AI-driven Methods for Patrol Vehicle Dispatch and Route Optimization Under Uncertainty

A security company needs a method to dispatch patrol vehicles and plan vehicle routes to reduce costs, maximize revenue, and deliver to customers the best value. Vehicle dispatching and routing planning are two well-known combinatorial optimization problems (COPs) that find wide applications in transportation (e.g., truck routing and luggage handling), logistics (e.g., warehouse planning and goods delivery), communication (e.g., parcel delivery and networking), and many other industries. For example, the ability to solve these COPs can lower transportation costs and cut carbon emissions, which leads to lower costs of gas and groceries and a better environment. This proposal develops hyperheuristics that can configure algorithms and their parameters to solve the COPs by leveraging machine learning tools, metaheuristics, heuristics, and classical mathematical approaches. Different algorithm configurations and information flow paths will be explored, developed, and tested. To tackle uncertainties, a deep reinforcement learning model will be built between unexpected events and optimal solutions, trained with a large number of optimized solutions. This is to avoid re-optimization but leverage existing experiences and hidden knowledge in the optimal solutions. The algorithms will be tested and applied to a selected region (Victoria Branch of the partner company) with a software prototype.

Faculty Supervisor:

Gary Wang

Student:

Partner:

Paladin Security Group Ltd.

Discipline:

Engineering

Sector:

Administrative and support, waste management and remediation services

University:

Simon Fraser University

Program: