Cat and mouse game tutorial builder socket

This tutorial will provide a more complex scenario, with a clear objective to accomplish for a robot, and that will show how to “close the loop” of sensors and actuators simulated in MORSE.

We will make a chase game, where a robot controlled by a human (the mouse) will be chased by a second robot (the cat) that is running software external to MORSE. This will be accomplished by the use of sensor and actuators configured properly.



Creating the scenario

We’ll use the Builder API to configure the robots in the scenario. First we will configure the mouse robot, which is a lot simpler.

  • Create a new ATRV robot, the mouse:

    from morse.builder import *
    mouse = ATRV()
    mouse.translate(x=1.0, z=0.2)
  • We will give it some special properties, so that it can be recognised by the semantic camera sensor: = True, Graspable = False, Label = "MOUSE")
  • Next we make it controllable by the keyboard, using the correct actuator. Also, we change the default speed, to make it more agile

    keyboard = Keyboard()

Now we’ll create the cat robot, with a sensor to detect the mouse, and an actuator to follow it.

  • Create another ATRV robot, the cat:

    cat = ATRV()
    cat.translate(x=-6.0, z=0.2)
  • Next add two semantic cameras to the robot. This will provide us with an easy way to follow our target. One is placed right and one left, to provide a fake stereo vision:

    semanticL = SemanticCamera()
    semanticL.translate(x=0.2, y=0.3, z=0.9)
    semanticR = SemanticCamera()
    semanticR.translate(x=0.2, y=-0.3, z=0.9)
  • Add also a v, omega actuator that will make the robot move:

    motion = MotionVW()
  • We configure these two components to use the sockets middleware:


And finally we complete the scene configuration:

  • We add data/environments/land-1/trees.blend environment:

    env = Environment('land-1/trees')
  • We setup the first person camera (CameraFP) and display the semanticL camera:

    env.set_camera_location([10.0, -10.0, 10.0])
    env.set_camera_rotation([1.0470, 0, 0.7854])

The last line indicates to MORSE that you want the images seen from the left camera to be displayed on the HUD screen, visible when you press v during the simulation. You can easily change it to display the view of the right camera.

The complete script can be found at: $MORSE_SRC/examples/tutorials/

Control program


This script uses pymorse, you need to have built MORSE with -DPYMORSE_SUPPORT=ON flag.

As a very simple example of how to use the data from a sensor to drive the robot, we’ll create a Python script to connect to MORSE and provide the “reasoning” of the cat robot.

The whole program can be found at: $MORSE_SRC/examples/clients/atrv/ Here we’ll explain the main parts of it:

  • The function is_mouse_visible will use the specified semantic camera to check if the mouse robot is anywhere in front:

    def is_mouse_visible(semantic_camera_stream):
        """ Read data from the semantic camera, and determine if a specific
        object is within the field of view of the robot """
        data = semantic_camera_stream.get()
        visible_objects = data['visible_objects']
        for visible_object in visible_objects:
            if visible_object['name'] == "MOUSE":
                return True
        return False
  • The main decision to move is made based on the information from the semantic cameras. There are four cases possible: The mouse can be seen by both cameras at once, only by the right, only by the left or by none of them. The cat‘s logic is very simple, it will move forward when the mouse is seen by both cameras, turn to the side of the only camera that sees the target or turn in place until it sees the target MOUSE.

    def main():
        """ Use the semantic cameras to locate the target and follow it """
        with Morse() as morse:
            semanticL =
            semanticR =
            motion =
            while True:
                mouse_seen_left = is_mouse_visible(semanticL)
                mouse_seen_right = is_mouse_visible(semanticR)
                if mouse_seen_left and mouse_seen_right:
                    v_w = {"v": 2, "w": 0}
                elif mouse_seen_left:
                    v_w = {"v": 1.5, "w": 1}
                elif mouse_seen_right:
                    v_w = {"v": 1.5, "w": -1}
                    v_w = {"v": 0, "w": -1}

Running the game

Run morse with the builder script to create the scenario. You will be able to control the mouse robot with the arrow keys on the keyboard:

$ cd MORSE_SRC/examples/tutorials
$ morse run

On the terminal you will get messages indicating the components, the available services and the datastream interfaces:

[    0.171] ------------------------------------
[    0.172] -        SIMULATION SUMMARY        -
[    0.172] ------------------------------------
[    0.172] Robots in the simulation:
[    0.172]     ROBOT: 'cat'
[    0.172]         - Component: 'cat.semanticR'
[    0.172]         - Component: 'cat.semanticL'
[    0.172]         - Component: 'cat.motion'
[    0.172]     ROBOT: 'mouse'
[    0.172]         - Component: 'mouse.keyboard'
[    0.172] Available services:
[    0.172]     - Interface morse.middleware.socket_request_manager.SocketRequestManager
[    0.173]         - communication: ['distance_and_view']
[    0.173]         - simulation: ['terminate', 'get_all_stream_ports', 'get_stream_port', 'activate', 'details', 'restore_dynamics', 'list_streams', 'quit', 'deactivate', 'list_robots', 'reset_objects', 'suspend_dynamics']
[    0.173] Modifiers in use:
[    0.173]     None
[    0.173]
[    0.173] Datastream interfaces configured:
[    0.173]     - 'morse.middleware.socket_datastream.Socket'

Then run the Python control script from another terminal. The cat will start moving and using the data from the semantic cameras to chase after the mouse:

$ python3

Note: The following consideration is deprecated but you may find it useful. As we use sockets for the introduced actuators and sensors, you can connect these ports using the telnet program on another terminal and you will seee the datastream of object visibility coming from the cameras. The socket port numbers are usually 60000+ (e.g. 60001 or 60002...):

$ telnet localhost 60001

For more information about sockets in MORSE, see the services tutorial<advanced_tutorials/request_tutorial>.

Going further

This example is very basic, but already provides a test of how the use of sensor data can help drive a robot. You can substitute the simple Python client that controls the cat for a more complex piece of software, implemented in other languages and middlewares. Here are some ideas of what you could do to improve the “intelligence” of the cat.

  • Use a single semantic camera and a Pose sensor to follow the mouse. You don’t really need two semantic cameras, since among the data one provides is the location of the detected object. Using that and the current position of the cat, it will be possible to chase, but you need to do some calculations to determine in which direction to turn.
  • Use other kinds of robots, like in the flying cat and mouse tutorial
  • Use a Laser Scanner to make the cat detect and avoid obstacles. This is more complex, since you have to handle a lot of data that is streamed by the Sick.
  • The target could hide behind an obstacle, so you could implement a strategy to move around the area searching for it.