Research And Implementation Of Crowdsourced Intelligent Delivery System For Instant Logistics

Instant logistics is a critical branch of modern logistics,characterized by its rapidity and time-sensitivity.With the gradual decline of the demographic dividend,the development of intelligent distribution for crowdsourced instant logistics holds significant research value.Presently,the primary challenges facing crowdsourced ondemand platforms include an imbalance between supply and demand and a low retention rate of crowdsourced riders.There is an absence of a task scheduling mechanism that can simultaneously satisfy the expectations of customers,crowdsourced riders,platforms,and merchants.This dissertation applies algorithms and artificial intelligence techniques to the crowdsourced instant delivery industry,which to some extent ensures the operational requirements of multiple objectives such as system delivery efficiency,service quality,and the health of the platform ecosystem.The main content includes the following four aspects:(1)In the context of instant logistics dispatch scenarios,there is often a supplydemand imbalance between the continuous increase in order volume and limited delivery capacity.This dissertation proposes a bundling model based on the acceptance willingness of crowdsourced riders.Under the premise of satisfying a certain degree of order similarity,the model generates a collection of order bundles with the objective function of maximizing the probability of rider acceptance.This approach not only ensures delivery efficiency and service quality but also effectively resolves the issue of high refusal rates among crowdsourced riders.The effectiveness of the method is proven through experimental analysis.(2)In the instant logistics dispatch scenario,orders are time-sensitive,and the rider’s acceptance experience as well as delivery efficiency must be considered.It is a complex multi-objective optimization problem for order assignment.The balance of dispatching is also an important aspect affecting the experience of crowdsourced riders.Based on this,the dissertation establishes a bilateral stable matching between orders and riders and proposes a dispatching balance scheme that considers multiple objectives including delivery efficiency,service quality,and rider experience.The effectiveness of the method is demonstrated through comparative experimental results.(3)In the instant delivery scenario,riders often deliver multiple orders at once.Various contextual information such as required delivery times,additional distances,merchant preparation times,and the rider’s prior knowledge can affect their choice of delivery routes.Consequently,this dissertation proposes a multitask deep learning network architecture based on the attention mechanism to predict delivery routes and durations.The core idea is to simulate the decision-making behavior of riders by predicting the order and the time consumed at each delivery node.Analysis of the experimental results indicates a significant improvement in the adoption rate of rider routes.(4)Lastly,this dissertation conducts an overall design of the instant logistics delivery system based on the concept of distributed service architecture.It covers aspects including demand analysis,core business process analysis,overall system design,and implementation of core module functions,while also providing implementation schemes for key front-end and back-end functionalities and the system operation interface.In summary,this dissertation investigates the platform dispatch methods in the context of crowdsourced instant logistics delivery,covering three core issues: order bundling strategy,dispatch strategy,and prediction of delivery routes and durations.It also integrates these research findings into system design,achieving a degree of intelligentization in the crowdsourced instant delivery system.