SCIENTIFIC NEWS AND
INNOVATION FROM ÉTS
An Algorithm to Increase the Speed of Feedback Systems - By : David Bensoussan,

An Algorithm to Increase the Speed of Feedback Systems


David Bensoussan
David Bensoussan Author profile
David Bensoussan teaches at the Department of Electrical Engineering of ÉTS. His research interests are control algorithms, fibre optic systems and digital telecommunications. He is also the co-inventor of home-automation products.

feedback systems

A new control algorithm has been patented by Professor David Bensoussan at the Electrical Engineering Department of the Montreal École de technologie supérieure (ÉTS). This new algorithm increases feedback system response time without reducing stability or robustness. Comparison with other feedback systems, and control methods has shown that this algorithm minimises feedback system response time.

Control Algorithm Characteristics

This control algorithm can provide very quick response times while maintaining good stability (figure 1). In general, in order to increase a system’s response time, a much larger gain is required; this principle, however, does not apply to all systems because it can lead to oscillations and self-destructive instabilities.

Figure 1 : Illustration of the Servo System

Figure 1 : Illustration of the Servo System

This new algorithm provides a more competitive response time through increased gain without reducing feedback systems stability or robustness.

Control Method History

In the 60s, the description of systems using state variables and the plethora of computer solutions led researchers to pursue methods referred to as “classical” at that time. These solutions were developed for stochastic systems with well-understood, though random, perturbations. Unfortunately, these models were only an approximation of reality. Many processes have varying degrees of uncertainty and the statistical properties of their perturbations are not always well-understood.

Classical control methods are well-suited for handling uncertainties because they are based on differential equations which are then translated into the frequency domain, resulting in a series of exponential time-functions.

In the 80s, a system “input-output” approach facilitated a reformulation of the control problem separating the quest for optimum sensitivity from the pursuit of stability and to optimise the phase compensation circuits for that purpose. Robustness methods were applied to minimise the effects of system perturbations and uncertainties but temporal performance, which is also an important criterion, was not a priority in the pursuit of robust control.

Figure 2: Time Response of Typical Controllers

Figure 2: Time Response of Typical Controllers

Past Research

Past research efforts sought to reformulate the performance problem, in both time and frequency domains, in order to design compensation methods that would ensure both stability and robustness. Phase circuits were introduced to improve temporal characteristics without actually affecting the frequency characteristics.

Professor Bensoussan had to rethink feedback systems in the light of new technologies: New sensors with greater precision, higher resolution and larger bandwidth. In addition, today’s Digital Signal Processors (DSPs), are much quicker and provide greater flexibility.

Figure 3: Temporal Response of the Compensator Designed by Professor Bensoussan

Figure 3: Temporal Response of the Compensator Designed by Professor Bensoussan

Improvements in the temporal characteristics, obtained using phase circuits, can now be modified by reducing compensator gain in order to avoid saturation at the inputs to the controlled process.

Experimentation

In collaboration with Professor Benoit Boulet of McGill University and a combined team of ÉTS-McGill researchers, this algorithm was tested in a laboratory using a magnetic levitation system. The experimental results have validated the performance of the theoretical results: The robustness margins are preserved and the temporal solution is competitive with respect to existing solutions, especially with respect to PID controllers (proportional-integral-differential). It should also be noted that, in this new algorithm, stabilisation and rise times are both equal.

Figure 4: Experimental Setup

Figure 4: Experimental Setup

One of the main advantages of the algorithm is that it can also be translated into the classical control domain whose graphical methods provide a more intuitive understanding of the solution.

Potential Applications

This algorithm is universal and can be applied to any system in which stability and response time are important. As one example, it could be applied to control the reading arms of hard-drives in order to increase their reading speeds. Automated assembly is another potential domain of applications where there is significant potential for improvement. Aligo Innovation, a partner company, specialises in the marketing of university innovations and is responsible for promoting this algorithm (US Patent Number 8,831,755).

Future Projects

On the academic level, there are many options for continued research including an application of the algorithm to multivariable systems and the accommodation of various system nonlinearities and processing times. On the practical side, this algorithm can be applied to a large range of control systems in which response time is a critical factor. A universal controller, provided as a set of software tools, could also be developed to adapt to various process characteristics and to the available computational power of a given microprocessor or digital signal processor (DSP).

David Bensoussan

Author's profile

David Bensoussan teaches at the Department of Electrical Engineering of ÉTS. His research interests are control algorithms, fibre optic systems and digital telecommunications. He is also the co-inventor of home-automation products.

Program : Electrical Engineering 

Author profile


Get the latest scientific news from ÉTS
comments

Leave a Reply

Your email address will not be published. Required fields are marked *