Current PhD projects
You may search for more projects and further descriptions in DTU's research database ORBIT - Online Research Database In Technology
Morten Kjærgaard
Modular platform for commercial mobile robots
Developing a robot application for sale on the commercial market creates a lot of new challenges. Such a robot is a very complex system, and yet it must be cost-efficient to develop and produce, and reliable and safe enough for people to actually use. Mobile robots already exist as successful commercial products, like lawnmowers, vacuum cleaners and for transportation. But the potential is much higher. The project is performed in cooperation with Prevas A/S.
By using higher level models to describe the robot system, the project strives to improve the ability to reuse functionality across different robot systems and avoid having to invent everything from scratch every time. Additionally, the project focuses on developing methods for automatic calibration, optimization and supervision of the parameters for the robot to improve reliability and safety.
The vision of the project is to close the gap between what is possible to do with robots in a research lab, and what a company like Prevas A/S can develop for the commercial market.
To be completed: 2013
Project co-founded by Prevas A/S
Enis Bayramoğlu
Robust and Reliable Robot Operation in Semi-Structured Environments
The field of robotics is currently very mature for operation in structured environments, such as an assembly line. However, a much wider range of applications require the ability to cope with semi-structured environments, where not everything is controllable or known. This occurs when there are living beings in the environment, or the environment itself is too large or complex to model accurately.
This project aims at using advanced sensors with recent artificial intelligence methods to enable robust and reliable robot operation in semi-structured environments. Robot localization, recognition of task-related objects and environment mapping, main challenges of this field, are in focus of this project. Robot vision is central to the approach due to its large information content and relatively lower cost, but other sensors will also be used to support it.
The results are expected to benefit the society through enabling the use of robots for many tasks that require human labour today.
To be completed: 2012.
Anders Billesø Beck
Situation Assessment for Mobile Robots
In scientific and popular literature, the emergence of intelligent service robots outside of laboratories has been predicted repeatedly for the last 10 years. Yet we still do not see intelligent service robots in our homes, in our workplace or at the shopping malls. The reason is that reliable operation in dynamic and indeterministic environments is a huge challenge, not only for the navigation systems but also for handling unplanned and even unexpected events.
In cooperation with the Danish Technological Institute, the project aims at developing methods and techniques for situation assessment for mobile robot systems in general. By understanding the situation of a robot, unexpected errors can be resolved or even avoided. Based on signal processing, sensor fusion and artificial intelligence (AI), the project will investigate how a general situation modelling framework and architecture can be developed and applied.
Commercial success of service robots relies on handling of errors in a new general sense. This project will contribute with methodology to make a robotic future one step closer.
To be completed in 2012.
Martin Choux
Robust and fault-tolerant control for nonlinear hydraulic-mechanical system
This project examines the combination of fault diagnosis with robust control for nonlinear hydraulic-mechanical (NHM) systems. Focus is on fault-tolerant control for NHM systems with applications focused on offshore equipment. The research is also relevant to machining plants, in robotics and for fatigue testing systems. Another objective is to integrate the control issues in the design of a hydraulic-mechanical system. New ways of selecting hydraulic components will be investigated.
The project will investigate the usefulness of control methods in this specialised application, including passivity techniques in nonlinear control, adaptive back stepping, H-infinity or mixed H_2 /H-infinity controllers and fault-tolerant control. The interaction between controls, hydraulic system design and optimisation techniques will also be considered. The project aims at advancing the theory in nonlinear and fault-tolerant control, as well as validating the theory with experiments.
The results are anticipated to have a large impact on control performance and the reliability of hydraulic-mechanical systems.
To be completed: 2011
Torben Green
Performance assessment and system monitoring for refrigeration systems
Supermarket refrigeration plants consume a large amount of energy. Consequently, it is beneficial to operate the systems in the best possible way. However, optimal operation of a refrigeration plant is also dependent on other costs such as the replacement of components due to excessive wear. Thus, assessing the performance of the refrigeration plant is important to ensure optimal operation with respect to the total cost of ownership.
This project examines the possibility of establishing a performance measure for refrigeration systems including the control system. The development of algorithms to actively assess the performance of a distributed control system will be part of the project. In addition, automatic accommodation for the cause of the performance degradation of a control system by means of online retuning and/or redesign will be addressed.
The project focuses on the development of methods and tools for active performance assessment of refrigeration systems and algorithms for accommodation of performance degradation.
To be completed: 2011
Søren Hansen
Fault Diagnosis and Fault Handling for Autonomous Airplanes
Small unmanned aerial vehicles have a great potential in many different applications. However the relative simple components and control methods often used on this type of aircraft can be the cause of dangerous crashes. Safety and reliability considerations used on piloted airplanes cannot directly be used on smaller unmanned airplanes because of weight and space limitations, and therefore other methods must be considered.
This project investigates diagnosis of sensor and actuator faults for unmanned aerial vehicles. The goal is to create a system that is able to diagnose faults by analysing telemetry data while in the air. The airplane operator is notified when a fault is discovered, and depending on the severity of the fault, appropriate actions are taken autonomously or suggested to the operator to bring down the plane in a safe manner and avoid hazardous situations.
Adding diagnosis systems to small unmanned aerial vehicles will make them able to complete their mission in a safer and more efficient way and thereby enhancing their usability.
To be completed: 2012
Ragnar Ingi Jónsson
Fault Diagnosis for Identification of Deviant Behaviour in Dairy Cows
Automatic detection of deviant behaviour amongst dairy cows is a task of growing interest in modern farming. A main focus in this perspective is to detect deviant behaviour caused by oestrus or lameness. Early detection of a cow in oestrus or a cow suffering from lameness can save the farmer from an economic loss and the animal from a prolonged period of pain/illness.
The main scope of this research is to study mathematical modelling of behaviour, extend methods from fault diagnosis and change detection to biometrical models for animal behaviour, and obtain a detection model. Methods of statistical change detection and automata-based detection theory are investigated,
i.e. methods developed for fault diagnosis in industry but which can be modified to cope with the complexity found with live animals.
This project results demonstrate how theory from technical sciences can be used within the much broader scope of biological detection problems and help aiding farmers’ decision making and
Ensure the wellbeing of cows.
To be completed: 2010
Completed PhD projects
Martin Birkelund Larsen
Nonlinear methods for spacecraft orbit control using an electrodynamic tether
An electrodynamic tether is a conducting cable used to induce a force on a spacecraft. The method utilises the Lorentz force, originating from the interaction between a current in the tether and the magnetic field of the Earth, to change the orbit of the spacecraft. The method is particularly interesting in connection with the de-orbiting of obsolete satellites, but it can be used in general to perform corrections of orbit drift.
The objective of the project is to develop control strategies ensuring a stable motion of the satellite-tether system during orbit manoeuvres. The nonlinear dynamics of the system and the time-periodic variation of the magnetic field during an orbit complicate the control problem and prevent the use of classical control methods, so novel control theory concepts are required.
This control approach permits orbit manoeuvring of a spacecraft independent of traditional propellant, but only by using electrical power obtained from its solar panels.
Completed: 2010
Morten Rufus Blas
Fault-tolerant navigation using vision and computer-based learning
Automatic navigation systems are already revolutionising farming by cutting costs and increasing output. Vision is an enabling technology, as it allows increased automation through new applications. However, computer-vision systems regularly produce artefacts in the image processing that could seriously disturb the safe function if used in closed loop steering of an automated tractor.
The project aimed at using vision in new ways by applying learning-based algorithms to the problem of navigation. Also, it contributes to increase the tolerance to faults by fusing the vision with multiple sensors such as odometry and GPS and by integrating fault-tolerance in the vision algorithms themselves.
The project has provided knowledge and experience concerning structured approaches to the design of fault-tolerant vision systems. Further, the project has shown methods to provide better reliability of vision systems for outdoor use. It will result in a significantly better product with regard to availability and safety, and results will be applied in products in the near future.
Completed: December 2009
Roberto Galeazzi
Autonomous Supervision and Control of Parametric Roll Resonance
Parametric roll is a resonance motion where large amplitude oscillations may develop in ships when the roll restoring moment varies in proportion with half of the roll eigenfrequency. Modern container ships are particularly prone to this due to their hull shape. The challenge is whether detection and stabilisation can be achieved in time to avoid damage.
In the project, stabilisation of parametric roll has been achieved directly, through increase of roll damping, or indirectly, through change in wave encounter frequency. Two control strategies - applying nonlinear Lyapunov-based methods - were combined to stabilise parametric roll within few roll cycles. Two novel detectors have been proposed, which work within a short-time prediction horizon, and issue early warnings of parametric roll inception within few roll cycles from its onset. The detectors have been validated against experimental data, and they have shown excellent performance and they have shown excellent performance in detection of the onset of parametric resonance.
The results are expected to be commercially interesting as ship-owners could avoid multi-million dollar damages every year if given the ability to avoid the roll resonance phenomenon. A proof-of-concept study has been started as a continuation of this research.
Completed: October 2009
Ole Fink Hansen
Maintenance free and sustainable high-level control in cement and mining industry
The use of high-level control systems in cement plants is well established. It is acknowledged that they can improve performance by increasing production and reducing the environmental impact of the plant. In spite of this, many systems are taken out of use after a short period due to changes in the plant which are not handled by the control system. Therefore, the systems need regular adjustment by expert personnel to maintain a high performance.
Much research has focused on making the control algorithms top-performing and ever more tolerant to such changes. Modern control algorithms have therefore become sophisticated and complex pieces of software which are very difficult to maintain. During this project, focus has turned onto the software aspects of control algorithms, and methods have been developed to monitor and diagnose the control algorithm rather than the plant.
In this way, an early warning is possible avoiding critical issues. Furthermore, man-hours for maintenance can be reduced by automatically locating the root cause of a problem.
Completed 2008, defended 2009