate from microconsolidation depots, small spaces in cities designed to store local delivery parcels [8]. These and other initiatives, such as carrier collaboration, bring new planning challenges. These include optimisation with multi-mode delivery platforms [3], coordination between carriers or sharing of collaboration benefits, which have been tackled by operational research methods such as combinatorial optimisation or game theory [4, 5]. These methods have typically translated and adapted traditional vehicle routing research to these new environments. In practice, however, the business models that sustain riders and porters tend to be grounded on the gig economy, where porters will be registered with multiple logistics platforms. During their workday, they will be presented with multiple work options and choose (or not) to perform some of them. Their choice will be driven by their preferences, and be impacted by the differences the routes have in terms of route aspects (reward it offers, length, weight it carries, locations it visits) and their own situation, typically unknown to the planner (where they live, what vehicle they have, what other activities they will be doing during the day, safety, etc.) [6, 7]. From the planner’s perspective, this carries a risk, as they might create unattractive routes for porters/riders which either require them to offer a high reward to be delivered or will not be undertaken by any worker.
To deal with this challenge, a growing body of literature in the mathematical sciences demonstrates that incorporation of discrete choice models into the operational management problem can mitigate this issue [10]. These models can exploit experimental data to understand individual choices (in this case whether to accept work or not) and construct porter/rider profiles which can inform optimisation models for planning. For large scale operations, experiments can be run by intelligent agents following principles of discrete choice experiments and dynamic pricing schemes. This project aims to perform foundational work to combine operational decisions related to vehicle routing with discrete choice models to develop intelligent agents capable of planning delivery work, making use of alternative delivery models, and doing so effectively in terms of sustainability, cost, and fairness and satisfaction of delivery workers.
