API Reference

The deepclaw library API consists of following parts:

  • Hardware Driver API

    • Arm

    • Hand

    • Camera

  • Modules Pool API

    • Segmentation

    • Recognition

    • Grasp Planning

    • Motion Planning

  • utils API

Hardware Driver API

The Hardware Driver API is used for controling the Hardware.

Arm

UR10e

class deepclaw.driver.arms.UR10eController(robot_configuration_file_path)

  • parameters[in]: robot_configuration_file_path, the robot configuration file, and the file format is yaml.

  • return: an instance of UR10e controller class

Methods

go_home()     
Description: move to pre-defined joints, the home joints is defined in robot configuration file.

move_j(joints_angle, velocity=None, acceleration=None, solution_space='Joint')   
Description: go to the target joints positions    
param[in]:
- joints_angle: target joint positions of each joints [rad];    
- velocity: joint acceleration of leading axis [rad/s];   
- accelerate: joint speed of leading axis [rad/s^2];   
- solution_space: move style, 'Joint' means it linear in joint-space(inverse kinematics is used to calculate the corresponding joints), and 'Space' means linear in tool-space   
return: bool, reaching target or not

move_p(position, velocity=None, acceleration=None,solution_space='Joint')   
Description: go to the target pose(Rotation vector)    
param[in]:
- position: target pose;    
- velocity: joint acceleration of leading axis [rad/s^2];   
- accelerate: joint speed of leading axis [rad/s];   
- solution_space: move style, 'Joint' means it linear in joint-space,and 'Space' means linear in tool-space(forward kinematics is used to calculate the corresponding pose)    
return: bool, reaching target or not

get_state()   
Description: get robot state    
return: dictionary, the whole states of the UR10e

verify_state(variable_name, target_value, error=0.0001, time_out=10)   
Description: verify the robot reaching the target pose(joint or cartesian) or not    
param[in]:
- variable_name: target style, joints('q_actual') or cartesian('tool_vector_actual');    
- target_value: target values;   
- error: threshold,if the difference between current state and target state is small than threshold, we say the robot reached the target;   
- time_out: Max time of the motion, [second]    
return: bool, reaching target or not

Hand

HandE

class deepclaw.driver.grippers.handE_controller.HandEController(robot_ip = "192.168.1.10",port = 30003)

  • parameters[in]: robot_ip, the UR ip which the gripper mounted on; port, the UR ip.

  • return: an instance of gripper controller class

Methods

close_gripper()     
Description: close the gripper.

open_gripper()     
Description: open the gripper.

Camera

Realsense D435

class deepclaw.driver.sensors.cameras.Realsense(camera_configuration_file_path)

  • parameters[in]: camera_configuration_file_path, the camera configuration file, and the file format is yaml.

  • return: an instance of camera controller class

Methods

get_frame()     
Description: get images from a camera.
return: a tuple, includes color image, depth image, point cloud, and infrared images.

get_intrinsics()     
Description: get intrinsics attributes of a camera.
return: a tuple, intrinsics parameters of the camera

Modules Pool API

Calibration

deepclaw.modules.calibration.EyeOnBase.Calibration

class deepclaw.modules.calibration.EyeOnBase.Calibration(arm, camera, configuration_file)

Parameters

  • arm
  • camera
  • configuration_file

Methods

  • run(self): Calibrate robot arm and camera

End-to-End

deepclaw.modules.end2end.effecientdet.efficientdet_predictor.efficientdet

class deepclaw.modules.end2end.effecientdet.efficientdet_predictor.efficientdet(compound_coef=0, weight_path=None, num_classes=204)

Parameters

  • compound_coef
  • weight_path
  • num_classes

Methods

  • run(self, image_np)
  • display(self, preds, imgs, save_path, imshow=False, imwrite=True)

Segmentation

deepclaw.modules.segmentation.ContourDetector.ContourDetector

class deepclaw.modules.segmentation.ContourDetector.ContourDetector(mode=cv2.RETR_LIST, method=cv2.CHAIN_APPROX_NONE, binary_threshold=127, area_threshold=(80, 200), with_angle=False)

Parameter
  - mode
  - method
  - binary_threshold
  - area_threshold
  - with_angle

Methods

  • run(self, color_image)