In our daily lives, we benefit from the application of Optimization theories and algorithms. They are used, for example, by IoT devices, GPS systems, by shipping companies delivering packages to our homes, by financial companies, airline reservations systems, etc. Optimization is a discipline that solves a great variety of real-world applied problems in diverse areas: transportation, supply chain, manufacturing, finance, government, economics, etc.
Our research group consists of academic staffs in the School of Information and Technology (SoICT) who are actively involved in teaching, supervising and researching across a wide range of topics and areas in Optimization. Our research focuses on the development, analysis and implementation of advanced theories and algorithms to provide high value solutions to complex real-world problems and challenges. We also have research collaboration with other optimization groups of Nanyang Technological University, Singapore, and University of Technology Sydney, Australia. See the slides here for more detail.
Contact: Assoc. Prof. Huynh Thi Thanh Binh, Email: firstname.lastname@example.org
- Logistics and transportation optimization: Transportation is a very relevant sector for contemporary society, both for companies and individuals. Every day, thousands types of commodities such as fresh food, frozen or dairy products, small to large packages, etc. are shipped between locations within or outside cities.Thanks to the development of digitization and automation technologies, in the past few years, we have seen an increase in the number of logistics companies. Their main goal is to transport different kinds of goods domestically and internationally based on customer needs. They thus all share the same challenges: how to manage limited fleets of vehicles, handle warehouses, mange inventory, etc. to make profit as much as possible, while enhancing customer satisfaction, improving working conditions for drivers, reduce carbon footprint, etc. Given these presence of multiple business constraints, our research is to analyze, design and develop planning and optimization methods to create better decision support to companies within the supply chains, transport sector.
- Multitask learning, transfer learning: Inspired by the human solving ability that routinely uses a pool of knowledge drawn from past experiences whenever faced with a new task, Transfer Learning and Multitask Learning have gained much attention within the Artificial Intelligence community. Besides, real-world problems seldom exist in isolation. For example, many routing problems are repetitive or and network designs share common characteristics or support each other. Knowledge from solving one problem may help solve other problems more efficiently. Transfer Learning aims at solving problems that occur sequentially, and the knowledge obtained when tackling preceding tasks is employed as external information when dealing with new problems/instances. In contrast, Multitasking Learning tackles multiple different tasks simultaneously by dynamically online exploiting synergies existing among them. Viewed from different angles, Transfer Learning and Multitask Learning can be applied to various optimization fields, ranging from theory to practice. Our work focuses on designing novel multitasking algorithms that can simultaneously solve a massive number of tasks. Besides, we will investigate these approaches to solve real-world optimization problems.
- Multi-domain network design optimization: A multi-domain network consists of multiple domains, in which each domain is administrated as a unit with the same rules and procedures. Routing is the fundamental problem showing the path of resources in the network. In a communication network, efficient routing provides more control to the operators and enables the delivery of services with quality of service across domain boundaries. In Military communication, various elements of military forces, such as the army, navy, air forces, and special units, cooperate to achieve speciﬁc tactical goals. Each one of these organizations has its network structure in one or more domains. Besides, freight routing planning aims at assigning optimal routes to move commodities from their origins to the respective destinations through the transportation networks. Although it is a short-term decision making in the transportation network design, freight routing planning is oriented directly on satisfying the customers’ demand, and its performance determines the competitiveness of a transportation carrier or a third-party logistics company in the freight market.
- Charging schedule optimization in Wireless Rechargeable Sensor Networks: Sensor nodes in conventional Wireless Sensor Networks are often powered by batteries, thus, they can only operate for a limited period of time depending on battery capacities. When some sensors deplete their energy, the network would become fragmented and the data from some parts of the sensing field can no longer be extracted. Wireless energy charging was proposed as a promising technique to address the energy provisioning problem. However, this technology also brings new challenges, including charging scheduling and energy forecasting for constructing an effective charging schedule. Solving these problems is of great significance, as the first step to exploiting the wireless charging technology and providing sustainable power for sensor networks. It can be applied in the harshness of environments, such as earthquakes, soil monitoring, large scale wireless sensor networks.
- Path planning with energy optimization for mobile robots: Mobile robots have become more commonplace in commercial and industrial settings. Hospitals have been using autonomous mobile robots to transfer materials. Warehouses have installed mobile robotic systems to efficiently move materials from stocking shelves to order fulfillment zones. Mobile robots are also found in industrial, military and security settings. To power a mobile robot usually use batteries. Battery power is limited. In order for the robot to work effectively, it is necessary to schedule the optimal robot path to optimize energy; need to predict exhausted energy to plan charging in time.
- Optimization techniques for Credit scoring problem: Credit scores are the most commonly used tool by financial institutions for determining consumer credit risk. In the last few decades, quantitative methods known as credit scoring models have been developed for the credit granting decision. The objective of quantitative credit scoring models is to assign credit applicants to one of two groups: a “good credit” group that is likely to repay the financial obligation, or a “bad credit” group that should be denied credit because of a high likelihood of defaulting on the financial obligation. However, quantitative methods have some limitations, such as they can not consider additionally constraints defined by risk management experts, they base completely on training dataset. This leads to our research direction, how to apply optimization techniques to build credit scoring models (non-parametric approaches), which not only are more accurate, but also can consider side constraints defined by users. Potential optimization techniques in this research are both complete search approaches (Integer Programming, Constraint Programming) and incomplete search approaches (Local Search, Meta-Heursitic).
- Resources management in the cloud-fog environment optimization: In recent years, the Internet of Things (IoT) has been one of the most popular technologies that facilitate new interactions among things and humans to enhance the quality of life. With the rapid development of IoT applications, fog computing is an emerging distributed computing paradigm that has recently attracted the attention of both the industry and academic community for guaranteeing the requests of computational applications in IoT smart devices. In the fog environment, IoT applications are executed by the intermediate computing nodes in the fog, as well as the physical servers in cloud data centers. Fog computing contributes to processing large amounts of data generated in smart transportation, smart grid, smart health, smart home, and smart home, and other latency-sensitive applications. On the other hand, due to the resource limitations, resource heterogeneity, dynamic nature, and unpredictability of the fog environment, it necessitates resource management issues as one of the challenging problems to be considered in the fog landscape. To solve the resource management challenge, we adopt several promising approaches such as heuristics, deep learning, reinforcement learning, …
- Optimization in the transportation, drone problems: Research optimization algorithms in the Trucks and Drones delivery problem, Door to Door sampling services, Multi-echelon distribution system in city logistics; Transportation Logistics Networks; minimum routing cost problems in the multi-domain networks;
- Multi-objective optimization, multi-tasking evolutionary algorithms: Research and application of multi-objective, multi-tasking evolutionary algorithms in solving interdisciplinary optimization problems.
- Multi-domain network design optimization: Develop multitasking evolutionary and heuristic algorithms to find the shortest path with uniqueness constraints and efficient network structures in the multi-domain network.
- Charging schedule optimization:study wireless charging models to prolong the life of the network; Propose algorithms to optimize the cycle of the charging robot, optimize the charging stop time; energy prediction to plan charging on time.
- Path planning with energy optimization for mobile robots: Propose optimization algorithms to find shortest path, minimizing energy for mobile robots; predict energy to plan charging in time.
- Optimization techniques for Credit scoring problem: Study optimization techniques – complete search approaches (Integer Programming, Constraint Programming) and incomplete search approaches (Local Search, Meta-Heursitic) to solve credit score problem.
- Reinforcement Learning for Combinatorial Optimization: Design Reinforcement Learning frameworks to solve complex combinatorial optimization instead of using traditional optimization algorithms.
- Resources management in the cloud-fog environment optimization: Propose heuristics, deep learning, reinforcement learning to solve resources management in fog computing.