You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 13 Next »

This example is a guideline on how to write a Matlab script that, together with the Middleware, communicates with a vehicle. This tutorial is based on our platoon example (which is just a series of vehicles following a preset path without much logic behind that behavior), which you can find in

https://github.com/embedded-software-laboratory/cpm_lab/tree/master/high_level_controller/examples/matlab/platoon

More specifically, we well take a look at main_vehicle_ids.m.

As you can see in the following sections, this script uses another script, called init_script.m.

Init script

The init script is similar (in its idea) to the cpm library for your C++ programs: It is supposed to take care of redundant work for you. In this case, it sets up all relevant DDS participants for the communication within the network. It also loads the required IDL files, loads XML files for QoS settings and sets the variable for the stop signal, which will be mentioned later on. It currently requires a given folder structure, which is automatically set up on the NUCs and should also be part of the software repository's structure - in an ideal case, you do not need to worry about that, as long as you import the init script from its original place (TODO: although other rules may apply for deploying remotely).

You can find it here:

https://github.com/embedded-software-laboratory/cpm_lab/blob/master/high_level_controller/examples/matlab/init_script.m

Working with RTI DDS in Matlab

Please consult RTI DDS and Matlab to get an idea of how to work with DDS in Matlab. The Middleware-specific IDL file is only relevant for the communication between Middleware and Matlab script and is thus not part of the cpm library.

Your Matlab script is not supposed to directly communicate with the vehicle or to perform timing operations with the LCC. This is the Middleware's task, so your task is just to properly communicate with the Middleware (if you only want to control the vehicle). 

The Middleware will wake your script up regularly to perform a computation with the currently given values, and you just need to respond in a proper way. This allows you to focus on problem solving instead of timing and communication.

Setup

The structure of your code

Your function head may differ, but you must use varargin as your last parameter to pass vehicle IDs to your script. This allows you to define, for your script, which vehicle(s) it should be responsible for. Any previous parameters you define are your own 'custom' parameters and need to be specified as additional parameters in the Lab Control Centers' UI before starting your script (if you wish to start it using the UI).

function main(varargin)
...
vehicle_ids = varargin;

In other words: If your function head looks like this

function main(some, params, varargin)

then you need to pass some, params as additional parameters in the LCC's UI (ignore the script path in the screenshot):

What your HLC scripts need to include

Import the init script

Set the Matlab domain ID to 1, go to the right directory for the init script, load the script, store the vehicle IDs. In the case the reader reads all vehicle states sent within the network aggregated by the Middleware, a waitset is set to wait upon taking data from the reader if no data is yet available. The timeout is set to 10 seconds.

function main_vehicle_ids(varargin)
	% Set the matlab domain ID for communicating with the middleware (which is always 1)
    matlabDomainID = 1;

	% Clear command window, store path of current script, go to that path
    clc
    script_directoy = fileparts([mfilename('fullpath') '.m']);
    cd(script_directoy)

    % Set the path of the init script relative to the path of your script (or absolute), assert that it exists, add it to the path so that it can be found and than call it
	% Then go back to the script directory
    init_script_path = fullfile('../', '/init_script.m');
    assert(isfile(init_script_path), 'Missing file "%s".', init_script_path);
    addpath(fileparts(init_script_path));
    [matlabParticipant, stateReader, trajectoryWriter, systemTriggerReader, readyStatusWriter, trigger_stop] = init_script(matlabDomainID);
    cd(script_directoy)

	% Remember the set vehicle IDs in a new variable
    vehicle_ids = varargin;

    %% Use a waitset for the reader of vehicle states, which is later used to indicate if a new computation should be started
	% The waitset makes sure that take() or read() from the reader's storage does not return if no new data is available (until the timeout is reached, here 10 seconds
	% so that one can still check for a stop signal regularly)
    stateReader.WaitSet = true;
    stateReader.WaitSetTimeout = 10;

    %% Do not display figures
    set(0,'DefaultFigureVisible','off');

Tell the Middleware that your script is ready to operate

You must send a ReadySignal message after initialization - only the ID string matters, which must be of the form "hlc_" + vehicle_id, where the latter is the ID of the vehicle the HLC and thus your script is responsible for:

% (Assuming that vehicle_ids contains the IDs your script is responsible for)
% Send first ready signal to the Middleware to indicate that your program is ready to operate
% The signal needs to be sent for all assigned vehicle ids separately (usually, one script on one NUC only manages one of the IDs)
% Also works for simulated time - period etc are set in Middleware, so you can ignore the timestamp field
for i = 1 : length(vehicle_ids)
	% Create a new msg of type ReadyStatus (idl)
    ready_msg = ReadyStatus;

	% Set the values of ready_msg
    ready_msg.source_id = strcat('hlc_', num2str(vehicle_ids{i}));
    ready_stamp = TimeStamp;
    ready_stamp.nanoseconds = uint64(0);
    ready_msg.next_start_stamp = ready_stamp;

	% Send the ready msg
    readyStatusWriter.write(ready_msg);
end

You use the ReadyStatus type here, which was defined in one of the IDL files that was imported for you using the init script. You can see the values of this type in the IDL file:

https://github.com/embedded-software-laboratory/cpm_lab/blob/master/cpm_lib/dds_idl/ReadyStatus.idl

This message tells the Middleware that your script is now ready to operate and to receive its data. Thus, at this point, your reader for vehicle states (here stateReader) should be initialized so that it can receive the data.

Consider the timing signals

This is very important. There are two signals you must consider: Start and stop signals. You have to check regularly for the stop signal, and once for the start signal:

  • Start signal: This indicates that your script is supposed to calculate a new e.g. trajectory for the vehicles it is responsible for. Start with calculation immediately and send the results back as soon as possible. The Middleware handles the rest.
  • Stop signal: trigger_stop = uint64(18446744073709551615) was already set by the init script. It is the highest number representable with a uint64_t type. This number indicates that the simulation was stopped, and thus your script should immediately terminate.
  • All other numbers are irrelevant. From here on, new messages received by the reader of VehicleStateList, stateReader, are interpreted as start signals for less redundancy.

Waiting for the initial start signal may look like this:

% Wait for start signal / stop if a stop signal was received
got_stop = false;
got_start = false;
while(true)
	% Read the newest system trigger message, if one exists
    trigger = SystemTrigger;
    sampleCount = 0;
    [trigger, status, sampleCount, sampleInfo] = systemTriggerReader.take(trigger);

	% Go through all received messages since the last check, if they exist (ssampleCount gives the number of read messages)
	% Any received message would indicate "start", but you have to check if a stop signal was received as well
    while sampleCount > 0
		% Check for stop signal
        if trigger.next_start().nanoseconds() == trigger_stop
            got_stop = true;
        elseif trigger.next_start().nanoseconds() >= 0
            got_start = true;
        end

		% Read the next sample and continue until sampleCount is zero again (all samples have been read then)
        [trigger, status, sampleCount, sampleInfo] = systemTriggerReader.take(trigger);
    end

	% Stop the loop if a signal was received, then proceed with handling a stop signal (stop the program) or a start signal (continue)
    if got_stop | got_start
        break;
    end
end

You need to repeat this procedure to check for the stop signal every time before you wait for the next start indicator sent by the Middleware (which is indicated by new samples in VehicleStateList). The loop, that includes calculating new data as well, may look like this:

% Again: Go through all samples and stop the program if you received a stop signal
% break_while is the name used in the sample program (as it breaks the enclosing while-loop of the whole program), but you could also name it e.g. stop_received
trigger = SystemTrigger;
sampleCount = 0;
[trigger, status, sampleCount, sampleInfo] = systemTriggerReader.take(trigger);

break_while = false;

while sampleCount > 0
    current_time = trigger.next_start().nanoseconds();
    if current_time == trigger_stop
        break_while = true;
    end

    [trigger, status, sampleCount, sampleInfo] = systemTriggerReader.take(trigger);
end

if break_while
    ...

Receive information about your vehicle

Information about your vehicle are contained in the VehicleStateList messages, which include the current states and observations of all vehicle as well as the current time. This signal is supposed to be the start signal for the HLC, so computation should start using this data directly after the message was received. 

These signals, which, in this example, you can obtain using the stateReader, contain two fields which are vital to the correct usage of your script.

  1. The messages contain current information about the whereabouts and settings of your vehicle as well as any other vehicle in the simulation. Use this data for planning.
  2. The messages contain timing information. Use these when you send commands to your vehicle - they contain the current time you set for the messages.

This signal has a third purpose as well. It shows the script when to start the computation, similar to a timing signal. The desired computation cycle is the following:

  • Wait for a new message
  • After receiving a message, start your computation
  • Then send the new command(s) to the vehicle
  • Remove messages that were received during computation, as they are old and potentially unusable, and wait for the next message
  • ...

Make sure to ignore signals received during computation, if that takes longer than a single period.

...
% Read the current vehicle information from the according reader, if such data is available
% Due to the set waitset, you wait for up to 10 seconds at take(...) if no data is available
sample = VehicleStateList;
status = 0;
stateSampleCount = 0;
sampleInfo = DDS.SampleInfo;
[sample, status, stateSampleCount, sampleInfo] = stateReader.take(sample);

% In this example, messages that were received during computation that should be ignored are not thrown away before waiting for new data
% If your script is faster than the period with which it is called, this is not problematic, in any other case you will probably end up using wrong timing and outdated vehicle data
% Thus, you should get rid of old data at the end of any iteration ('clear' stateReader)

% Check if any new message was received, then proceed with handling the data, computing your solution and finally sending back a command to the vehicle(s) your script controls
if stateSampleCount > 0 
    ...

Send commands to your vehicle

You need to send vehicle command messages as a result of your computation including the vehicle ID to the Middleware, which propagates these to the vehicle. The implementation of the computation of e.g. the vehicle's trajectory is not explained here and depends on your task. You are only given an example of how to send a simple trajectory here.

In the script which we were looking at until now, we use another script within the same folder (only that is allowed if you want to use several scripts) to compute the next trajectory segment:

% Call the programs which calculate the trajectories of all HLCs, then send the results back to the vehicles
if (size(vehicle_ids) > 0)
    for i = 1 : length(vehicle_ids)
        msg_leader = leader(vehicle_ids{i}, sample.t_now);

		% Send resulting trajectory to the vehicle
        trajectoryWriter.write(msg_leader);
    end
end

In the following, you see an example of how to set up a trajectory message, that can then be sent using the trajectoryWriter:

%Create msg of type trajectory
trajectory = VehicleCommandTrajectory;

% Set the vehicle ID
trajectory.vehicle_id = uint8(vehicle_id);

% In this example, we only send one trajectory point (which is not sufficient for interpolation)
trajectory_points = [];
point1 = TrajectoryPoint;

% This trajectory point should be considered in the future, so add some nanoseconds to the current time, then set the time stamp for the trajectory point
% t_eval here is the time that we got from the middleware in the VehicleStateList message - you are not supposed to use your own clock / timer (else you would have trouble working with simulated time)
time = t_eval + 400000000;
stamp = TimeStamp;
stamp.nanoseconds = uint64(time);
point1.t = stamp;

% Set other trajectory values - position and velocity in 2D coordinates
point1.px = trajectory_point(1);
point1.py = trajectory_point(2);
point1.vx = trajectory_point(3);
point1.vy = trajectory_point(4);

% Append to the list of all trajectory points
trajectory_points = [trajectory_points [point1]];
trajectory.trajectory_points = trajectory_points;

% Send the trajectory to the vehicle
trajectoryWriter.write(trajectory);

Bash script

If you want to start your script using a bash script, you can do it like this:

/opt/MATLAB/R2019a/bin/matlab -logfile matlab.log -sd $script_dir -batch "$script_name(1, ${vehicle_id})"

Usually, you would want to start it using the LCC instead, which takes care of these things for you.



  • No labels