Artificial Intelligence-Assisted Decision Making in Dynamic Vehicle Routing Problems

Brief Description:

One of the important factors for efficient supply chain management is to optimise the supply chain's physical flow, improve customer satisfaction, and reduce overall costs in the supply chain. An integrated and resilient approach is required for the decision-makers to forecast demands in different periods and route the vehicles to meet those demands. Although some studies are available in the literature for demand forecasting and vehicle routing, none of them studied the problems in an integrated manner. Thus, there is a gap between the theory for supply chain management and actual practice. After investigating limiting assumptions of previous studies of demand forecasting and vehicle routing problem to address the practical challenges and academic research gap, it is found that the demand forecasts from the artificial intelligence-based machine learning approach could be integrated with distribution decisions to create optimal value and thus increase customer satisfaction. Hence the study can be aided by the answers identified in this project to five research questions:

1. Why it is necessary to integrate demand forecasting and vehicle routing in the supply chain for on-time delivery?
2. How does a machine learning algorithm perform the demand prediction more precisely and scientifically?
3. What is the minimum required percentage of demand forecast accuracy for making informed decisions?
4. How do some critical practical factors (e.g., seasonality, product type or categories, geography, time frame, accuracy, vehicle numbers) influence forecast demands and route the vehicles to meet those demands simultaneously?
5. How to develop an integrated model considering both demand forecasting and vehicle routing scheduling?

Advanced machine learning methods can improve forecasting performance, result in lower costs due to decreased inventory and higher customer satisfaction, and improve on-time deliveries. The aim of this research is to resolve these issues by developing a model that serves as an understanding of the problem and then solving it by considering both solution performance and computational time. Outcomes of this project will develop models considering the problem characteristics and machine learning and meta-heuristic algorithms based integrated approach to ensure an improvement in operational performances of the supply chain.
 

Research Aim:

• To predict potential demand ahead of time to minimise the manufacturer's lead time. 
• Investigate the utility of advanced machine learning techniques in forecasting manufacturer’s distorted demand. 
• Identify and analyse the trade-off in forecast demand accuracy and shipment punctuality.
• Consider parameters that are uncertain in real-world practical scenarios. 
• Development of an integrating mathematical formulation of demand forecasting and vehicle routing to meet those demands on time in the supply chain through different optimised techniques considering the situational resource constraint. 
• Compare different demand prediction network optimisation techniques for solving synchronised objectives in the supply chain. 

Project start date: February 2022.

Expected completion date: July 2025.