FHO_PPT-Vorlage_Wissenschaftliche

Model based design (MBD) –
a free tool-chain
Simon Mayr, (Gernot Grabmair)
Simon Mayr, [email protected], University of Applied Sciences Upper Austria , Austria
University of Applied Sciences Upper Austria
Projects
Projects dealing with Scilab/XCos
PROTOFRAME – Framework und frontend for semi-automated matching
of real and virtual prototypes
Work in progress
MOdoPS – MOdel based Design by OPen Source
Project finished
Project result: Scilab/XCos example library
Seite 2
Overview
Model based design (MBD)
Code generation from XCos
Example (cart and pendulum)
Conclusion
Seite 3
Model based design (MBD)
Definition
Mathematical and visual method applied in designing embedded software to
address problems associated with
Complex control
Signal processing
Communication systems
Seite 4
Model based design (MBD)
Applications
Common fields of application are:
Motion control applications
Industrial equipment
Aerospace applications
Automotive applications
…
Seite 5
Model based design (MBD)
Main steps of model based (controller) design
System modelling and basic model verification
Parameter identification and model verification
Control design and closed-loop simulation
Code generation and transfer to target
Seite 6
Model based design (MBD)
Main steps of model based (controller) design
Seite 7
Model based design (MBD)
Advantages
Faster and more cost-efficient development
Errors in system design can be located and corrected in early stage of the
project, when financial and time impacts of the system redesign are relatively
small
Extension and/or modification of an existing system is relatively easy
Seite 8
Model based design (MBD)
Common commercial tool-chains
Typical examples of commercial tools are:
Matlab/Simulink
Dymola
…
Advantages:
Advanced and well-proven software
Complete tool-chains
Disadvantages:
Quite expensive
Unsuitable for small and medium-sized companies
Seite 9
Model based design (MBD)
Free tool-chain
Scilab/XCos
Advantages:
Plant modeling
Control design & simulation
Disadvantages:
Code generation is not implemented
Solution:
Use an external application to generate code from XCos diagram
Seite 10
Code generator for Scilab/XCos
State of the art
Existing code generators for the outdated Scilab/Scicos:
RTAI [3]
Gene-Auto [4]
Scicos-FLEX [5]
Code generators for Scilab/XCos:
Project-P [6]
X2C from JKU-Linz (Upper Austria) [2]
Seite 11
Code generator for Scilab/XCos
Code generator X2C
The predecessor of X2C was developed more than 10 years ago at the JKUniversity Linz, Austria as a Simulink extension
X2C natively includes into XCos and can be simulated in parallel with plant
and controller
Seite 12
Code generator for Scilab/XCos
Code generator X2C
X2C-blocks are full featured XCos-blocks extended with an parameter editor
and the connection to the back-end for code generation
In XCos simulation the X2C-blocks are implementing exactly the code which
will run on the target
Model transformation and code generation is executed by a simple mouse
click. All non-X2C-blocks are ignored during this process.
Seite 13
Code generator for Scilab/XCos
Code generator X2C
The central tool is the so called „Communicator“. It‘s the interface between
simulation environment and target.
The Communicator features
Code generation
Change parameters in the model or in the communicator, and the
parameters on the target are updated instantly
Scope (software oscilloscope)
Seite 14
Code generator for Scilab/XCos
Communicator and scope
Seite 15
Code generator for Scilab/XCos
User defined X2C-blocks
It’s possible to generate user defined X2C blocks easily with the help of a
dedicated block generator
Inputs, outputs, control parameters and data types are specified by the user
This information is used to generate a code template automatically
The behavior of the block is included by the user
This blocks can be used for simulation and implementation on target
Seite 16
Code generator for Scilab/XCos
X2C-block generator
Seite 17
Code generator for Scilab/XCos
Code template and user code
Seite 18
Experiment
Cart and pendulum
System modeling
Plant simulation
Parameter identification (pendulum length)
Adaptive STC control
Code generation
Measurements
Seite 19
Plant modeling
Cart and pendulum
Plant modeling can be either done by using mathematical terms (e.g. ODEs)
or by using the Modelica-based Scilab-addon „Coselica“
Plant modeling is shown by the cart and pendulum example.
m1
d1
m2
d2
l2
... vehicle mass
… linear friction coefficient (cart)
… pendulum mass
… linear friction coefficient (pendulum)
… pendulum length
x … distance (cart)
v ... velocity (cart)
φ … angle (pendulum)
ω … angular velocity (pendulum)
l2, m2
Seite 20
Plant modeling
Cart and pendulum
Nonlinear system equations can be computed using the Lagrange formalism
with the vector of generalized coordinates  = [, ] and  =  .
Furthermore static friction FC is ignored, because it‘s compensated.
The linearized model (around  = [ ,  ,  ,  ] = [0, , 0,0] , k = 0,2,…)
can be written as
Seite 21
Plant modeling
Coselica & ODE
Seite 22
Parameter identification
Identifying the unknown (but constant) pendulum length l2
Assumption(s):
Pendulum friction is set to zero (d2 = 0)
The 4th line of the linearized model is used for identification
To get rid of the time derivatives, the system equation is transformed into the
laplace-domain
Seite 23
Parameter identification
We apply realizable stable filters F0(s) and F1(s) = sF0(s) with free coefficients
to the whole equation [1]
The inverse laplace transformation leads to one data line, linear in the
unknown parameter (* indicates the convolution operator in time-domain)
Estimation of the parameter using recursive least square algorithm
Seite 24
Adaptive STC control
Design a linear state control law parameterized in pendulum length l2
Seite 25
STC measurements
Seite 26
STC measurements
Discovery-board
Seite 27
STC measurements
Seite 28
Summary and outlook
Complete free (or low cost if hardware is included) tool-chain based on
Scilab/XCos
Ongoing development is targeted towards
efficient handling of vectorized signal lines in X2C
more block libraries
Industrial targets
adaption of the FMI (functional mockup interface) for model exchange
Thank you for your attention!
Seite 29
References
[1]
JJE. Slotine, W. Li, Applied nonlinear control, Prantice-Hall, 1991
[2]
X2C in Scilab/XCos, 2013, http://www.mechatronic-simulation.org
[3]
Roberto Bucher, et al., RTAI-Lab tutorial: Scilab, Comedi and real-time control,
2006
[4]
Ana-Elena Rugina, et al., Gene-Auto: Automatic Software Code Generation for
Real-Time Embedded Systems, DASIA 2008
[5]
Scicos-FLEX code generator, http://erika.tuxfamily.org/drupal/scilabscicos.html
[6]
Project-P code generator, http://www.open-do.org/projects/p/
Seite 30