After you checkou the repository, ource your ROS environment and make sure the additional dependencies such as the eclipse paho mqtt client library is installed.
source /opt/ro/noetic/setup.bash
pip install paho-mqtt celery requestsmuto agent requires network connectivity to the muto Twins (ditto) and MQTT servers. The address for the muto sandbox is:
- The twin servers and API: http://sandbox.composiv.ai
- The mqtt server: mqtt://sandbox.composiv.ai:1883
The muto agent must be launched and the target device and the devices should have access to all the packages required to launch the stacks. Fo these purposes you can either choose to run a containerizedf version of muto or launch it direct from the devices after all the packages (such as the muto learning modules) are build and installed on the AV.
source devel/setup.bash
roslaunch muto_agent agent.launchwhere agent.launch is:
<?xml version="1.0"?>
<launch>
<node pkg="muto_agent" name="muto_agent" type="muto_agent.py">
<rosparam command="load" file="$(find muto_agent)/params.yaml" />
</node>
</launch>and params.yaml is:
muto:
nav_topic: /nav
type: simulator
twin_url: "http://ditto:ditto@sandbox.composiv.ai"
mqtt:
host: sandbox.composiv.ai
port: 1883
keep_alive: 60
user: none
password: none
thing:
namespace: ai.composiv.sandbox.f1tenth
anonymous: False # Use this for automatically generated id (uuid based)
# if anonymous is True or anynoymous param is missing, name/id will be auto generated
name: simulator-monster-01There are two definition types for thins on the wtin server:
- Stack - The model os the muto adaptive software stack that can be running on an AV
- Thing definition id for stack is: ai.composiv.sandbox.f1tenth:Stack:1.0.0
- Vehicle - The device twin for the AV, where parameters, telemetry data and stack associations are managed
- Thing definition id for vehicle is: ai.composiv.sandbox.f1tenth:TestCar:1.0.0
Examples of these definitions can be sampled by sending http rest requests to the sandbox server http://sandbox.composiv.ai. Muto Twin Server
curl 'http://sandbox.composiv.ai/api/2/search/things?filter=eq(definition,"ai.composiv.sandbox.f1tenth:Stack:1.0.0")'returns:
{
"name": "Muto Learning Simulator with Gap Follwer",
"context": "eteration_office",
"stackId": "ai.composiv.sandbox.f1tenth:composir_simulator_gf.launch",
"stack": [
{
"thingId": "ai.composiv.sandbox.f1tenth:composiv_simulator.launch"
}
],
"node": [
{
"name": "cass_gap_follower",
"pkg": "cass_gap_follower",
"exec": "cass_gap_follower",
"param": [
{ "from": "$(find cass_gap_follower)/params.yaml" }
]
}
]
}This is a stack with a single node, "cass_gap_follower". However, it includes another stack (with many other nodes and parameters) that it requires with a stackId reference ai.composiv.sandbox.f1tenth:muto_simulator.launch. The elements of the stack model resembles a ROS launch XML, so it should be fairly straightforward to understand if you have experience with it.
New stack can be easily stored on the sandbox server using the things API (put). See the ditto documentation for many examples [https://www.eclipse.org/ditto/intro-overview.html]. For example, to add a new stack to the repository we can use the thing PUT api as follows:
$ curl -X PUT -H "Content-Type: application/json" -d '
{
"name": "Muto Learning Simulator with Gap Follower",
"context": "eteration_office",
"stackId": "ai.composiv.sandbox.f1tenth:composir_simulator_gf.launch",
"stack": [
{
"thingId": "ai.composiv.sandbox.f1tenth:composiv_simulator.launch"
}
],
"node": [
{
"name": "cass_gap_follower",
"pkg": "cass_gap_follower",
"exec": "cass_gap_follower",
"param": [
{ "from": "$(find cass_gap_follower)/params.yaml" }
]
}
]
}
' http://sandbox.composiv.ai/api/2/things/ai.composiv.sandbox.f1tenth:composiv_simulator_gf.launch
We can use the TWINS to directly communicating commands to the vehicle itself. The twin server supports special mqtt channels for these purposes called twin and live channels. For example the following command can be published to the sandbox MQTT server to activate a stack on a car. Each vehicle has its dedicated twin and live channels:
topic: ai.composiv.sandbox.f1tenth:simulator-monster-01/stack/commands/activepayload: {
"name": "Muto Learning Simulator with Gap Follwer",
"context": "eteration_office",
"stackId": "ai.composiv.sandbox.f1tenth:composiv_simulator_gf.launch",
"stack": [
{
"thingId": "ai.composiv.sandbox.f1tenth:composiv_simulator.launch"
}
],
"node": [
{
"name": "cass_gap_follower",
"pkg": "cass_gap_follower",
"exec": "cass_gap_follower",
"param": [
{ "from": "$(find cass_gap_follower)/params.yaml" }
]
}
]
}You can use any open-source mqtt client to issue these commands and monitor various muto twin messages MQTTX. Another option is the mosquitto_pub, which is a simple MQTT version 5/3.1.1 client that will publish a single message on a topic and exit. You can publish the message described above using the commandline:
mosquitto_pub -d -h sandbox.composiv.ai -p 1883 -t "ai.composiv.sandbox.f1tenth:simulator-monster-01/stack/commands/active" -m '{"name":"Muto Learning Simulator with Gap Follwer","context":"eteration_office","stackId":"ai.composiv.sandbox.f1tenth:muto_simulator_gf.launch","stack":[{"thingId":"ai.composiv.sandbox.f1tenth:composiv_simulator.launch"}],"node":[{"name":"cass_gap_follower","pkg":"cass_gap_follower","exec":"cass_gap_follower","param":[{"from":"$(find cass_gap_follower)/params.yaml"}]}]}'