/ORTD

Permalink
Switch branches/tags
Nothing to show
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
Raw Blame History
365 lines (264 sloc) 14.2 KB
//
// This file is part of OpenRTDynamics.sf.net, the Real-Time Dynamics Framework
//
// A demonstration for automating experiments using the macro ld_AutoOnlineExch_dev2
// --------------------------------------------------------------------------------
//
// Rev 2. 10.8.15
//
// NOTE: At least Revision 591 of ORTD is required to run this 2nd revision of this template.
//
//
// This example shows how to automate sequences of experiments including
// their evaluation and/or the ongoing design of control systems. Normal Scilab
// code can be used to automatically perform calibration / controller design
// procedures, while the overall control system is running. Concepts like
// embedded Scilab and on-line replaceable schematics are involved.
//
//
// Definitions of terms:
//
// This example involves the execution of Scilab code to define schematics as
// well as to perform calculations, while the control
// system is running by using Scilab embedded into the ORTD interpreter
// (ld_scilab4). If comments in the code state that a certain part of Scilab
// commands are executed "On-line" or "at runtime" it means these commands will be
// run in the embedded Scilab instance, when the ortd interpreter is running.
// Parts marked by the term "Off-line" are executed when this Scilab-Script
// (RTmain.sce) is executed prior to running the ortd-interpreter.
//
// The schematic that is responsible for performing e.g. calibration
// or implementing control systems is called "experiment controller".
// Such schematics are defined using the function "ExperimentCntrl" in
// this example. This function may run Off-line (to define an initial
// experiment controller) as well as On-line to define experiment controllers
// during runtime. The execution of this function is triggered each time
// an experiment controller decides to finish its execution, which is indicated
// by setting the return-signal "finished" to one. In a next step, typically,
// a replacement for the current experiment controller is defined that is automatically
// loaded.
//
// By implementing "ExperimentCntrl" using a state machine, complex experiment control
// logic can be implemented. Herein, each state may be used to define new experiment
// controllers potentially based on results (e.g. collected data) of previously
// active experiment controllers.
//
//
// In this example the following steps are performed
//
// 1) An initial, dummy schematic for the experiment controller is
// active that finishes executing after 10 execution steps.
// 2) The function "ExperimentCntrl" is called using the embedded
// Scilab instance. The state "init" is active and used to define
// an experiment controller that is supposed to collect measured
// data.
// 3) The experiment controller is loaded and executed. Collection of data
// is progressing until NcalibSamples=20 samples are collected.
// 4) Again, he function "ExperimentCntrl" is called using the embedded
// Scilab instance to define a next experiment controller. Now, the state "control"
// is active. The previously collected data is used obtain some parameters
// (e.g. a mean value) using normal Scilab commands to design a control system.
// 5) The control system is loaded and executed.
// 6) after some time, the control system decides to perform a re-calibration;
// Step 2) and the ongoing logic is performed again.
//
//
// Please note: All error messages occurring when embedded Scilab code is executed
// go to stderr of the ortd interpreter. E.g. append something like
// ortd <...> 2> Errors.log
//
// The name of the program
ProgramName = 'RTmain'; // must be the filename without .sce
thispath = get_absolute_file_path(ProgramName+'.sce');
cd(thispath);
function [sim] = AutoCalibration(sim, Signal)
function [sim, finished, outlist, userdata] = ExperimentCntrl(sim, ev, inlist, userdata, CalledOnline)
// Define parameters. They must be defined once again at this place, because this will also be called at
// runtime.
Data_vecsize = 2; NcalibSamples=20;
if CalledOnline == %t then
// The contents of this part will be compiled on-line, while the control
// system is running. The aim is to generate a new compiled schematic for
// the experiment.
// Please note: Since this code is only executed on-line, most potential errors
// occuring in this part become only visible during runtime.
printf("Compiling a new control system\n");
if userdata.isInitialised == %f then
//
// State variables can be initialise at this place
//
userdata.Acounter = 0;
userdata.State = "calibration";
userdata.isInitialised = %t; // prevent from initialising the variables once again
end
//
// Example for a state update: increase the counter
//
userdata.Acounter = userdata.Acounter + 1;
// Build an info-string
SchematicInfo = "On-line compiled in iteration #" + string(userdata.Acounter);
//
// Define a new experiment controller schematic depending on the currently active state
//
[sim, zero] = ld_const(sim, ev, 0);
[sim, one] = ld_const(sim, 0, 1);
// default output (dummy)
outlist=list(zero);
//
// Here a state-machine is implemented that may be used to implement some automation
// logic that is executed during runtime using the embedded Scilab interpreter.
// In this example, a calibration run succeeded by the design/compilation/execution
// of a control-system is implemented. The schematics defined in each state are loaded
// at runtime.
//
select userdata.State
case "calibration" // define a controller to perform a calibration experiment
Signal = inlist(1);
[sim] = ld_printf(sim, 0, Signal, "Calibration active; collecting data: ", 2);
// Store the sensor data into a shared memory.
[sim, Data_vecsize__] = ld_const(sim, ev, Data_vecsize);
[sim, zero] = ld_const(sim, ev, 0);
[sim, writeI] = ld_counter(sim, 0, count=Data_vecsize__, reset=zero, resetto=zero, initial=1);
[sim] = ld_write_global_memory(sim, 0, data=Signal, index=writeI, ...
ident_str="CalibrationData", datatype=ORTD.DATATYPE_FLOAT, ...
ElementsToWrite=Data_vecsize);
// Wait until a number of time steps has passed, then notify that
// the experiment has finished by setting "finished" to 1.
[sim, finished] = ld_steps2(sim, ev, activation_simsteps=NcalibSamples, values=[0,1] );
[sim, out] = ld_const(sim, ev, 0);
outlist=list(out);
// chose the next state to enter when the calibration experiment controller has finished
userdata.State = "control";
case "control" // design a controller based on the parameters obtained during the calibration
// Use the data collected during the experiment that was defined by state "calibration".
// Data comming out of userdata.InputData is the output ToScilab of "PreScilabRun".
data = userdata.InputData;
A=matrix( userdata.InputData , Data_vecsize, NcalibSamples )';
printf("Got the following data for iteration %d:\n", userdata.Acounter); disp(A);
// Perform some calibration tasks...
Signal1 = A(:,1); mean_S1 = mean(Signal1); stdev_S1 = stdev(Signal1);
// No real control system at this place -- just one printf to show the current parameters.
[sim] = ld_printf(sim, 0, zero, "The control system was compiled and is active now. Parameters are: mean_S1="+string(mean_S1)+" stdev_S1="+string(stdev_S1), 1);
// Do something useful in this demo: subtract to obtained mean values from the Signal
// S1
Signal = inlist(1); [sim, S1] = ld_demux(sim, 0, 2, Signal);
[sim, S1_minus_ofs] = ld_add_ofs(sim, 0, S1(1), -mean_S1);
[sim] = ld_printf(sim, 0, S1_minus_ofs, "The offset compensated input signal:" , 1);
// Wait until a number of time steps has passed, then notify that
// the experiment (control system in this case) has finished.
[sim, finished] = ld_steps2(sim, 0, activation_simsteps=50-1, values=[0,1] );
[sim, MaxCount] = ld_const(sim, 0, 50);
[sim, Counter] = ld_modcounter(sim, 0, in=one, initial_count=0, mod=50+1);
[sim, CountDown] = ld_add(sim, 0, list(Counter, MaxCount), [ -1, 1 ] );
[sim] = ld_printf(sim, 0, CountDown, "Recalibration after time-steps:" , 1);
// next state: recalibrate by going to "init" again after "finished" is set to 1
userdata.State = "calibration";
end
end // CalledOnline == %t
// When RTmain.sce is executed, this part will be run. It may be used to define an initial experiment in advance to
// the execution of the whole control system.
if CalledOnline == %f then
SchematicInfo = "Off-line compiled";
// default output (dummy)
[sim, out] = ld_const(sim, 0, 0);
outlist=list(out);
[sim, finished] = ld_steps2(sim, 0, activation_simsteps=10, values=[0,1] );
end
endfunction
function [sim, outlist, HoldState, userdata] = whileComputing_example(sim, ev, inlist, CalibrationReturnVal, computation_finished, par);
[sim, HoldState] = ld_const(sim, 0, 0);
[sim] = ld_printf(sim, 0, HoldState, "calculating ... " , 1);
// While the computation is running this is called regularly
[sim, out] = ld_const(sim, ev, 0);
outlist=list(out);
endfunction
function [sim, ToScilab, userdata] = PreScilabRun(sim, ev, par)
userdata = par.userdata;
// get the stored sensor data
[sim, readI] = ld_const(sim, 0, 1); // start at index 1
[sim, ToScilab] = ld_read_global_memory(sim, 0, index=readI, ident_str="CalibrationData", ...
datatype=ORTD.DATATYPE_FLOAT, ...
ElementsToRead=NcalibSamples*Data_vecsize);
endfunction
Data_vecsize = 2;
NcalibSamples = 20;
// initialise a global memory for storing the sensor-data for the calibration
[sim] = ld_global_memory(sim, ev, ident_str="CalibrationData", ...
datatype=ORTD.DATATYPE_FLOAT, len=NcalibSamples*Data_vecsize, ...
initial_data=[zeros(NcalibSamples*Data_vecsize,1)], ...
visibility='global', useMutex=1);
// Start the experiment
ThreadPrioStruct.prio1=ORTD.ORTD_RT_NORMALTASK;
ThreadPrioStruct.prio2=0, ThreadPrioStruct.cpu = -1;
insizes=[2]; outsizes=[1];
intypes=[ORTD.DATATYPE_FLOAT]; outtypes=[ORTD.DATATYPE_FLOAT];
CallbackFns.experiment = ExperimentCntrl;
CallbackFns.whileComputing = whileComputing_example;
CallbackFns.PreScilabRun = PreScilabRun;
// Please note ident_str must be unique.
userdata = [];
param.scilab_path = "BUILDIN_PATH";
[sim, finished, outlist, userdata] = ld_AutoOnlineExch_dev2(sim, 0, inlist=list(Signal), ...
insizes, outsizes, intypes, outtypes, ...
ThreadPrioStruct, CallbackFns, ident_str="AutoCalibDemo", userdata, param);
// [sim] = ld_printf(sim, 0, finished, "State ", 1);
endfunction
// The main real-time thread
function [sim, outlist, userdata] = Thread_MainRT(sim, inlist, userdata)
// This will become a main-loop running in a thread
[sim, Tpause] = ld_const(sim, ev, 1/4); // The sampling time that is constant at 4 Hz in this example
[sim, out] = ld_ClockSync(sim, ev, in=Tpause); // synchronise this simulation
// print the time interval
// [sim] = ld_printf(sim, ev, Tpause, "Time interval [s]", 1);
// save the absolute time into a file
[sim, time] = ld_clock(sim, ev);
// [sim] = ld_savefile(sim, ev, fname="AbsoluteTime.dat", source=time, vlen=1);
//
// Add your own control system here
//
rand('seed',0);
[sim,S1] = ld_play_simple(sim, 0, r=rand(1000,1) ); // some noise signal
[sim,S2] = ld_play_simple(sim, 0, r=1+sin(1:0.1:100) );
[sim, Signal] = ld_mux(sim, 0, 2, list(S1, S2) );
[sim] = AutoCalibration(sim, Signal);
outlist = list();
endfunction
// This is the main top level schematic
function [sim, outlist] = schematic_fn(sim, inlist)
//
// Create a thread that runs the control system
//
ThreadPrioStruct.prio1=ORTD.ORTD_RT_NORMALTASK; // or ORTD.ORTD_RT_REALTIMETASK
ThreadPrioStruct.prio2=0; // for ORTD.ORTD_RT_REALTIMETASK: 1-99 as described in man sched_setscheduler
// for ORTD.ORTD_RT_NORMALTASK this is the nice-value (higher value means less priority)
ThreadPrioStruct.cpu = -1; // The CPU on which the thread will run; -1 dynamically assigns to a CPU,
// counting of the CPUs starts at 0
[sim, StartThread] = ld_initimpuls(sim, ev); // triggers your computation only once
[sim, outlist, computation_finished] = ld_async_simulation(sim, ev, ...
inlist=list(), ...
insizes=[], outsizes=[], ...
intypes=[], outtypes=[], ...
nested_fn = Thread_MainRT, ...
TriggerSignal=StartThread, name="MainRealtimeThread", ...
ThreadPrioStruct, userdata=list() );
// output of schematic (empty)
outlist = list();
endfunction
//
// Set-up (no detailed understanding necessary)
//
thispath = get_absolute_file_path(ProgramName+'.sce');
cd(thispath);
z = poly(0,'z');
// defile ev
ev = [0]; // main event
// set-up schematic by calling the user defined function "schematic_fn"
insizes = []; outsizes=[];
[sim_container_irpar, sim]=libdyn_setup_schematic(schematic_fn, insizes, outsizes);
// pack the simulation into a irpar container
parlist = new_irparam_set();
parlist = new_irparam_container(parlist, sim_container_irpar, 901); // pack simulations into irpar container with id = 901
par = combine_irparam(parlist); // complete irparam set
save_irparam(par, ProgramName+'.ipar', ProgramName+'.rpar'); // Save the schematic to disk
// clear
par.ipar = []; par.rpar = [];