Skip to content

Latest commit

 

History

History
502 lines (385 loc) · 19.5 KB

README.md

File metadata and controls

502 lines (385 loc) · 19.5 KB
Raw

abb_motion_program_exec

abb_motion_program_exec provides a simple way to download and run a sequence of MoveAbsJ, MoveJ, MoveL, MoveC, and WaitTime commands on an ABB IRC5 robot controller. This program is intended to be a proof of concept for more sophisticated controller interfaces. Multi-move control of two robots is also supported. Externally Guided Motion (EGM) is also supported for joint target, pose target, and path correction modes.

Documentation can be found at: https://abb-motion-program-exec.readthedocs.io/

Installation

abb-motion-program-exec is avaliable on PyPi.

pip install abb-motion-program-exec

Begin by installing the software for the robot controller. This software can be installed manually by copying files to the robot controller and importing configuration files.

This contains the robot-side code, that reads and executes the contents of motion_program.bin. motion_program.bin contains the sequence of instructions to run, encoded in binary format for fast interpretation.

Only one instance of a Robot Studio virtual controller can be run at a time. Only instances of Robot Studio can be run at a time running a single virtual controller. This is due to the controller using TCP port 80 on the local computer to accept REST commands from Python. If more than one controller is started, TCP port 80 will already be in use and can cause unpredictable behavior. Restart the computer if connections cannot be made from Python to the controller. Multiple real robots can be used concurrently since they will each have a unique IP address to bind port 80.

Python 3.6 Linux Install (Ubuntu Bionic)

Older versions of Python are not supported by the currently available protobuf package. Use the apt version instead.

sudo apt install python3-virtualenv python3-protobuf python3-numpy python3-wheel python3-setuptools
python3 -m pip install --user abb-motion-program-exec

Usage

The robot must be in "Auto" mode for this driver to operate. A non-descriptive error code will be returned if attempting to execute a motion program in manual mode.

Once the abb_motion_program_exec.mod has been loaded on the controller, the Python module can be used to command motion sequences. The class MotionProgram contains is used to build the sequence of motions. It has the following commands of interest:

  • MoveAbsJ(to_joint_pos: jointtarget, speed: speeddata, zone: zonedata) - Move the robot to a specified joint waypoint.
  • MoveJ(to_point: robtarget, speed: speeddata, zone: zonedata) - Move the robot to the specified Cartesian target using joint interpolation.
  • MoveL(to_point: robtarget, speed: speeddata, zone: zonedata) - Move the robot to the specified Cartesian target using linear interpolation.
  • MoveC(cir_point: robtarget, to_point: robtarget, speed: speeddata, zone: zonedata) - Move the robot to the specified Cartesian target circularly using an intermediate waypoint.
  • WaitTime(t: float) - Wait a specified time in seconds.

Calling each of these functions adds the command to the sequence.

The constructor for MotionProgram optionally takes a tool parameter. This parameter is expected to be type tooldata and will be passed to each of the move commands. Because the tool is expected to be a PERS type by the ABB software, it can't be modified for each motion command without a significant performance penalty.

my_motion_program = MotionProgram(tool=my_tool)

The following types are defined as subclasses of NamedTuple:

class speeddata(NamedTuple):
    v_tcp: float
    v_ori: float
    v_leax: float
    v_reax: float

class zonedata(NamedTuple):
    finep: bool
    pzone_tcp: float
    pzone_ori: float
    pzone_eax: float
    zone_ori: float
    zone_leax: float
    zone_reax: float

class jointtarget(NamedTuple):
    robax: np.ndarray # shape=(6,)
    extax: np.ndarray # shape=(6,)

class pose(NamedTuple):
    trans: np.ndarray # [x,y,z]
    rot: np.ndarray # [qw,qx,qy,qz]

class confdata(NamedTuple):
    cf1: float
    cf4: float
    cf6: float
    cfx: float

class robtarget(NamedTuple):
    trans: np.ndarray # [x,y,z]
    rot: np.ndarray # [qw,qx,qy,qz]
    robconf: confdata # 
    extax: np.ndarray # shape=(6,)

class loaddata(NamedTuple):
    mass: float
    cog: np.ndarray # shape=(3,)
    aom: np.ndarray # shape=(4,)
    ix: float
    iy: float
    iz: float

class tooldata(NamedTuple):
    robhold: bool
    tframe: pose
    tload : loaddata

See the ABB Robotics manual "Technical reference manual RAPID Instructions, Functions and Data types" for more details on these data types. Note that pos, orient, robjoint, and extjoint are implemented using numpy arrays or lists.

The following standard speeddata are available in the module: v5, v10, v20, v30, v40, v50, v60, v80, v100, v200, v300, v400, v500, v600, v800, v1000, v1500, v2000, v2500, v3000, v4000, v5000, v6000, v7000, vmax.

The following standard zonedata are available in the module: fine, z0, z1, z5, z10, z15, z20, z30, z40, z50, z60, z80, z100, z150, z200.

The following tooldata are available in the module: tool0

Once the program is complete, it can be executed on the robot using MotionProgramExecClient. The constructor is by default:

mp_client = MotionProgramClient(base_url='http://127.0.0.1:80', username='Default User', password='robotics')

The base_url, username, and password should be adjusted to the actual robot. The client using ABB Web Services. base_url must be set to the IP address of the robot, or using localhost if using the simulator.

Once the client is constructed, it can be used to execute the program:

log_results = mp_client.execute_motion_program(mp)

log_results is a tuple containing the results of the motion:

class MotionProgramResultLog(NamedTuple):
    timestamp: str
    column_headers: List[str]
    data: np.array

timestamp is a string timestamp for the data. column_headers is a list of strings containing the labels of the columns of data. data contains a float32 numpy 2D array of the data, with each row being a sample.

For a single robot, the data has the following columns:

  • timestamp - The time of the row. This is time from the startup of the logger task in seconds. Subtract the initial time from all samples to get a 0 start time for the program.
  • cmd_num - The currently executing command number. Use get_program_rapid() to print out the program with command numbers annotated.
  • J1 - Joint 1 position in degrees
  • J2 - Joint 2 position in degrees
  • J3 - Joint 3 position in degrees
  • J4 - Joint 4 position in degrees
  • J5 - Joint 5 position in degrees
  • J6 - Joint 6 position in degrees

The field column_headers contains a list of the column headers.

Python module installation

The abb_motion_program_exec module is available on PyPI and can be installed using pip:

pip install abb_motion_program_exec

For development, use of a virtualenv is recommended. Use editable install with the virtualenv:

pip install -e .

Externally Guided Motion (EGM)

See egm.md for instructions on using EGM.

Example

import abb_motion_program_exec as abb

j1 = abb.jointtarget([10,20,30,40,50,60],[0]*6)
j2 = abb.jointtarget([-10,15,35,10,95,-95],[0]*6)
j3 = abb.jointtarget([15,-5,25,83,-84,85],[0]*6)

my_tool = abb.tooldata(True,abb.pose([0,0,0.1],[1,0,0,0]),abb.loaddata(0.001,[0,0,0.001],[1,0,0,0],0,0,0)) 

mp = abb.MotionProgram(tool=my_tool)
mp.MoveAbsJ(j1,abb.v1000,abb.fine)
mp.MoveAbsJ(j2,abb.v5000,abb.fine)
mp.MoveAbsJ(j3,abb.v500,abb.fine)
mp.MoveAbsJ(j2,abb.v5000,abb.z50)
mp.MoveAbsJ(j3,abb.v500,abb.z200)
mp.MoveAbsJ(j2,abb.v5000,abb.fine)
mp.WaitTime(1)

r1 = abb.robtarget([350., -100., 600.], [ 0.0868241, -0.0868241, 0.9924039, 0.0075961 ], abb.confdata(-1,0,-1,0),[0]*6)
r2 = abb.robtarget([370., 120., 620. ], [ 0.0868241, 0.0868241, 0.9924039, -0.0075961], abb.confdata(0,-1,0,0),[0]*6)

r3 = abb.robtarget([400., -200., 500.], [0.7071068, 0., 0.7071068, 0.], abb.confdata( -1.,  -3., 2.,  0.),[0]*6)
r4 = abb.robtarget([400., 0., 580.], [0.7071068, 0., 0.7071068, 0.], abb.confdata(0.,  -3., 2.,  0.), [0]*6)
r5 = abb.robtarget([400., 200., 500.], [0.7071068, 0., 0.7071068, 0.], abb.confdata(0.,  -2., 1.,  0.),[0]*6)

mp.MoveJ(r1,abb.v500,abb.fine)
mp.MoveJ(r2,abb.v400,abb.fine)
mp.MoveJ(r1,abb.v500,abb.z100)
mp.MoveJ(r2,abb.v400,abb.z100)
mp.MoveJ(r1,abb.v500,abb.fine)
mp.WaitTime(1.5)

mp.MoveJ(r3,abb.v5000,abb.fine)
mp.MoveL(r4,abb.v200,abb.fine)
mp.MoveL(r3,abb.v200,abb.fine)
mp.MoveL(r4,abb.v1000,abb.z100)
mp.MoveL(r3,abb.v1000,abb.z100)
mp.MoveL(r4,abb.v1000,abb.fine)
mp.WaitTime(2.5)

mp.MoveJ(r3,abb.v5000,abb.fine)

mp.MoveC(r4,r5,abb.v200,abb.z10)
mp.MoveC(r4,r3,abb.v50,abb.fine)

# Print out RAPID module of motion program for debugging
print(mp.get_program_rapid())

# Execute the motion program on the robot
# Change base_url to the robot IP address
client = abb.MotionProgramExecClient(base_url="http://127.0.0.1:80")
log_results = client.execute_motion_program(mp)

# log_results.data is a numpy array
import matplotlib.pyplot as plt
fig, ax1 = plt.subplots()
lns1 = ax1.plot(log_results.data[:,0], log_results.data[:,2:])
ax1.set_xlabel("Time (s)")
ax1.set_ylabel("Joint angle (deg)")
ax2 = ax1.twinx()
lns2 = ax2.plot(log_results.data[:,0], log_results.data[:,1], '-k')
ax2.set_ylabel("Command number")
ax2.set_yticks(range(-1,int(max(log_results.data[:,1]))+1))
ax1.legend(lns1 + lns2, log_results.column_headers[2:] + ["cmdnum"])
ax1.set_title("Joint motion")
plt.show()

Multi-Move Robot Example

Two robots can be controlled using ABB Multi-Move. See robot_multimove_setup_manual.md for setup instructions.

They must have exactly the same number of motion commands. The commands are passed with the \ID parameter corresponding to the command number. SyncMoveOn is activated to cause the robots to move in sync. The execute_multimove_motion_program() function of MotionProgramExecClient is used to send multi-move programs to the robot.

import abb_motion_program_exec as abb

# Fill motion program for T_ROB1
t1 = abb.robtarget([575,-200,1280],[0,-.707,0,.707],abb.confdata(0,0,-1,1),[0]*6)
t2 = abb.robtarget([575,200,1480],[0,-.707,0,.707],abb.confdata(-1,-1,0,1),[0]*6)
t3 = abb.robtarget([575,0,1280],[0,-.707,0,.707],abb.confdata(-1,-1,0,1),[0]*6)

my_tool = abb.tooldata(True,abb.pose([0,0,0.1],[1,0,0,0]),abb.loaddata(0.001,[0,0,0.001],[1,0,0,0],0,0,0)) 

mp = abb.MotionProgram(tool=my_tool)
mp.SyncMoveOn()
mp.MoveAbsJ(abb.jointtarget([5,-20,30,27,-11,-27],[0]*6),abb.v1000,abb.fine)
mp.MoveL(t1,abb.v1000,abb.fine)
mp.MoveJ(t2,abb.v5000,abb.fine)
mp.MoveL(t3,abb.v500,abb.fine)
mp.WaitTime(1)
mp.MoveL(t1,abb.v5000,abb.z50)
mp.MoveJ(t2,abb.v500,abb.z200)
mp.MoveL(t3,abb.v5000,abb.fine)

# Fill motion program for T_ROB2. Both programs must have
# same number of commands
t1_2 = abb.robtarget([250,-200,1280],[.707,0,.707,0],abb.confdata(-1,-1,0,1),[0]*6)
t2_2 = abb.robtarget([250,200,1480],[.707,0,.707,0],abb.confdata(0,0,-1,1),[0]*6)
t3_2 = abb.robtarget([250,0,1280],[.707,0,.707,0],abb.confdata(0,0,0,1),[0]*6)

my_tool2 = abb.tooldata(True,abb.pose([0,0,0.5],[1,0,0,0]),abb.loaddata(0.1,[0,0,0.1],[1,0,0,0],0,0,0)) 

mp2 = abb.MotionProgram(tool=my_tool2)
mp2.SyncMoveOn()
mp2.MoveAbsJ(abb.jointtarget([1,1,40,2,-40,-2],[0]*6),abb.v1000,abb.fine)
mp2.MoveJ(t1_2,abb.v1000,abb.fine)
mp2.MoveL(t2_2,abb.v5000,abb.fine)
mp2.MoveL(t3_2,abb.v500,abb.fine)
mp2.WaitTime(1)
mp2.MoveL(t1_2,abb.v5000,abb.z50)
mp2.MoveL(t2_2,abb.v500,abb.z200)
mp2.MoveL(t3_2,abb.v5000,abb.fine)


# Execute the motion program on the robot
# Change base_url to the robot IP address
client = abb.MotionProgramExecClient(base_url="http://127.0.0.1:80")

# Execute both motion programs simultaneously
log_results = client.execute_multimove_motion_program([mp,mp2])

# log_results.data is a numpy array
import matplotlib.pyplot as plt
fig, ax1 = plt.subplots()
lns1 = ax1.plot(log_results.data[:,0], log_results.data[:,2:8])
ax1.set_xlabel("Time (s)")
ax1.set_ylabel("Joint angle (deg)")
ax2 = ax1.twinx()
lns2 = ax2.plot(log_results.data[:,0], log_results.data[:,1], '-k')
ax2.set_ylabel("Command number")
ax2.set_yticks(range(-1,int(max(log_results.data[:,1]))+1))
ax1.legend(lns1 + lns2, log_results.column_headers[2:8] + ["cmdnum"])
ax1.set_title("Robot 1 joint motion")
fig, ax1 = plt.subplots()
lns1 = ax1.plot(log_results.data[:,0], log_results.data[:,8:])
ax1.set_xlabel("Time (s)")
ax1.set_ylabel("Joint angle (deg)")
ax2 = ax1.twinx()
lns2 = ax2.plot(log_results.data[:,0], log_results.data[:,1], '-k')
ax2.set_ylabel("Command number")
ax2.set_yticks(range(-1,int(max(log_results.data[:,1]))+1))
ax1.legend(lns1 + lns2, log_results.column_headers[8:] + ["cmdnum"])
ax1.set_title("Robot 2 joint motion")
plt.show()

Multi-Move example using relative work object:

# Multi-Move example using relative robot end effector poses

import abb_motion_program_exec as abb


# Fill motion program for T_ROB1

# Hold constant relative position for this example
t1 = abb.robtarget([0,0,-200],[1,0,0,0],abb.confdata(0,0,1,1),[0]*6)
t2 = t1
t3 = t1


my_tool = abb.tooldata(True,abb.pose([0,0,0],[1,0,0,0]),abb.loaddata(0.001,[0,0,0.001],[1,0,0,0],0,0,0))
my_wobj = abb.wobjdata(False,False,"ROB_2",abb.pose([0,0,0],[1,0,0,0]),abb.pose([0,0,0],[0,0,1,0]))

mp = abb.MotionProgram(tool=my_tool,wobj=my_wobj)
mp.SyncMoveOn()
mp.MoveAbsJ(abb.jointtarget([5,-20,30,27,-11,172],[0]*6),abb.v1000,abb.fine)
mp.WaitTime(0.1)
mp.MoveJ(t1,abb.v1000,abb.fine)
mp.MoveJ(t1,abb.v1000,abb.fine)
mp.MoveL(t1,abb.v1000,abb.fine)

# Fill motion program for T_ROB2. Both programs must have
# same number of commands
t1_2 = abb.robtarget([250,-200,1280],[.707,0,.707,0],abb.confdata(-1,-1,0,1),[0]*6)
t2_2 = abb.robtarget([250,200,1480],[.707,0,.707,0],abb.confdata(0,0,-1,1),[0]*6)
t3_2 = abb.robtarget([250,0,1280],[.707,0,.707,0],abb.confdata(0,0,0,1),[0]*6)

my_tool2 = abb.tooldata(True,abb.pose([0,0,0.5],[1,0,0,0]),abb.loaddata(0.1,[0,0,0.1],[1,0,0,0],0,0,0)) 

mp2 = abb.MotionProgram(tool=my_tool2)
mp2.SyncMoveOn()
mp2.MoveAbsJ(abb.jointtarget([1,1,40,2,-40,-2],[0]*6),abb.v1000,abb.fine)
mp2.WaitTime(0.1)
mp2.MoveL(t1_2,abb.v1000,abb.fine)
mp2.MoveL(t2_2,abb.v5000,abb.fine)
mp2.MoveL(t3_2,abb.v500,abb.fine)

# Execute the motion program on the robot
# Change base_url to the robot IP address
client = abb.MotionProgramExecClient(base_url="http://127.0.0.1:80")

# Execute both motion programs simultaneously
log_results = client.execute_multimove_motion_program([mp,mp2])

# log_results.data is a numpy array
import matplotlib.pyplot as plt
fig, ax1 = plt.subplots()
lns1 = ax1.plot(log_results.data[:,0], log_results.data[:,2:8])
ax1.set_xlabel("Time (s)")
ax1.set_ylabel("Joint angle (deg)")
ax2 = ax1.twinx()
lns2 = ax2.plot(log_results.data[:,0], log_results.data[:,1], '-k')
ax2.set_ylabel("Command number")
ax2.set_yticks(range(-1,int(max(log_results.data[:,1]))+1))
ax1.legend(lns1 + lns2, log_results.column_headers[2:8] + ["cmdnum"])
ax1.set_title("Robot 1 joint motion")
fig, ax1 = plt.subplots()
lns1 = ax1.plot(log_results.data[:,0], log_results.data[:,8:])
ax1.set_xlabel("Time (s)")
ax1.set_ylabel("Joint angle (deg)")
ax2 = ax1.twinx()
lns2 = ax2.plot(log_results.data[:,0], log_results.data[:,1], '-k')
ax2.set_ylabel("Command number")
ax2.set_yticks(range(-1,int(max(log_results.data[:,1]))+1))
ax1.legend(lns1 + lns2, log_results.column_headers[8:] + ["cmdnum"])
ax1.set_title("Robot 2 joint motion")
plt.show()

Robot Raconteur Service

A Robot Raconteur service is available to allow a client to execute multi-move programs. The service uses a standards track service definition experimental.robotics.motion_program that will provide interoperability between multiple robot types. Due to the differences in the robot commands between controller implementations there may be slight differences in the commands between robots, but in general the commands are very similar.

When installing the module, include the robotraconteur option to get the required dependencies:

python -m pip install abb-motion-program-exec[robotraconteur]

Start the service, specifying a robot info file:

abb-motion-program-exec-robotraconteur --mp-robot-info-file=config/abb_multimove_motion_program_robot_default_config.yml

Optionally start using a module if the entrypoint does not work:

python -m abb_motion_program_exec.robotraconteur --mp-robot-info-file=config/abb_multimove_motion_program_robot_default_config.yml

The following options are supported:

  • --mp-robot-info-file= - The info file that specifies information about the robot
  • --mp-robot-base-url= - The connection URL for Robot Web Services on the robot. Defaults to http://127.0.0.1:80 for use with Robot Studio virtual controllers. Set 127.0.0.1 to the WAN IP address of the robot.
  • --mp-robot-username= - The robot controller username. Defaults to "Default User"
  • --mp-robot-password= - The robot controller password. Defaults to "robotics"

Examples for a single robot and multi-move robots are in the examples/robotraconteur directory. The motion programs make heavy use of varvalue types to allow for flexibility in the motion program contents. The Python type RR.VarValue is used to represent the varvalue type. See the Robot Raconteur Python documentation for more inforamtion on how RR.VarValue works and why it is necessary.

Connection Info

  • URL: rr+tcp://localhost:59843?service=mp_robot
  • Node Name: experimental.robotics.motion_program
  • Device Name: abb_robot
  • Service Name: mp_robot
  • Root Object Type: experimental.robotics.motion_program.MotionProgramRobot

License

Apache 2.0 License, Copyright 2022 Wason Technology, LLC, Rensselaer Polytechnic Institute

Acknowledgment

This work was supported in part by Subaward No. ARM-TEC-21-02-F19 from the Advanced Robotics for Manufacturing ("ARM") Institute under Agreement Number W911NF-17-3-0004 sponsored by the Office of the Secretary of Defense. ARM Project Management was provided by Christopher Adams. The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the official policies, either expressed or implied, of either ARM or the Office of the Secretary of Defense of the U.S. Government. The U.S. Government is authorized to reproduce and distribute reprints for Government purposes, notwithstanding any copyright notation herein.

This work was supported in part by the New York State Empire State Development Division of Science, Technology and Innovation (NYSTAR) under contract C160142.