Research directions are the general, abstract topics on which we work. We conduct concrete research projects under these directions.
Logistics and transport are among the most important services for any industry or society. Without them, the economy would simply break down or fall back to pre-industrial levels. However, they also turn oil (which is getting less) into pollution. Using optimization algorithms for logistic planning means to find ways to transport goods or people in an efficient way. Depending on the problem, efficient could mean to travel short distances, use few vehicles and less manpower, or to be otherwise cheap. This often equates to being environmentally friendlier. Research on this domain therefore is highly relevant and we want to contribute to it. But logistics is not limited to vehicles driving outside. It may also concern tasks as diverse as the routing of vehicles inside of automated warehouses or factory floors, the routing of work pieces on assembly lines, and even finding the optimal sequence to drill a given set of holes into and place a certain set of components onto a circuit board.
This research direction can be considered as a complement of our work on operations research, together with which we aim to provide a set of tools of optimal control and management of highly-automated manufacturing and delivery processes for a modern industry.
Many decisions to be made in management or planning of an enterprise are optimization problems. The field of research concerned with solving them is called operations research [運籌學]. We are concerned with questions such as
- How can we divide blocks of raw material into parts of exactly the right sizes needed for the work pieces the customer wants, with minimum wastage?
- If a company has already several dependencies in China and wants to open a new office, in which city should that be located to maximize customer coverage and company profit?
- How should incoming customer orders be assigned to machines for manufacturing so that all orders are completed on time and as fast as possible (also considering that an order may need to be processed by multiple machines in multiple steps)?
These few examples show that the direction borders directly to our logistics direction and topics relevant to the topic of Industry 4.0. Our institute can provide a framework for the optimized operation in a highly automated manufacturing environment.
Optimization is a technology which has the potential to become even more important than data mining and big data are now, as it can improve the efficiency of each and every aspect of an enterprises operation and products. Optimization means to find approximate solutions for hard problems. Optimization algorithms can find short routes in logistic planning scenarios, construction plans for work pieces that require only little amounts of material, or efficient schedules in production planning. They can help to reduce costs and pollution at the same time.
The term "optimization algorithms" implies a plural, and there are already many different families of algorithms, including exact methods, heuristics, metaheuristics sub-families such as evolutionary computation and swarm intelligence, as well as local search and hybrid algorithms. Each of these families contains dozens of general and specialized algorithms. We want to know how we can find out which method is best for which problem and when and why. First, using the best algorithm will give us the best solutions, which is what we want. Knowing why an algorithm is best could help us to create even better ones. Thus, benchmarking of optimization algorithms in a statistically sound, robust, and ideally automatic way seems to be an important topic to me. Yet it is entirely under-represented in literature. We want to emphasize how important it is, as for practical applications of optimization, we simply need to know that we are using the best and most reliable methods.
Distributed computing is the huge field of applications that involve multiple computers connected over a network. This includes cloud computing, service-oriented architectures (SOAs, sensor networks, the internet of things, and the internet in general. It is thus an area of computer science which touches almost every aspect of our daily life, ranging from providing the news that we consume, the apps we use for ordering products, to the infrastructure that powers [[enterprise application environments and even the sensors, networks, and controllers enabling the Industry 4.0, i.e., the next evolutionary step of our industry. Now this, at first glance, has nothing to do with optimization. But only at first glance, because we do not just want to build such systems, we want to build them to be efficient, robust, and cheap both in construction and maintenance.
Thus, for every of the above-mentioned topics, various questions arise, such as: How can networks be designed to be as robust and fast but also as cheap as possible? Where should we put sensors for a reliable, complete surveillance at a low cost? How should we best assign resources in a cloud? How can we design robust and self-healing and automatically (optimally) configuring SOAs?