Tormach Robot Programming Language prompts

[machinekoder]

I'm working on the Tormach RPL, basically a Python-based programming language for the Tormach manipulators, and here are my first attempts to generate manipulator programs. The TRPL has been released only a few months ago, so ChatGPT does not know about it. However, since it's actually Python, ChatGPT can successfully create manipulator programs. With the right tuning, and the fact that this pure Python (3rd party modules can be installed), there are potentially a lot of use cases.

The TRPL docs can be found here: TRPL Docs

So here is my first attempt, I hope the formatting is right. Model: gpt-3.5-turbo from openai
Temperature: 0, Maximum length: 2048

system

Act as a code generator, explain the code by adding comments.

the TRPL is a python based robot programming language, a basic program looks as follows

"""
from robot_command.rpl import *
set_units("mm", "deg")
waypoint_1 = p[544.500, 0.001, 939.000, 124.908, -89.996, 55.087]
waypoint_2 = p[523.641, -354.892, 494.062, 179.929, -50.440, 0.048]

def main():
    movej(waypoint_1)
    sleep(2.5)
    movej(waypoint_2)
    sleep(2.0)
"""

 movej is the joint move command

a waypoint is a pose consisting of xyz cartesian position and abc euler rotation orientation, alternatively it can be created as follows: Pose(x=100, a=45)

waypoint * Pose(x=10,y=10) to offset the pose

user

create a TRPL program that moves between 4 waypoints which are the edges of a square, the center should be defined with a pose, the width of the square should be a parameter

assistant

"""
from robot_command.rpl import *
set_units("mm", "deg")

# Define the center pose and the width of the square
center_pose = Pose(x=500, y=500, z=500, a=0, b=0, c=0)
square_width = 200

# Define the four waypoints of the square
waypoint_1 = center_pose * Pose(x=square_width/2, y=square_width/2)
waypoint_2 = center_pose * Pose(x=square_width/2, y=-square_width/2)
waypoint_3 = center_pose * Pose(x=-square_width/2, y=-square_width/2)
waypoint_4 = center_pose * Pose(x=-square_width/2, y=square_width/2)

def main():
    # Move to waypoint 1
    movej(waypoint_1)
    sleep(2.5)

    # Move to waypoint 2
    movej(waypoint_2)
    sleep(2.5)

    # Move to waypoint 3
    movej(waypoint_3)
    sleep(2.5)

    # Move to waypoint 4
    movej(waypoint_4)
    sleep(2.5)
"""

In this program, we first define the center pose and the width of the square as variables. We then use these variables to calculate the four waypoints of the square using the `*` operator to offset the center pose by half the square width in the x and y directions.

In the `main()` function, we move the robot to each of the four waypoints in turn using the `movej()` command, and then pause for 2.5 seconds using the `sleep()` command.

user

create a TRPL program that moves in polygon shape with n corners and a given size and center waypoint

and so on...

The results look pretty good so far. I'm looking for ideas how to feed in the rest of the documentation or to train it with the documentation, so the potential applications go beyond just moving the end-effector in some simple shapes.

[muzixiaosanshi]

Hello Machinekoder,
@machinekoder I think it's amazing and funny project , But i have one question how can you feed the functions to chatgpt which they don't know it before?