7.4 KiB
Tutorial: Parameter Management
In this tutorial, we will discuss how parameters are managed in Stretch Body and show examples of how to customize your robot by overriding parameters.
Overview
Stretch Body shares a global set of parameters across all of the hardware it manages. All members of the Device class have an instance of RobotParams. This class constructs a dictionary of the device parameters as well as the global parameters for each device. For example, from iPython try:
import stretch_body.arm
a=stretch_body.arm.Arm()
a.params
Out[7]:
{'usb_name': '/dev/hello-motor-arm',
'force_N_per_A': 55.9
'chain_pitch': 0.0167,
'chain_sprocket_teeth': 10,
'gr_spur': 3.875,
'i_feedforward': 0,
'calibration_range_bounds': [0.515, 0.525],
'contact_model': 'effort_pct',
'contact_model_homing': 'effort_pct',
'contact_models': {'effort_pct': {'contact_thresh_calibration_margin': 10.0,
'contact_thresh_max': [-90.0, 90.0],
'contact_thresh_default': [-45.0, 45.0],
'contact_thresh_homing': [-45.0, 45.0]}},
'motion': {'default': {'accel_m': 0.14, 'vel_m': 0.14},
'fast': {'accel_m': 0.3, 'vel_m': 0.3},
'max': {'accel_m': 0.4, 'vel_m': 0.4},
'slow': {'accel_m': 0.05, 'vel_m': 0.05},
'trajectory_max': {'vel_m': 0.4, 'accel_m': 0.4}},
'range_m': [0.0, 0.52]}
or to access another device params:
a.robot_params['lift']
Out[9]:
{'usb_name': '/dev/hello-motor-lift',
'force_N_per_A': 75.0
'calibration_range_bounds': [1.094, 1.106],
'contact_model': 'effort_pct',
'contact_model_homing': 'effort_pct',
'contact_models': {'effort_pct': {'contact_thresh_calibration_margin': 10.0,
'contact_thresh_max': [-100, 100],
'contact_thresh_default': [-69.0, 69.0],
'contact_thresh_homing': [-69.0, 69.0]}},
'belt_pitch_m': 0.005,
'motion': {'default': {'accel_m': 0.2, 'vel_m': 0.11},
'fast': {'accel_m': 0.25, 'vel_m': 0.13},
'max': {'accel_m': 0.3, 'vel_m': 0.15},
'slow': {'accel_m': 0.05, 'vel_m': 0.05},
'trajectory_max': {'accel_m': 0.3, 'vel_m': 0.15}},
'pinion_t': 12,
'i_feedforward': 1.2,
'range_m': [0.0, 1.1]}
Parameter Organization
Stretch Body utilizes a prioritized parameter organization such that default settings can be easily overridden
| Priority | Name | Location | Description |
|---|---|---|---|
| 1 | user_params | $HELLO_FLEET_PATH/$HELLO_FLEET_ID/ stretch_user_params.yaml | Yaml file for users to override default settings and to define custom configurations. |
| 2 | configuration_params | $HELLO_FLEET_PATH/$HELLO_FLEET_ID/ stretch_configuration_params.yaml | Robot specific data (eg, serial numbers and calibrations). Calibration tools may update these. |
| 3 | external_params | Imported via a list defined as params in stretch_user_params.yaml |
External Python parameter dictionaries for 3rd party devices and peripherals. |
| 4 | nominal_params | stretch_body.robot_params_RE2V0.py | Generic systems settings (common across all robots of a given model. |
| 5 | nominal_system_params | stretch_body.robot_params.py | Generic systems settings (common across all robot models). |
This allows the user to override any of the parameters by defining it in their stretch_user_params.yaml. It also allows Hello Robot to periodically update parameters defined in the Python files via Pip updates.
The tool stretch_params.py will print out all of the robot parameters as well as their origin. For example:
stretch_params.py
############################################################ Parameters for stretch-re2-xxxx
Origin Parameter Value
--------------------------------------------------------------------------------------------------------------------------------- ...
stretch_body.robot_params.nominal_params param.arm.chain_pitch 0.0167
stretch_body.robot_params.nominal_params param.arm.chain_sprocket_teeth 10 ...
stretch_configuration_params.yaml param.arm.contact_models.effort_pct.contact_thresh_default [-45.0, 45.0]
Manually Querying and Modifying Parameters
A quick way to query parameters is with the stretch_params.py tool. For example, to look at parameters relating to the arm motion:
stretch_params.py | grep arm | grep motion
stretch_body.robot_params.nominal_params param.arm.motion.default.accel_m 0.14
stretch_body.robot_params.nominal_params param.arm.motion.default.vel_m 0.14
stretch_body.robot_params.nominal_params param.arm.motion.fast.accel_m 0.3
stretch_body.robot_params.nominal_params param.arm.motion.fast.vel_m 0.3
stretch_body.robot_params.nominal_params param.arm.motion.max.accel_m 0.4
stretch_body.robot_params.nominal_params param.arm.motion.max.vel_m 0.4
stretch_body.robot_params.nominal_params param.arm.motion.slow.accel_m 0.05
stretch_body.robot_params.nominal_params param.arm.motion.slow.vel_m 0.05
...
The tool displays each parameter's value as well as which parameter file it was loaded from.
For example, if you want to override the default motion settings for the arm, you could add the following to your stretch_user_params.yaml:
arm:
motion:
default:
vel_m: 0.1
accel_m: 0.1
Run the tool again and we see:
stretch_params.py | grep arm | grep motion | grep default
stretch_body.robot_params.nominal_params param.arm.motion.default.accel_m 0.1
stretch_body.robot_params.nominal_params param.arm.motion.default.vel_m 0.1
!!! note The factory parameter settings should suffice for most use cases.
Programmatically Modifying and Storing Parameters
A user can compute the value of a parameter programmatically and modify the robot settings accordingly. For example, in the Stretch Factory tool REx_base_calibrate_wheel_separation.py we see that the parameter wheel_separation_m is recomputed as the variable d_avg. This new value could be used during the robot execution by simply:
robot.base.params['wheel_separation_m']=d_vag
or it could be saved as a user override:
robot.write_user_param_to_YAML('base.wheel_separation_m', d_avg)
This will update the file stretch_user_params.yaml.