I ask for food delivery services everyday back in my country due to its fast speed and relatively cheap delivery fee. One strong competitor in this industry is Meituan. It’s a technical company using AI, machine learning etc to develop its delivery app. Though the technology is so developed, it still can’t expect the real issues the deliverymen meet. Deliveryman also becomes a high-risk occupation because the algorithm is continuously optimized to shorten the delivery time without accounting the unexpected issues like extralong waiting time during elevator peak in office building.

This articles aim to talk about the important areas and pain points briefly in food delivery platform. Since the algorithm part is complex, if you have interest in how they actually solve based on that, you can check the last resource link. Here is what Meituan mainly aim to solve:

1. Takeaway order allocation problem This kind of problem can generally be expressed as: there are a certain number of deliverymen, each already has a number of orders in the delivery process, a batch of new orders have been generated in the past period of time (such as 1 minute), and the rider’s driving speed, The driving distance between any two points, and delivery time of each order (the time required for the rider to deliver the order to the user after arriving at the user’s location), how to allocate this batch of new orders at the correct time?
2. Intelligent Regional Planning When a user orders food from Meituan, who is the rider serving him? How is it determined? These are determined by the delivery area boundaries. The delivery area boundary refers to the range corresponding to some merchant collections. In traditional logistics, the most critical point affecting the efficiency of terminal distribution is the familiarity of the distributor with the area he is responsible for. This is also one of the reasons why in the traditional logistics field, distribution stations or delivery personnel are fixedly responsible for certain communities. Because the more familiar they are, the higher the delivery efficiency will be. The real-time delivery scenario is also similar. Each rider needs to be familiar with a merchant or delivery area as fixedly as possible. At the same time, as far as management is concerned, the scope of delivery station management is relatively clear. In addition, if a new merchant comes online, it is also easy to determine which delivery station will provide the delivery service for it.
3. Smart Rider Scheduling This is a project derived from the longer and longer business hours of food delivery. The distribution team finally chooses the method of group scheduling, dividing all riders into several groups, and specifying the start time of each group. Then everyone can rotate in groups, and everyone will have a turn for each shift.
4. Rider Path Planning The situation is that there are many delivery tasks on a rider, and there are various constraints in these delivery tasks. How to choose the optimal delivery order of the tasks.

Resources link: https://tech.meituan.com/2020/02/20/meituan-delivery-operations-research.html