Machine Learning Delivery Cluster Assignment Balancing Driver Workload

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for assigning packages to delivery persons struggle to balance the number of delivery destinations among drivers and consider travel routes effectively, making it difficult to improve the assignment process.

Innovation Solution

A delivery management system that uses machine learning to classify delivery destinations into clusters based on past data, including position information, to optimize package assignment by training a model that predicts delivery clusters and adjusts assignments to balance the workload and consider travel efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of manufacture

If a general clustering method is used for assigning packages to delivery persons, then the assignment process is simple, but the assignment quality is poor and cannot adapt to detailed requirements such as reducing imbalance among delivery persons and considering travel routes

Engineering Contradiction:
Improveease of assignment processVSAvoidassignment quality
Core Design Contradiction:
Ease of manufactureVSManufacturing precision

Solution Approach 1:

The patent transforms the assignment problem from a general clustering task to a supervised learning task by introducing labeled training data with ground truth delivery cluster assignments. The model is trained using parameters derived from actual delivery data, including delivery person assignments, travel routes, and delivery destination characteristics, enabling it to learn optimal assignment patterns that generalize to new delivery scenarios.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The system incorporates feedback loops where delivery performance data is continuously collected and used to retrain the machine learning model. The training data generation means creates training data from actual delivery results, and the training means updates the model parameters based on this feedback, allowing the assignment system to progressively improve its accuracy and adapt to changing delivery requirements.

Inventive Principle:
Principle #23Feedback

2Productivity

If packages are assigned to delivery persons and then delivery routes are determined, then the delivery process is systematic, but it is difficult to achieve balanced assignment and optimal routes simultaneously

Engineering Contradiction:
Improvedelivery efficiencyVSAvoidassignment and routing complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The system performs preliminary actions by pre-training the machine learning model using historical delivery data before actual package assignment occurs. The training data generation means prepares training data from past delivery records, and the training means pre-computes optimal assignment patterns and route characteristics. This preliminary training enables the system to quickly and accurately assign packages and determine routes in real-time without complex computational overhead during actual delivery operations.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP4141762A1Delivery management system, delivery management method, and program
Publication Date: 2023.03.01 RAKUTEN GROUP INC
  • EP4141762A1 patent drawingFigure 1
  • EP4141762A1 patent drawingFigure 2
  • EP4141762A1 patent drawingFigure 3

AI summary

Provided is a delivery management system configured to: generate, through clustering processing, training data including: delivery destination information on a plurality of delivery destinations, the delivery destination information including information indicating positions of past delivery destinations; and ground truth data indicating delivery clusters into each of which a corresponding one of the plurality of delivery destinations is classified, and each of which is associated with a delivery person; and input, to a machine learning model, the delivery destination information on the plurality of delivery destinations included in the training data, and train the machine learning model based on information indicating predicted delivery clusters, which is output from the machine learning model, and based on the delivery clusters into which the delivery destinations are classified, and which are included in the training data.