ACODUASIS

Automatic Control Design Using Advanced Simulation Software

Automation is more and more part of our everyday life: car driving is increasingly computer assisted, new metros are now automatically driven, and robots are progressively used in many different areas, to transport goods, people, or to explore submarine places. That follows a logical trend: to automate the control of equipments so as to increase people’s safety, quality of life, or scientific knowledge.

That explains the increasing demand to have more efficient, reliable and safe automated systems in a short time-to-market. But although most technological barriers have been overcome (from a mechanical point of view, for example), one issue remains and limits the development of new automated systems: the design of complex automatic control systems. Indeed, it requires important skills in mathematics, non-linear, complex modelling and programming, in order to test, simulate and validate the control modules.

Nowadays, there is only a limited number of tools existing on the market and they are not complete. There is a need to benefit from innovative approaches to be able to design control systems that are both efficient, reliable, while consuming limited time so as to meet the market requirements. The ACODUASIS project does well answer this need.

The purpose of this project is to make them benefit from the innovative approach of EICAS, an Italian SME, specialised in automatic control design which has develop an advanced software for its internal use. This software has several advantages over competing solutions. The following technically innovative aspects can be highlighted: Automated algorithm and code generation, Model identification, Control parameters numerical optimisation and Default fine model class. It allows reducing the time spent in the design phase and will increase companies competitiveness and time to response.

Within the project, two different versions of the software will be developed and transferred to industrials through intervention of two specialises bodies (UNIKARL and UNINOVA):
· A general purpose simulator, that will be used by automatic control designers, able to solve any control issue
· A customised simulator, specifically developed for robotics. Tested and validate by COMAU (fixed robot representative) and TELELIFT (mobile robot representative), this simulator will require less skills than the first version, be more user-friendly and easy-to-use so as to be successfully implemented and used by industrials.

The software structure is compose by several tools, the most important are:
· SIMBUILDER: is the tool which helps the user in developing control algorithms and related codes and in programming the simulated environment for testing the control design. It includes the Automated Algorithm Generation functions which help the user in the control algorithm design.
· SIM: is the tool which enables the performance of simulation trials. A customized SIM program is created by means of the SIMBUILDER for each specific control design project. It includes the simulation of the plant, of its environment and of the control system. It includes also the algorithm for the model identification and the algorithm for the control optimisation, when these options are required by the user.
· POST: is the tool which allows the analysis off-line of the simulation trial results.

SIMBUILDER PECULIARITIES
Two distinct areas exist, the PLANT area and the CONTROL area.
The first area is used to program the simulation of the plant and its environment. Continuous or discrete or hybrid plants can be simulated. It is also possible to use data recorded in experimental trials performed in actual plants.
The second area is used to design the plant control system, which is conceived as a potential multi-processor system, where each processor may perform multi control functions. Each control function has its own sampling frequency and the user must schedule it by means of a suitable scheduler. Data can be exchanged among the control functions running within the same processor and among different processors. The data transmission is simulated and the transmission time is simulated.
Automated Algorithm Generation functions are available to help the user in the control algorithm design. The user can choose within a set of predefined feedback control algorithm design. The user can choose within a set of predefined feedback control architectures or state estimators and forecasting models, all related to linear dynamic models. The use of the Automated Algorithm Generation functions requires that a plant linear model be available and that the required control or forecasting performance be stated according to defined standard performance indicators. Algorithm and related code are automatically generated, which are strictly in conformity with the stated performance indicators.
In order to be practically useful, the automated algorithm and code generation techniques have been conceived in such a way to be specifically oriented towards given technological sectors.
The options, which activate the algorithms for the model identification and/or control optimisation, are selected at the SIMBUILDER level. The identification and the optimisation activities are performed by the SIM tool.

SIM PECULIARITIES
The SIM tool enables to assess the control and/or forecasting algorithm performance by means of simulated trials, which can be performed both by varying the plant and the environment parameters, that is by modifying the testing conditions, as well as by varying the control parameters, that is by performing an experimental control tuning.
When the related option has been selected, the SIM tool performs the identification of the model parameter set stated by the user. The identification can be performed both from input and output data recorded during trials carried out on the actual plant, as well as from input and output data obtained by simulating the plant using a fine model, more sophisticated than the one which has to be identified.
An identification algorithm is implemented which is specifically oriented toward the identification of models suited for the control design. The above algorithm has been originally developed by EICAS.
Similarly, when the related option has been selected, the SIM tool performs the optimisation of the control parameters set stated by the user. The optimisation is carried out by repeating the same simulated trial and varying the control parameters according to a numerical algorithm which optimises the cost function defined by the user.
The above algorithm, originally developed by EICAS, implements an approximation of the “conjugate gradient” method.

POST PECULIARITIES
The POST program enables to visualise the variables resulting from the simulation, which have been selected with the aid of the SIMBUILDER during the programming phase. The POST offers in a very friendly way all typical functions of a plotting program and moreover some other functions which are often very useful for analysing the result of a simulation test. Among the classical post-processing functions (like sum, product, division of variables, raising to a power, logarithm…), it is relevant to mention the availability of tools for:
· The statistical analysis of the behaviour of a variable in time, with calculation of average value and standard deviation.
· The Discrete Fourier transform
· The Auto and Mutual correlation
· The harmonic analysis and power spectrum
The POST program is also a useful tool for preparing technical reports. In fact it allows to capture images in the desired format for printing or being included in word processors. Such images are created starting from diagrams of variables on video, adding captions, arrows and notes according to the user wish.

Last updated: 2004-10-28