Results
On this page you can find an overview of the development cases studied in LeCoPro and a selection of the project results achieved on these cases.
Development cases:
 Transmission |
 Badminton demonstrator |
 Tractor with implement |
 Tractor in LEGO MINDSTORMS |
 Sliding mass with optical sensor |
 Hydrostatic drivetrain |
Transmission
In this development case, learning control is applied to the engagement of wet clutches. The objective is to learn the optimal current profile to the electro-hydraulic valve, which controls the pressure of the oil to the clutch. A wet clutch is a system with one input, which is characterised by a hard non-linearity when the piston of the clutch gets in contact with the friction plates. These clutches are typically used in power transmissions of off-road vehicles, which operate under strongly varying environmental conditions. These features make the control of wet clutches a suitable case to validate the techniques of WP 1. The validation experiments are carried out on a dedicated test bench (Fig. 1), where an electromotor (30kW) drives a flywheel (25 kgm2) via a torque convertor and two mechanical transmissions. The developed control scheme is tested on the first range clutch of the left transmission while the right transmission is used to vary the actual load observed by the first transmission. This set-up will also be used to test the global learning control techniques developed in WP 2. The techniques of WP 2 will be applied to control the current to the valves of the opening and closing clutch during a gear shifting.
Fig. 1: Transmission test set-up
Selected results:
-
Fine tuning of a wet clutch engagement by means of a genetic algorithm
Y. Zhong, A. Dutta, B. Wyns, C.M. Ionescu, G. Pinte, W. Symens, J. Stoev and R. De Keyser
8th AIAI (Artificial Intelligence Applications and Innovations) Conference
Halkidiki, Greece, September 27 - 30, 2012 -
Switched nonlinear predictive control with adaptive references for engagement of wet clutches
A. Dutta, C. Ionescu, B. Wyns, R. De Keyser, J. Stoev, G. Pinte , W. Symens
IFAC Conference on Nonlinear Model Predictive Control 2012
Noordwijkerhout, the Netherlands , August 23 - 27, 2012 - A reference free iterative learning strategy for wet clutch control
B. Depraetere, G. Pinte, J. Swevers
American Control Conference 2011
San Francisco, California, USA, June 29 - July 1, 2011 -
Improving wet clutch engagement with Reinforcement Learning
K. Van Vaerenbergh, A. Rodr?guez, M. Gagliolo, P. Vrancx, A. Nowé, J. Stoev, S. Goossens, G. Pinte, W. Symens
International Joint Conference on Neural Networks 2012
Brisbane, Australia, June 10 - 15, 2012
Badminton demonstrator
To demonstrate its mechatronic competences, FMTC has constructed a badminton robot (Fig. 2). A visual system with two high-definition black and white cameras is used to localize the fast moving shuttle. The robot itself consists of one linear motor and two rotational motors. In the LeCoPro project, the application of learning techniques to the motion control of these different motors will be investigated. The main challenge is the development of fast learning algorithms, which can take into account the uncertainty of the estimated shuttle trajectory.
A movie of the badminton robot in action can be found here.
Fig. 2: Badminton demonstrator
Selected results:
- Model-free and model-based time-optimal control of a badminton robot
M. Liu, B. Depraetere, G. Pinte, I. Grondman and R. Babuska
To appear in the 9th IEEE Asian Control Conference (ASCC 2013)
Istanbul, Turkey, June 23 - 26, 2013 - Energy Optimal Point-to-Point Motion using Model Predictive Control
X. Wang, J. Stoev, G. Pinte and J. Swevers
2012 ASME Dynamic Systems and Control Conference - 11th Motion and Vibration Conference
Fort Lauderdale, Florida, USA, October 17 - 19, 2012 - Energy-optimal time allocation for a series of point-to-point motions
P. Janssens, G. Pipeleers, Moritz Diehl, J. Swevers
31st Benelux Meeting on Systems and Control,
Heijen/Nijmegen, The Netherlands, March 27 - 29, 2012
Tractor with implement
This development case consists of a small tractor (Fig. 3), that can be controlled automatically (steering, speed), and a carried implement, that has rear wheels to steer relative to the tractor. In a first phase, learning control will be applied at the subsystem level to control the tractor. In a second phase the entire system with two subsystems (tractor + implement) will be considered. The main challenges in this application for the developed learning controllers will lie in the effect of varying soil conditions and the interaction between the tractor and the implement.
A movie of the tractor and implement action can be found here.
Fig. 3: Tractor with implement
Selected results:
- Model Predictive Control of the Yaw Dynamics of an Autonomous Tractor
E. Kayacan, E. Kayacan, H. Ramon and W. Saeys
International Conference on Robotics and Associated High-Technologies and Equipment for Agriculture
Pisa, Italy, September 19 - 21, 2012 -
Intelligent Control of a Tractor-Implement System Using Type-2 Fuzzy Neural Networks
E. Kayacan, W. Saeys, E. Kayacan, H. Ramon, O. Kaynak
The 2012 IEEE World Congress on Fuzzy systems
Brisbane, Australia, June 10 - 15, 2012 - Iterative feedback tuning to learn steering control of an autonomous tractor
E. Hostens, G. Pinte, W. Symens
31st Benelux Meeting on Systems and Control,
Heijen/Nijmegen, The Netherlands, March 27 - 29, 2012
Automous tractor built in LEGO MINDSTORMS
For the ease of developing model-free controllers in a safe environment, a scaled version of the autonomous tractor has been built in LEGO MINDSTORMS. The result is shown in Fig. 4, and a movie of the tractor in action can be seen here
Fig. 4: Scaled version of tractor built in LEGO MINDSTORMS
Sliding mass with optical sensor
A fourth development case deals with the motion control of a sliding mass which is excited by the impact of a mass-spring-damper system (Fig. 5). The input signal to the motor which actuates the mass-spring-damper system, has to be controlled such that the impacted mass follows a desired motion. Only a single optical sensor is installed to check when the excited mass passes a certain position. The objective behind this set-up is to evaluate the potential of learning control on systems where the controlled output can not be measured continuously, but only at discrete positions. Learning control algorithms, which can deal with the restricted position information, will be developed.
Fig. 5: Sliding mass with optical sensor
Selected results:
-
Identification of linear systems with binary outputs using short independent experiments
B. Depraetere, Y. Stoev, G. Pinte, J. Swevers
16th IFAC Symposium on System Identification, SYSID 2012
Brussels, Belgium, July 11 - 13, 2012 - Policy gradient methods for controlling systems with discrete sensor information
M. Gagliolo, K. van Vaerenbergh, A. Rodriguez, A. Nowé, S. Goossens, G. Pinte, W. Symens
20th Annual Belgian-Dutch Conference on Machine Learning, BeNeLearn 2011
May 20, The Hague, The Netherlands
Hydrostatic drivetrain
A fifth development case is a hydrostatic drivetrain (Fig. 6), consisting of an electromotor, a hydraulic pump and two hydraulic motors. Heavy duty mobile machines such as road construction vehicles, agricultural and forestry machines are frequently driven by this type of drivetrain. As this development case consists of several strongly interacting subsystems both in the electrical, mechanical and hydraulic domain, it is an ideal development case for global learning control algorithms.
Fig. 6: Hydrostatic drivetrain