fourier-grx 0.1.1rc3

Robotics Control System for the General Robotics (GR)

fourier-grx

Fourier GRX is a python package that provides a guidance for Fourier GR-X series robot.

Sturcture

src/fourier_grx

• control_system: control system for Fourier GR-X series robot. It is mainly the control the main running loop of the robot.
• robot: robot model definition, robot related algorithms, and predefined tasks for Fourier GR-X series robot.
• algorithm: algorithm for Fourier GR-X series robot. It is mainly the common used algorithm for robot.
• task: task for Fourier GR-X series robot. It is mainly the predefined task for robot.
• other_controller: other controller for Fourier GR-X series robot, including the communication of MPC-WBC controller for Fourier GR-X series robot.
• tools: tools for Fourier GR-X series robot. It is mainly the tools for the robot, including the calibration tools, the data processing tools, etc.
• webots: Webots code for Fourier GR-X series robot. It is mainly the simulation of the robot in webots.
• zenoh: zenoh for Fourier GR-X series robot. It is mainly the communication between the robot and the computer.
• sdk: sdk for Fourier GR-X series robot. It is mainly the communication between the robot and the computer.

fourier-core

fourier-core is the core package for Fourier GR-X series robot. It builds the basic structure for Fourier GR-X series robot or other robot. It is mainly the basic structure for the robot, including the robot model, the robot controller, the robot algorithm, etc.

tests

test code for Fourier GR-X series robot.

data

data for Fourier GR-X series robot:

• pictures: pictures of the GR-X series robot.
• sensor_offset: absolute encoder sensor offset value data for different robots.

webots

Webots environment for Fourier GR-X series robot.

demo

demo code for Fourier GR-X series robot.

• developer_mode: use developer_mode interface to test the demos.
• zenoh: use zenoh interface to test the demos.
• state_estimator: config files for state estimator of the robot.

---

Process to test and add new algorithm

Develop new algorithm directly in the fourier-grx, or add algorithm to the fourier-grx (legacy method)

Besides using zenoh interface to add the new algorithm, developers can also directly change the code of fourier-grx to add new algorithms. This is mainly for the developers who want to add the algorithm to the fourier-grx package, and develop built-in algorithms for the robot.

The main pipeline to develop the algorithm directly in the fourier-grx package is as following:

[Algorithm] -> [Task] -> [Robot] -> Test on the robot

1. Change directory to src/algorithm/ and create the algorithm file in the corresponding robot folder.

   • GR-1's algorithm should be in src/algorithm/gr1/
   • GR-2's algorithm should be in src/algorithm/gr2/
   • etc.

2. Define the algorithm name and write its logic code.

   • using the class definition to define the algorithm.
   • using the def run() interface to define the algorithm input and output.

# example code
class AlgorithmGRX:
    def __init__(self):
        super().__init__()

    def run(self,
            algorithm_input=None):
        algorithm_output = None

        return algorithm_output

3. Create and bind the algorithm with customized task in robot class file.

   • GR-1's task should be in src/task/gr1/
   • GR-2's task should be in src/task/gr2/
   • etc.

4. Add the task Enum value to TaskGRX, and use the enum value as the key. Then, bind the enum value with the task model and function.

# add the task enum value
class TaskGR1(Enum):
    TASK_CUSTOMIZED = 0x66666666


# define the task function
def task_customized_function(self):
    pass


# write the `add_task_customized` to add the task to the robot task manager
def add_task_customized(self):
    # imports
    from fourier_grx.algorithm.grx.fi_algorithm_customized import AlgorithmCustomized

    # create task algorithm models
    self.algorithm_customized_model = AlgorithmCustomized()
    self.task_customized_function = task_customized_function

    self.tasks.extend(
        [
            TaskGRX.TASK_CUSTOMIZED,  # task key
        ]
    )
    self.task_algorithm_models.update(
        {
            TaskGRX.TASK_CUSTOMIZED: self.algorithm_customized_model,  # task algorithm model
        }
    )
    self.task_functions.update(
        {
            TaskGRX.TASK_CUSTOMIZED: self.task_customized_function,  # task function
        }
    )

5. Finally, call add_task_customized() in the update_tasks() function to add the task to the robot task manager.
   • GR-1's task should be in src/task/gr1/fi_task_gr1.py
     • GR-1 T1's task should be in src/task/gr1/fi_task_gr1_t1.py
     • GR-1 T2's task should be in src/task/gr1/fi_task_gr1_t2.py
     • etc.
   • GR-2's task should be in src/task/gr2/fi_task_gr2.py
   • etc.

def update_tasks(self):
    from fourier_grx.task.grx.fi_task_customized import add_task_customized

    add_task_customized(self)

6. Till now, your algorithm is added to the robot, and you can test it on the real robot.
   • You can run python main.py --config=config/config_xxx.yaml to start the program.
   • Then, use the R1 button of the joystick to switch to the customized task.
   • Later, you can use the R2 button of the joystick to confirm the task. The robot will then run the customized task.

Develop new algorithm based on zenoh interface (quick demo -> feature method)

Please check the demos in the demo/zenoh folder. It is suggested to read the demo_robot_client.py to understand how to get state of the robot and set the control command to the robot.

1. The main pipeline to develop the demo based on zenoh interface is as following:

#################################################
# In command line
# 1. run run_server.py to start the robot server.
# python run_server.py
#################################################

#################################################
# In demo code file
# 1. call RobotClient to connect to the robot
client = RobotClient(FREQUENCY)

# 2. set enable to True to enable the robot
client.set_enable(True)

# 3. get the state of the robot
joint_measured_position_urdf_deg = client.states["joint"]["position"].copy()
joint_measured_velocity_urdf_deg = client.states["joint"]["velocity"].copy()


# 4. write and run your algorithm
class Algorithm:
   def __init__(self):
      self.algorithm_input = None
      self.algorithm_output = None

   def run(algorithm_input=None):
      self.algorithm_input = algorithm_input
      # algorithm calculation...
      return self.algorithm_output


algorithm = Algorithm()
ALGORITHM_OUTPUT = algorithm.run(algorithm_input=ALGORITHM_INPUT)

# 5. set the control command to the robot
client.move_joints(group=ControlGroup.ALL, positions=joint_target_position_deg)
#################################################

Notice: It is suggested to write all algorithms with class definition, and provide only def run() interface for the exchange of algorithm input and output. This will make the transfer of algorithm from demo to feature more convenient.

2. After the demo has been tested independently, you can create a task in the fourier-grx package to add the algorithm to the robot.

   • Change directory to fourier-grx/src/task and create the task file in the corresponding robot folder.
     • GR-1's task should be in src/task/gr1/
     • GR-2's task should be in src/task/gr2/
     • etc.
   • Use fi_task_demo.py as a reference to create the task class, and attach the algorithm to the task.
   • Add add_task_demo() to the task file to enable adding the task to the robot task manager.

3. Finally, call add_task_demo() in the update_tasks() function to add the task to the robot task manager in fi_zenoh_client.py.

4. Till now, your algorithm is added to the robot, and you can test it on the real robot.

   • You can run python main_zenoh.py --config=config/config_xxx.yaml to start the program.
   • Then, use the R1 button of the joystick to switch to the customized task.
   • Later, you can use the R2 button of the joystick to confirm the task. The robot will then run the customized task.