This paper treats L 1 adaptive hovering control of an unmanned surface vehicle in a station-keepi... more This paper treats L 1 adaptive hovering control of an unmanned surface vehicle in a station-keeping mode where a region of zero control authority and under-actuation are main challenges. Low-speed and reversing dynamics are identified from full scale sea trials, and parameter uncertainty is estimated. With significant parameter variation, an L 1 adaptive controller is employed for heading control. The L 1 family of controllers allows for several topologies and an architecture is suggested that suits heading control of a vessel, the requirements of which differ from that of previous L 1 literature. The control design is tackled directly in discrete time to allow a fast embedded implementation in the vehicle. Analysis of robustness, tracking performance and wave disturbance response are detailed in the paper.
Gas bearings are popular for their high speed capabilities, low friction and clean operation, but... more Gas bearings are popular for their high speed capabilities, low friction and clean operation, but su er from poor damping, which poses challenges for safe operation in presence of disturbances. Enhanced damping can be achieved through active lubrication techniques using feedback control laws. Such control design requires models with low complexity, able to describe the dominant dynamics from actuator input to sensor output over the relevant range of operation. The mathematical models based on first principles are not easy to obtain, and in many cases, they cannot be directly used for control design due to their complexity and parameter uncertainties. As an alternative, this paper presents an experimental technique for ”in situ” identification of low complexity models of the entire rotor-bearingactuator system. Using grey-box identification techniques, the approach is shown to be easily applied to industrial rotating machinery with gas bearings and to allow for subsequent control design. The paper shows how piezoelectric actuators in a gas bearing are eciently used to perturb the gas film for identification over relevant ranges of rotational speed and gas injection pressure. Parameter-varying linear models are found to capture the dominant dynamics of the system over the range of operation. Based on the identified models, decentralised proportional control is designed and is shown to obtain the required damping in theory as well as in a laboratory test rig.
An H-inf robust control of wind turbine inverters employing an LCL filter is proposed in this pap... more An H-inf robust control of wind turbine inverters employing an LCL filter is proposed in this paper. The controller dynamics are designed for selective harmonic filtering in an offshore transmission network subject to parameter perturbations. Parameter uncertainty in the network originates from the grid and the number of wind turbines connected. Power converter based turbines inject harmonic currents, which are attenuated by passive filters. A robust high order active filter controller is proposed to complement the passive filtering. The H1 design of the control loop enables desired tracking with integral effect while bounding the induced change. The design was tested in an aggregated model of the London Array offshore wind power plant and compared with traditional PI controller designs. Robust stability and performance and a reduction of control effort by 25% are obtained over the full envelope of operation.
Autonomous robots require many types of information to obtain intelligent and safe behaviours. Fo... more Autonomous robots require many types of information to obtain intelligent and safe behaviours. For outdoor operations, semantic mapping is essential and this paper proposes a stochastic automaton to localise the robot within the semantic map. For correct modelling and classification under uncertainty, this paper suggests quantising robotic perceptual features, according to a probabilistic description, and then optimising the quantisation. The
IEEE Transactions on Aerospace and Electronic Systems
Airspeed sensor faults are common causes for un-manned aerial vehicle incidents. Pitot tube clogg... more Airspeed sensor faults are common causes for un-manned aerial vehicle incidents. Pitot tube clogging or icing are common causes. Timely diagnosis of such faults or artifacts could potentially prevent crashes. This paper employs parameter adaptive estimators to provide analytical redundancies and a dedicated diagnosis scheme is designed. Robustness is investigated on sets of flight data to estimate distributions of test statistics. The result is robust diagnosis with adequate balance between false alarm rate and sensitivity.
Fault prognosis,diagnosis and fault tolerant control play important roles in continued developmen... more Fault prognosis,diagnosis and fault tolerant control play important roles in continued development of different kinds of industrial systems. In some industries,fault detection,isolation and fault tolerant control have already been well established,e.g. fault diagnosis is a standard component in automotive engines for reasons of pollution control and efficiency, in the aerospace industry on-board diagnostic algorithms can reduce maximal structure stress and reduce weight,and in other industries, fault prognosis, diagnosis and fault accommodation are technologies that gradually enter in to the new design process. This field is also very active in the academic community, where new methods are being developed to meet the challenging industrial needs. The purpose of this special issue is to gain experience in the field from both academic and industrial perspectives. Submissions to this special issue should include theory and/or methods that directly target real world problems and cases o...
This paper describes an obstacle detection system for a high-speed and agile unmanned surface veh... more This paper describes an obstacle detection system for a high-speed and agile unmanned surface vehicle (USV), running at speeds up to 30 m/s. The aim is a real-time and high performance obstacle detection system using both radar and vision technologies to detect obstacles within a range of 175 m. A computer vision horizon detector enables a highly accurate attitude estimation despite large and sudden vehicle accelerations. This further facilitates the reduction of sea clutter by utilising a attitude based statistical measure. Full scale sea trials show a significant increase in obstacle tracking performance using sensor fusion of radar and computer vision.
Pack-off is a partially or complete blocking of the circulation flow in oil and gas drilling, whi... more Pack-off is a partially or complete blocking of the circulation flow in oil and gas drilling, which can lead to costly delays. Early detection and localization of a pack-off is crucial in order to take necessary actions avoiding downtime. This incident will affect physical friction parameters in the well. A model-based adaptive observer is used to estimate these friction parameters as well as flow rates. Detecting changes to these estimates can then be used for pack-off diagnosis, which due to measurement noise is done using statistical change detection. Isolation of incident type and location is done using a multivariate generalized likelihood ratio test, determining the change direction of the estimated mean values. The method is tested on simulated data from the commercial high-fidelity multi-phase simulator OLGA, where three different pack-offs at different locations and with different magnitudes are successfully detected at an early stage and with low false alarms.
This paper investigates a minimalistic laser-based range sensor, used for underwater inspection b... more This paper investigates a minimalistic laser-based range sensor, used for underwater inspection by Autonomous Underwater Vehicles (AUV). This range detection system system comprise two lasers projecting vertical lines, parallel to a camera's viewing axis, into the environment. Using both lasers for distance estimation, the sensor offers three dimensional interpretation of the environment. This is obtained by triangulation of points extracted from the image using the Hough Transform. We evaluate the system in simulation and by physical proof-of-concept experiments on an OpenROV platform.
A nonlinear adaptive controller is proposed for the exhaust gas recirculation system on large two... more A nonlinear adaptive controller is proposed for the exhaust gas recirculation system on large two-stroke diesel engines. The control design is based on a control oriented model of the nonlinear dynamics at hand that incorporates load and engine speed changes as known disturbances to the exhaust gas recirculation. The paper provides proof of exponential stability for closed loop control of the model given. Difficulties in the system include that certain disturbance levels will make a desired setpoint in O 2 unreachable, for reasons of the physics of the system, and it is proven that the proposed control will make the system converge exponentially to the best achievable state. Simulation examples confirm convergence and good disturbance rejection over relevant operational ranges of the engine.
This paper presents a fault tolerant navigation system for a remotely operated vehicle (ROV). The... more This paper presents a fault tolerant navigation system for a remotely operated vehicle (ROV). The navigation system uses hydro-acoustic position reference (HPR) and Doppler velocity log (DVL) measurements to achieve an integrated navigation. The fault tolerant functionality is based on a modified particle filter. This particle filter is able to run in an asynchronous manner to accommodate the measurement drop out problem, and it overcomes the measurement outliers by switching observation models. Simulations with experimental data show that this fault tolerant navigation system can accurately estimate the ROV kinematic states, even when sensor failures appear frequently.
IEEE Transactions on Aerospace and Electronic Systems, 2014
Airspeed sensor faults are common causes for unmanned aerial vehicle incidents. Pitot tube cloggi... more Airspeed sensor faults are common causes for unmanned aerial vehicle incidents. Pitot tube clogging or icing are common causes. Timely diagnosis of such faults or artifacts could potentially prevent crashes. This paper employs parameter adaptive estimators to provide analytical redundancies and a dedicated diagnosis scheme is designed. Robustness is investigated on sets of flight data to estimate distributions of test statistics. The result is robust diagnosis with adequate balance between false alarm rate and sensitivity.
As a system's property, the classic observability of a linear system is dependent upon the system... more As a system's property, the classic observability of a linear system is dependent upon the system's structure (internal connections) and values of the system's parameters. Observability analysis of structured linear systems explores the generic nature of the observability as it seeks when the system is observable for almost all values that the system's parameters take. Using a graph-theoretic approach, this paper deals with observability of a specific state and input in structured linear systems.
9th IFAC Conference on Manoeuvring and Control of Marine Craft, 2012, 2012
This work addresses the issue of designing an adaptive robust control system to govern the steeri... more This work addresses the issue of designing an adaptive robust control system to govern the steering of a high speed unmanned personal watercraft (PWC) maintaining equal performance across the craft's envelope of operation. The maneuvering dynamics of a high speed PWC is presented and a strong variation over the envelope of operational conditions, including speed, is highlighted. The complexity of the nonlinear dynamics is overcome through identification of linear models at different speed regimes. A gray-box identification is conducted from full scale experiments and results in a four degrees-of-freedom surge-sway-yaw-roll model. An L 1 adaptive autopilot is then designed, which allows to achieve fast adaption to system parameters' changes and robustness of the closed loop system.
This paper treats L 1 adaptive hovering control of an unmanned surface vehicle in a station-keepi... more This paper treats L 1 adaptive hovering control of an unmanned surface vehicle in a station-keeping mode where a region of zero control authority and under-actuation are main challenges. Low-speed and reversing dynamics are identified from full scale sea trials, and parameter uncertainty is estimated. With significant parameter variation, an L 1 adaptive controller is employed for heading control. The L 1 family of controllers allows for several topologies and an architecture is suggested that suits heading control of a vessel, the requirements of which differ from that of previous L 1 literature. The control design is tackled directly in discrete time to allow a fast embedded implementation in the vehicle. Analysis of robustness, tracking performance and wave disturbance response are detailed in the paper.
Gas bearings are popular for their high speed capabilities, low friction and clean operation, but... more Gas bearings are popular for their high speed capabilities, low friction and clean operation, but su er from poor damping, which poses challenges for safe operation in presence of disturbances. Enhanced damping can be achieved through active lubrication techniques using feedback control laws. Such control design requires models with low complexity, able to describe the dominant dynamics from actuator input to sensor output over the relevant range of operation. The mathematical models based on first principles are not easy to obtain, and in many cases, they cannot be directly used for control design due to their complexity and parameter uncertainties. As an alternative, this paper presents an experimental technique for ”in situ” identification of low complexity models of the entire rotor-bearingactuator system. Using grey-box identification techniques, the approach is shown to be easily applied to industrial rotating machinery with gas bearings and to allow for subsequent control design. The paper shows how piezoelectric actuators in a gas bearing are eciently used to perturb the gas film for identification over relevant ranges of rotational speed and gas injection pressure. Parameter-varying linear models are found to capture the dominant dynamics of the system over the range of operation. Based on the identified models, decentralised proportional control is designed and is shown to obtain the required damping in theory as well as in a laboratory test rig.
An H-inf robust control of wind turbine inverters employing an LCL filter is proposed in this pap... more An H-inf robust control of wind turbine inverters employing an LCL filter is proposed in this paper. The controller dynamics are designed for selective harmonic filtering in an offshore transmission network subject to parameter perturbations. Parameter uncertainty in the network originates from the grid and the number of wind turbines connected. Power converter based turbines inject harmonic currents, which are attenuated by passive filters. A robust high order active filter controller is proposed to complement the passive filtering. The H1 design of the control loop enables desired tracking with integral effect while bounding the induced change. The design was tested in an aggregated model of the London Array offshore wind power plant and compared with traditional PI controller designs. Robust stability and performance and a reduction of control effort by 25% are obtained over the full envelope of operation.
Autonomous robots require many types of information to obtain intelligent and safe behaviours. Fo... more Autonomous robots require many types of information to obtain intelligent and safe behaviours. For outdoor operations, semantic mapping is essential and this paper proposes a stochastic automaton to localise the robot within the semantic map. For correct modelling and classification under uncertainty, this paper suggests quantising robotic perceptual features, according to a probabilistic description, and then optimising the quantisation. The
IEEE Transactions on Aerospace and Electronic Systems
Airspeed sensor faults are common causes for un-manned aerial vehicle incidents. Pitot tube clogg... more Airspeed sensor faults are common causes for un-manned aerial vehicle incidents. Pitot tube clogging or icing are common causes. Timely diagnosis of such faults or artifacts could potentially prevent crashes. This paper employs parameter adaptive estimators to provide analytical redundancies and a dedicated diagnosis scheme is designed. Robustness is investigated on sets of flight data to estimate distributions of test statistics. The result is robust diagnosis with adequate balance between false alarm rate and sensitivity.
Fault prognosis,diagnosis and fault tolerant control play important roles in continued developmen... more Fault prognosis,diagnosis and fault tolerant control play important roles in continued development of different kinds of industrial systems. In some industries,fault detection,isolation and fault tolerant control have already been well established,e.g. fault diagnosis is a standard component in automotive engines for reasons of pollution control and efficiency, in the aerospace industry on-board diagnostic algorithms can reduce maximal structure stress and reduce weight,and in other industries, fault prognosis, diagnosis and fault accommodation are technologies that gradually enter in to the new design process. This field is also very active in the academic community, where new methods are being developed to meet the challenging industrial needs. The purpose of this special issue is to gain experience in the field from both academic and industrial perspectives. Submissions to this special issue should include theory and/or methods that directly target real world problems and cases o...
This paper describes an obstacle detection system for a high-speed and agile unmanned surface veh... more This paper describes an obstacle detection system for a high-speed and agile unmanned surface vehicle (USV), running at speeds up to 30 m/s. The aim is a real-time and high performance obstacle detection system using both radar and vision technologies to detect obstacles within a range of 175 m. A computer vision horizon detector enables a highly accurate attitude estimation despite large and sudden vehicle accelerations. This further facilitates the reduction of sea clutter by utilising a attitude based statistical measure. Full scale sea trials show a significant increase in obstacle tracking performance using sensor fusion of radar and computer vision.
Pack-off is a partially or complete blocking of the circulation flow in oil and gas drilling, whi... more Pack-off is a partially or complete blocking of the circulation flow in oil and gas drilling, which can lead to costly delays. Early detection and localization of a pack-off is crucial in order to take necessary actions avoiding downtime. This incident will affect physical friction parameters in the well. A model-based adaptive observer is used to estimate these friction parameters as well as flow rates. Detecting changes to these estimates can then be used for pack-off diagnosis, which due to measurement noise is done using statistical change detection. Isolation of incident type and location is done using a multivariate generalized likelihood ratio test, determining the change direction of the estimated mean values. The method is tested on simulated data from the commercial high-fidelity multi-phase simulator OLGA, where three different pack-offs at different locations and with different magnitudes are successfully detected at an early stage and with low false alarms.
This paper investigates a minimalistic laser-based range sensor, used for underwater inspection b... more This paper investigates a minimalistic laser-based range sensor, used for underwater inspection by Autonomous Underwater Vehicles (AUV). This range detection system system comprise two lasers projecting vertical lines, parallel to a camera's viewing axis, into the environment. Using both lasers for distance estimation, the sensor offers three dimensional interpretation of the environment. This is obtained by triangulation of points extracted from the image using the Hough Transform. We evaluate the system in simulation and by physical proof-of-concept experiments on an OpenROV platform.
A nonlinear adaptive controller is proposed for the exhaust gas recirculation system on large two... more A nonlinear adaptive controller is proposed for the exhaust gas recirculation system on large two-stroke diesel engines. The control design is based on a control oriented model of the nonlinear dynamics at hand that incorporates load and engine speed changes as known disturbances to the exhaust gas recirculation. The paper provides proof of exponential stability for closed loop control of the model given. Difficulties in the system include that certain disturbance levels will make a desired setpoint in O 2 unreachable, for reasons of the physics of the system, and it is proven that the proposed control will make the system converge exponentially to the best achievable state. Simulation examples confirm convergence and good disturbance rejection over relevant operational ranges of the engine.
This paper presents a fault tolerant navigation system for a remotely operated vehicle (ROV). The... more This paper presents a fault tolerant navigation system for a remotely operated vehicle (ROV). The navigation system uses hydro-acoustic position reference (HPR) and Doppler velocity log (DVL) measurements to achieve an integrated navigation. The fault tolerant functionality is based on a modified particle filter. This particle filter is able to run in an asynchronous manner to accommodate the measurement drop out problem, and it overcomes the measurement outliers by switching observation models. Simulations with experimental data show that this fault tolerant navigation system can accurately estimate the ROV kinematic states, even when sensor failures appear frequently.
IEEE Transactions on Aerospace and Electronic Systems, 2014
Airspeed sensor faults are common causes for unmanned aerial vehicle incidents. Pitot tube cloggi... more Airspeed sensor faults are common causes for unmanned aerial vehicle incidents. Pitot tube clogging or icing are common causes. Timely diagnosis of such faults or artifacts could potentially prevent crashes. This paper employs parameter adaptive estimators to provide analytical redundancies and a dedicated diagnosis scheme is designed. Robustness is investigated on sets of flight data to estimate distributions of test statistics. The result is robust diagnosis with adequate balance between false alarm rate and sensitivity.
As a system's property, the classic observability of a linear system is dependent upon the system... more As a system's property, the classic observability of a linear system is dependent upon the system's structure (internal connections) and values of the system's parameters. Observability analysis of structured linear systems explores the generic nature of the observability as it seeks when the system is observable for almost all values that the system's parameters take. Using a graph-theoretic approach, this paper deals with observability of a specific state and input in structured linear systems.
9th IFAC Conference on Manoeuvring and Control of Marine Craft, 2012, 2012
This work addresses the issue of designing an adaptive robust control system to govern the steeri... more This work addresses the issue of designing an adaptive robust control system to govern the steering of a high speed unmanned personal watercraft (PWC) maintaining equal performance across the craft's envelope of operation. The maneuvering dynamics of a high speed PWC is presented and a strong variation over the envelope of operational conditions, including speed, is highlighted. The complexity of the nonlinear dynamics is overcome through identification of linear models at different speed regimes. A gray-box identification is conducted from full scale experiments and results in a four degrees-of-freedom surge-sway-yaw-roll model. An L 1 adaptive autopilot is then designed, which allows to achieve fast adaption to system parameters' changes and robustness of the closed loop system.
Uploads
Papers by Mogens Blanke
damping, which poses challenges for safe operation in presence of disturbances. Enhanced damping can be achieved
through active lubrication techniques using feedback control laws. Such control design requires models with low
complexity, able to describe the dominant dynamics from actuator input to sensor output over the relevant range of
operation. The mathematical models based on first principles are not easy to obtain, and in many cases, they cannot
be directly used for control design due to their complexity and parameter uncertainties. As an alternative, this paper
presents an experimental technique for ”in situ” identification of low complexity models of the entire rotor-bearingactuator
system. Using grey-box identification techniques, the approach is shown to be easily applied to industrial
rotating machinery with gas bearings and to allow for subsequent control design. The paper shows how piezoelectric
actuators in a gas bearing are eciently used to perturb the gas film for identification over relevant ranges of rotational
speed and gas injection pressure. Parameter-varying linear models are found to capture the dominant dynamics of the
system over the range of operation. Based on the identified models, decentralised proportional control is designed and
is shown to obtain the required damping in theory as well as in a laboratory test rig.
Parameter uncertainty in the network originates from the grid
and the number of wind turbines connected. Power converter
based turbines inject harmonic currents, which are attenuated
by passive filters. A robust high order active filter controller is
proposed to complement the passive filtering. The H1 design
of the control loop enables desired tracking with integral effect
while bounding the induced change. The design was tested in
an aggregated model of the London Array offshore wind power
plant and compared with traditional PI controller designs. Robust stability and performance and a reduction of control effort by 25% are obtained over the full envelope of operation.
damping, which poses challenges for safe operation in presence of disturbances. Enhanced damping can be achieved
through active lubrication techniques using feedback control laws. Such control design requires models with low
complexity, able to describe the dominant dynamics from actuator input to sensor output over the relevant range of
operation. The mathematical models based on first principles are not easy to obtain, and in many cases, they cannot
be directly used for control design due to their complexity and parameter uncertainties. As an alternative, this paper
presents an experimental technique for ”in situ” identification of low complexity models of the entire rotor-bearingactuator
system. Using grey-box identification techniques, the approach is shown to be easily applied to industrial
rotating machinery with gas bearings and to allow for subsequent control design. The paper shows how piezoelectric
actuators in a gas bearing are eciently used to perturb the gas film for identification over relevant ranges of rotational
speed and gas injection pressure. Parameter-varying linear models are found to capture the dominant dynamics of the
system over the range of operation. Based on the identified models, decentralised proportional control is designed and
is shown to obtain the required damping in theory as well as in a laboratory test rig.
Parameter uncertainty in the network originates from the grid
and the number of wind turbines connected. Power converter
based turbines inject harmonic currents, which are attenuated
by passive filters. A robust high order active filter controller is
proposed to complement the passive filtering. The H1 design
of the control loop enables desired tracking with integral effect
while bounding the induced change. The design was tested in
an aggregated model of the London Array offshore wind power
plant and compared with traditional PI controller designs. Robust stability and performance and a reduction of control effort by 25% are obtained over the full envelope of operation.