Erik Hellström
Research Subjects
Erik Hellström has a total of 43 co-authors. Here are the top co-authors by number of shared publications:
Li Jiang
Robert Bosch LLC, Farmington Hills, MI, USA
21 shared publicationsAnna G. Stefanopoulou
Department of Mechanical Engineering, University of Michigan, Ann Arbor, MI, USA
14 shared publicationsJacob Larimore
Department of Mechanical Engineering, University of Michigan, Ann Arbor, MI, USA
10 shared publicationsShyam Jade
Department of Mechanical Engineering, University of Michigan, Ann Arbor, MI, USA
9 shared publicationsAnna Stefanopoulou
University of Michigan, Ann Arbor, MI
8 shared publicationsLars Nielsen
Linköping University, Linköping, Sweden
6 shared publications
Latest Publications
- 25 November 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Control Systems Letters
- Vol. 6, 1664-1669
Control Barrier Functions With Unmodeled Input Dynamics Using Integral Quadratic Constraints
This letter considers the design of controllers that achieve safety for systems with unmodeled dynamics at the plant input. Simplified, low-order models are often used in the design of such controllers. Unmodeled dynamics (e.g., actuator dynamics, time delays, etc.) can lead to safety violations. This letter proposes a method to achieve safety in the presence of these unmodeled dynamics. The approach combines control barrier functions (CBFs) and integral quadratic constraints (IQCs). First, the input-output behavior of the unmodeled dynamics is bounded in the time domain using an$\alpha $ -IQC. Next, the$\alpha $ -IQC is incorporated into the CBF constraint to ensure safety. A safe controller is implemented by solving, in real-time, a convex optimization subject to the CBF constraint with the$\alpha $ -IQC. The approach is demonstrated with a simple example.- 25 November 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Control Systems Letters
- Vol. 6, 1664-1669
Control Barrier Functions With Unmodeled Input Dynamics Using Integral Quadratic Constraints
This letter considers the design of controllers that achieve safety for systems with unmodeled dynamics at the plant input. Simplified, low-order models are often used in the design of such controllers. Unmodeled dynamics (e.g., actuator dynamics, time delays, etc.) can lead to safety violations. This letter proposes a method to achieve safety in the presence of these unmodeled dynamics. The approach combines control barrier functions (CBFs) and integral quadratic constraints (IQCs). First, the input-output behavior of the unmodeled dynamics is bounded in the time domain using an$\alpha $ -IQC. Next, the$\alpha $ -IQC is incorporated into the CBF constraint to ensure safety. A safe controller is implemented by solving, in real-time, a convex optimization subject to the CBF constraint with the$\alpha $ -IQC. The approach is demonstrated with a simple example.- 3 April 2018
- Published by SAE International in SAE International Journal of Advances and Current Practices in Mobility
- Vol. 1
Improving Transient Torque Response for Boosted Engines with VCT and EGR
Modern gasoline engines have increased part-load fuel economy and specific power output through technologies such as downsizing, turbocharging, direct injection, and exhaust gas recirculation. These engines tend to have higher sensitivity to driving behavior because of the steady-state efficiency versus output characteristics (e.g., sweet spot at lower output) and the dynamic response characteristics (e.g., turbo lag). It has been observed that the technologies aimed at increased engine efficiency may improve fuel economy for less aggressive cycles and drivers while hurting fuel economy for more aggressive cycles and drivers. The higher degrees of freedom in these engines and the increased sensitivity make controls and calibration more complex and more important at the same time. With the interactions between the dynamic response characteristics of the powertrain and the driver in mind, a dynamic control strategy for variable cam timing (VCT) and exhaust gas recirculation (EGR) is developed. The strategy allows actuator positions at steady-state optimal values when possible yet a fast response proportional to the driver request in transients. The aim is to strike a balance, which is tunable, between steady state efficiency and transient response. Most of the calibration process is algorithmic and based on standard engine mapping data. Experimental results for fuel economy on drive cycles and performance testing from powertrain and chassis dynamometers for two powertrain configurations are reported. Analysis shows improvements in terms of fuel economy and driver demand tracking on drive cycles as well as improved performance metrics. In particular, it is demonstrated that it is possible to simultaneously improve transient performance and fuel economy.
- 3 April 2018
- Published by SAE International in SAE International Journal of Advances and Current Practices in Mobility
- Vol. 1
Improving Transient Torque Response for Boosted Engines with VCT and EGR
Modern gasoline engines have increased part-load fuel economy and specific power output through technologies such as downsizing, turbocharging, direct injection, and exhaust gas recirculation. These engines tend to have higher sensitivity to driving behavior because of the steady-state efficiency versus output characteristics (e.g., sweet spot at lower output) and the dynamic response characteristics (e.g., turbo lag). It has been observed that the technologies aimed at increased engine efficiency may improve fuel economy for less aggressive cycles and drivers while hurting fuel economy for more aggressive cycles and drivers. The higher degrees of freedom in these engines and the increased sensitivity make controls and calibration more complex and more important at the same time. With the interactions between the dynamic response characteristics of the powertrain and the driver in mind, a dynamic control strategy for variable cam timing (VCT) and exhaust gas recirculation (EGR) is developed. The strategy allows actuator positions at steady-state optimal values when possible yet a fast response proportional to the driver request in transients. The aim is to strike a balance, which is tunable, between steady state efficiency and transient response. Most of the calibration process is algorithmic and based on standard engine mapping data. Experimental results for fuel economy on drive cycles and performance testing from powertrain and chassis dynamometers for two powertrain configurations are reported. Analysis shows improvements in terms of fuel economy and driver demand tracking on drive cycles as well as improved performance metrics. In particular, it is demonstrated that it is possible to simultaneously improve transient performance and fuel economy.
- 28 March 2017
- Published by SAE International in SAE International Journal of Advances and Current Practices in Mobility
- Vol. 1
Automated Calibration for Compressor Recirculation Valve Control
Turbocharger compressors are susceptible to surge – the instability phenomena that impose limitations on the operation of turbocharged engines because of undesired noise, engine torque capability constraints, and hardware strain. Turbocharged engines are typically equipped with a binary compressor recirculation valve (CRV) whose primary function is to prevent compressor surge. Calibration of the associated control strategy requires in-vehicle tests and usually employs subjective criteria. This work aims to reduce the calibration effort for the strategy by developing a test procedure and data processing algorithms. An automated calibration for CRV control is developed that will generate a baseline calibration that avoids surge events. The effort to obtain the baseline calibration, which can be further fine-tuned, is thereby significantly reduced. Procedures are developed for testing and analysis that combines a previously developed surge characterization method with unconstrained nonlinear optimization. An adaptive algorithm is also presented that can update the CRV control calibration online. The developed procedures are demonstrated on vehicle data and compared with manually obtained calibrations.
- 28 March 2017
- Published by SAE International in SAE International Journal of Advances and Current Practices in Mobility
- Vol. 1
Automated Calibration for Compressor Recirculation Valve Control
Turbocharger compressors are susceptible to surge – the instability phenomena that impose limitations on the operation of turbocharged engines because of undesired noise, engine torque capability constraints, and hardware strain. Turbocharged engines are typically equipped with a binary compressor recirculation valve (CRV) whose primary function is to prevent compressor surge. Calibration of the associated control strategy requires in-vehicle tests and usually employs subjective criteria. This work aims to reduce the calibration effort for the strategy by developing a test procedure and data processing algorithms. An automated calibration for CRV control is developed that will generate a baseline calibration that avoids surge events. The effort to obtain the baseline calibration, which can be further fine-tuned, is thereby significantly reduced. Procedures are developed for testing and analysis that combines a previously developed surge characterization method with unconstrained nonlinear optimization. An adaptive algorithm is also presented that can update the CRV control calibration online. The developed procedures are demonstrated on vehicle data and compared with manually obtained calibrations.
- 1 January 2016
- Published by Elsevier in IFAC-PapersOnLine
- Vol. 49 (11) , 702-707
Characterizing and Detecting Surge and Co-Surge in Automotive Compressors
- 1 January 2016
- Published by Elsevier in IFAC-PapersOnLine
- Vol. 49 (11) , 702-707
Characterizing and Detecting Surge and Co-Surge in Automotive Compressors
- 1 July 2015
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 13 (07431619) , 3342-3347
A driver model for velocity tracking with look-ahead
Understanding the dynamics of the combined system of driver and vehicle opens up for improving the performance of powertrain controls. To this end, a new model structure is formulated for accelerator pedal behavior of human drivers. The structure is comprised of feedback and feedforward with preview of the reference. A system identification approach is developed for determining the parameters in the structure. Models are identified on data from experiments with different driver and vehicle combinations where the drivers are tracking the common FTP and US06 drive cycles. Experiments where the same individual drives different cycles show the driver adapting in response to how demanding the drive cycle is and individual driving styles are studied with another data set with different drivers on the same cycle. The results show that drivers have similar feedforward for a given vehicle while the look-ahead and feedback is variable based on drive cycle and driving style. The potential of using the driver model for prediction is demonstrated. For the first time, for longitudinal experiments, it is shown that the open loop frequency response for the driver cascaded with the vehicle is approximated by the so-called crossover model near the crossover frequency.
- 1 July 2015
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 13 (07431619) , 3342-3347
A driver model for velocity tracking with look-ahead
Understanding the dynamics of the combined system of driver and vehicle opens up for improving the performance of powertrain controls. To this end, a new model structure is formulated for accelerator pedal behavior of human drivers. The structure is comprised of feedback and feedforward with preview of the reference. A system identification approach is developed for determining the parameters in the structure. Models are identified on data from experiments with different driver and vehicle combinations where the drivers are tracking the common FTP and US06 drive cycles. Experiments where the same individual drives different cycles show the driver adapting in response to how demanding the drive cycle is and individual driving styles are studied with another data set with different drivers on the same cycle. The results show that drivers have similar feedforward for a given vehicle while the look-ahead and feedback is variable based on drive cycle and driving style. The potential of using the driver model for prediction is demonstrated. For the first time, for longitudinal experiments, it is shown that the open loop frequency response for the driver cascaded with the vehicle is approximated by the so-called crossover model near the crossover frequency.
- 2 March 2015
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Control Systems Technology
- Vol. 23 (6) , 2144-2154
Adaptive Control of a Recompression Four-Cylinder HCCI Engine
An adaptive controller is presented for the control of combustion phasing in a multicylinder homogeneous charge compression ignition engine. Adaptive parameter estimation is used to modify a model-based feedforward controller for each cylinder's start of injection (SOI) timing in an effort to mitigate model errors and increase the feedforward control accuracy. In-cylinder pressure measurements are used to calculate combustion phasing, which is compared with the prediction of an online nonlinear engine model to drive the parameter estimation that adapts the feedforward controller. It is demonstrated through experiments that the adaptive parameter can reduce the parameterization effort by allowing the model to adapt and match the response of each cylinder. It is also shown that the adaptive feedforward control is more accurate in the sense that load transitions can be achieved with less correction from the feedback controller. Overall, an average reduction of 41% in the absolute value of the SOI feedback component at steady state is achieved.
- 2 March 2015
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Control Systems Technology
- Vol. 23 (6) , 2144-2154
Adaptive Control of a Recompression Four-Cylinder HCCI Engine
An adaptive controller is presented for the control of combustion phasing in a multicylinder homogeneous charge compression ignition engine. Adaptive parameter estimation is used to modify a model-based feedforward controller for each cylinder's start of injection (SOI) timing in an effort to mitigate model errors and increase the feedforward control accuracy. In-cylinder pressure measurements are used to calculate combustion phasing, which is compared with the prediction of an online nonlinear engine model to drive the parameter estimation that adapts the feedforward controller. It is demonstrated through experiments that the adaptive parameter can reduce the parameterization effort by allowing the model to adapt and match the response of each cylinder. It is also shown that the adaptive feedforward control is more accurate in the sense that load transitions can be achieved with less correction from the feedback controller. Overall, an average reduction of 41% in the absolute value of the SOI feedback component at steady state is achieved.
- 8 October 2014
- Published by SAGE Publications in International Journal of Engine Research
- Vol. 16 (3) , 474-484
Real-time internal residual mass estimation for combustion with high cyclic variability
In this work, a physics-based method of estimating the residual mass in a recompression homogeneous charge compression ignition engine is developed and analyzed for real-time implementation. The estimation routine is achieved through in-cylinder pressure and exhaust temperature measurements coupled with energy and mass conservation laws applied during the exhaust period. Experimental results on a multicylinder gasoline homogeneous charge compression ignition engine and dynamic analysis demonstrate the estimation routine’s ability to perform across a wide range of operating conditions as well as on a cycle-by-cycle basis for highly variable combustion phasing data.
- 8 October 2014
- Published by SAGE Publications in International Journal of Engine Research
- Vol. 16 (3) , 474-484
Real-time internal residual mass estimation for combustion with high cyclic variability
In this work, a physics-based method of estimating the residual mass in a recompression homogeneous charge compression ignition engine is developed and analyzed for real-time implementation. The estimation routine is achieved through in-cylinder pressure and exhaust temperature measurements coupled with energy and mass conservation laws applied during the exhaust period. Experimental results on a multicylinder gasoline homogeneous charge compression ignition engine and dynamic analysis demonstrate the estimation routine’s ability to perform across a wide range of operating conditions as well as on a cycle-by-cycle basis for highly variable combustion phasing data.
- Open Access
- 5 September 2014
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Control Systems Technology
- Vol. 23 (3) , 868-881
Controlled Load and Speed Transitions in a Multicylinder Recompression HCCI Engine
A model-based control strategy to track combustion phasing during load and speed transitions in the homogeneous charge compression ignition (HCCI) operating region of a multimode combustion engine is presented in this paper. HCCI transitions can traverse regions of high cyclic variability (CV), even if the steady state transition end points are stable with low CV. A control-oriented HCCI model for both the stable, low CV region and the oscillatory, high CV, late phasing region is used to design a controller that uses valve and fuel injection timings to track combustion phasing. Novel aspects of the controller include nonlinear model-inversion-based feedforward and gain scheduled feedback based on unburned fuel that reduces transient CV. Transitions tested on a multicylinder HCCI engine include load transitions at fixed engine speeds, simultaneous load, and speed transitions, and select FTP75 drive-cycle transitions with high load slew rates. Good combustion phasing tracking performance is demonstrated, and misfires are prevented.
- 5 September 2014
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Control Systems Technology
- Vol. 23 (3) , 868-881
Open AccessControlled Load and Speed Transitions in a Multicylinder Recompression HCCI Engine
A model-based control strategy to track combustion phasing during load and speed transitions in the homogeneous charge compression ignition (HCCI) operating region of a multimode combustion engine is presented in this paper. HCCI transitions can traverse regions of high cyclic variability (CV), even if the steady state transition end points are stable with low CV. A control-oriented HCCI model for both the stable, low CV region and the oscillatory, high CV, late phasing region is used to design a controller that uses valve and fuel injection timings to track combustion phasing. Novel aspects of the controller include nonlinear model-inversion-based feedforward and gain scheduled feedback based on unburned fuel that reduces transient CV. Transitions tested on a multicylinder HCCI engine include load transitions at fixed engine speeds, simultaneous load, and speed transitions, and select FTP75 drive-cycle transitions with high load slew rates. Good combustion phasing tracking performance is demonstrated, and misfires are prevented.
- 1 July 2014
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Control Systems Technology
- Vol. 22 (4) , 1408-1421
Reference Governor for Load Control in a Multicylinder Recompression HCCI Engine
This paper presents a model-based control strategy designed to regulate combustion phasing during load transitions in a recompression homogeneous charge compression ignition (HCCI) engine. A low-order discrete-time control-oriented model for recompression HCCI combustion is developed that represents the strong thermal and composition coupling between engine cycles. A baseline two-input single-output controller is designed to regulate combustion phasing, using the amount of negative valve overlap and the fuel injection timing as actuators. This controller is augmented by a reference or fuel governor, which modifies transient fuel mass commands during large load transitions, when future actuator constraint violations are predicted. This approach is shown in experiments to improve combustion phasing and load responses, preventing engine misfires in some cases. The fuel governor enables larger load transitions than were possible with the baseline controller alone. The governor acts only when future actuator constraint violations are predicted. The complexity and computational overhead of the governor are reduced by developing a linearized fuel governor. Satisfactory performance is demonstrated experimentally for a range of engine speeds.
- 1 July 2014
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Control Systems Technology
- Vol. 22 (4) , 1408-1421
Reference Governor for Load Control in a Multicylinder Recompression HCCI Engine
This paper presents a model-based control strategy designed to regulate combustion phasing during load transitions in a recompression homogeneous charge compression ignition (HCCI) engine. A low-order discrete-time control-oriented model for recompression HCCI combustion is developed that represents the strong thermal and composition coupling between engine cycles. A baseline two-input single-output controller is designed to regulate combustion phasing, using the amount of negative valve overlap and the fuel injection timing as actuators. This controller is augmented by a reference or fuel governor, which modifies transient fuel mass commands during large load transitions, when future actuator constraint violations are predicted. This approach is shown in experiments to improve combustion phasing and load responses, preventing engine misfires in some cases. The fuel governor enables larger load transitions than were possible with the baseline controller alone. The governor acts only when future actuator constraint violations are predicted. The complexity and computational overhead of the governor are reduced by developing a linearized fuel governor. Satisfactory performance is demonstrated experimentally for a range of engine speeds.
- 1 June 2014
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- No. 07431619,p. 849-854
Mode switches among SI, SACI, and HCCI combustion and their influence on drive cycle fuel economy
A vehicle and driver model from (Nuesch et al., 2013) is here integrated with a finite state model describing mode switches among spark-ignited (SI) combustion and the advanced combustion modes spark assisted compression ignition (SACI) and homogeneous charge compression ignition (HCCI). The model is used to quantify the influence of mode switch fuel penalties on drive cycle fuel economy considering the federal test procedure (FTP-75) and highway fuel economy test (HWFET). The simulation under the assumed fuel penalties and dynamics shows a very small fuel loss due to harmful switches. Harmful switches are mode switches that lead to short stays in the advanced combustion, where the penalty in switching is greater than the benefit achieved after the switch. The simulations also highlight the benefits of integrating the SACI mode. Apart from extending the fuel efficiency improvements beyond the HCCI range, it is postulated that SACI combustion is easier to reach (lower fuel penalty and faster response than the HCCI-SI switch) from both SI and HCCI, creating a bridge and an economic destination in the overall speed-load space. The combined operation of SACI and HCCI leads to substantially higher improvements, especially for the FTP-75 drive cycle, than using either mode individually.
- 1 June 2014
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- No. 07431619,p. 849-854
Mode switches among SI, SACI, and HCCI combustion and their influence on drive cycle fuel economy
A vehicle and driver model from (Nuesch et al., 2013) is here integrated with a finite state model describing mode switches among spark-ignited (SI) combustion and the advanced combustion modes spark assisted compression ignition (SACI) and homogeneous charge compression ignition (HCCI). The model is used to quantify the influence of mode switch fuel penalties on drive cycle fuel economy considering the federal test procedure (FTP-75) and highway fuel economy test (HWFET). The simulation under the assumed fuel penalties and dynamics shows a very small fuel loss due to harmful switches. Harmful switches are mode switches that lead to short stays in the advanced combustion, where the penalty in switching is greater than the benefit achieved after the switch. The simulations also highlight the benefits of integrating the SACI mode. Apart from extending the fuel efficiency improvements beyond the HCCI range, it is postulated that SACI combustion is easier to reach (lower fuel penalty and faster response than the HCCI-SI switch) from both SI and HCCI, creating a bridge and an economic destination in the overall speed-load space. The combined operation of SACI and HCCI leads to substantially higher improvements, especially for the FTP-75 drive cycle, than using either mode individually.