Thursday, June 21, 2012

Accuracy Soars, Research Costs Stay Grounded Using Quanser HIL Solution

Two of the most valuable benefits of Hardware-in-the-Loop (HIL) development are increased research accuracy and dramatically decreased research costs. Recent work supervised by Professor Manfredi Maggiore of theDepartment of Electrical and Computer Engineering, University of Toronto amply underscore that fact. Professor Maggiore spoke with us recently after a visit to Quanser headquarters.

Quanser: In terms of teaching and research, you’ve worked with Quanser experiments and solutions for a number of years. What is the recent high-level problem you were trying to solve or investigate?
Professor Manfredi Maggiore: My M.A.Sc. students, Farid Zare Seisan and Ashton Roza, have developed a method to control the position of a class of autonomous aerial vehicles (e.g., quadrotor and coaxial helicopters). We were particularly interested in testing this approach on a coaxial helicopter steered using a moving mass actuator.
A 3D representation of Professor Maggiore's concept helicopter
using QUARC's 3D visualization tools.

Quanser: What was the methodology you used to tackle the problem?
Prof. Maggiore: Developing helicopter prototypes is expensive and time-consuming. Additionally, significant discrepancies exist between the actual behaviour of the helicopter and that of its mathematical model. Such discrepancies are so relevant that they can invalidate a theoretical control design. One of the main causes of the above discrepancies is the uncertainty in the models of sensors and actuators.
Concerning sensors, it is well-known that the estimate of the helicopter attitude using IMU (inertial measurement unit) measurements and complementary filters is noisy and inaccurate. With regard to actuators, the aerodynamic models typically used to model the lift generated by propellers are overly simplified, and do not take into account the motor dynamics and a number of aerodynamic effects. For our coaxial helicopter, we imagined that the moving mass actuator would limit the performance of the closed-loop system, but we had no concrete idea of the extent of this limitation.
We wanted to test our controllers in a realistic scenario that would take into account sensor and actuator limitations, but without having to spend the time and money to develop a full helicopter prototype. We used the Quanser 3 DOF Gyroscope experiment to do a hardware-in-the-loop simulation of the coaxial helicopter.
Shown:  HIL rapid prototyping device consisting of Quanser's 3 DOF Gyroscope
retrofitted with Quanser's HiQ avionics sensor board.

This experiment works like this. A microcontroller with IMU is mounted on a plate placed at the centre of a fully actuated 3 DOF gyroscope. The gyroscope is driven by the Simulink simulation of the coaxial helicopter. In turn, the simulation receives inputs from real sensors and actuators. Specifically, the feedback in Simulink is implemented using actual IMU measurements, and the helicopter dynamics are affected by the actual displacement of the moving mass actuator, which is driven by the reference signals generated by the position controller in the Simulink diagram.

                                                                                                                      - Quanser Video 

Quanser: In using the hardware-in-the-loop platform to study the problem, did you discover anything about the problem that you would not have seen otherwise?
Prof. Maggiore: We realized that the moving mass actuator had a time delay of the order of 0.1 seconds which severely limited the stability and performance of the closed-loop system. This forced us to detune their controller to make it less aggressive. We also realized that although the sensor noise can be significant, it does not pose a stability problem. It only induces a steady-state error in the helicopter's position.
The moral of the story for us was that the design of a moving mass actuator is crucial, and it should be improved before trying to develop a prototype of a coaxial helicopter. The HIL scenario allowed us to do that quickly and without incurring major research costs.

Quanser: What benefits do you see for using this type of hardware-in-the-loop system in the classroom, undergraduate or graduate, or for research?
Prof. Maggiore: The hardware-in-the-loop system described above (the Quanser 3 DOF gyroscope with an external moving mass actuator) allowed us to take an intermediate step between a purely theoretical analysis and the full implementation of a coaxial helicopter with a moving mass actuator. Quanser’s real-time control software, QUARC, provides the ideal development tools to implement such a system from theoretical analysis to hardware-in-the-loop simulations, and even full hardware implementation.
On the research side, more work can be done to exploit this platform. For instance, one could include actual motors and propellers in the hardware-in-the-loop simulation to test the impact of the motor dynamics on the closed-loop system. On the teaching side, one could use the Quanser 3 DOF gyroscope to present an evocative simulation of a UAV, one in which certain practical issues are taken into account.

Funding a full UAV lab for teaching courses is a challenge for many professors, especially when you consider the maintenance involved in running such a lab and the safety precautions that would have to be adopted. All in all, we were extremely happy with the 3 DOF Gyroscope experiment, and we learned a great deal from it.

Quanser: Thank you, Professor Maggiore.

Dr. Manfredi Maggiore is an Associate Professor in the Department of Electrical and Computer Engineering at the University of Toronto, in Toronto, Canada.  He has been  associated with the University's Systems Control Group since September, 2000.

No comments: