Robot Motion Controller

Team: Andy Payne
Developed by Andy Payne, this project aims to bring physical input and output closer together through the design of purpose-built tools for fabrication, potentially leading to many new creative opportunities for designers.

Working from observations about the way architects design, this project explores the development of a novel 3D drawing tool or customized 5-axis digitizing arm that takes real-time input and translates movement patterns directly into machine code for robotic fabrication.  An improved workflow for robotic simulation was also developed as part of this project; using design tools that are already familiar to architects and designers, such as Rhino and Grasshopper.  The work was presented at the 2011 ACADIA Conference: Integration Through Computation held in Banff, Canada from October 13th-16th. Click here to download the full paper.

Conceptually, this project started by looking at the joint and axis configurations of the ABB-IRB 140 robot.  The design challenge then, was to design a tangible controller around these constraints.  By using the same joint and axis configurations, the digitizing arm has a one to one relationship with the larger industrial robot.  It is intuitive, and users understand that when they move the digitizing arm in a certain way, the robot will respond in kind.

 

Outside of the development of a new robotic workflow, one of the primary goals of the project was to minimize costs. The entire controller was built for less than $200 dollars. The aluminum body was water jet cut and all of the hardware can be purchased from local hardware stores or online retailers. All of the sensors, including the five high-precision potentiometers and the two digital sensors on the tool tip were also purchased from online retailers and were chosen because of their affordability.

To create a real-time robotic simulation, data from each of the embedded sensors on the tangible controller are streamed into the computer using a plug-in for Grasshopper called Firefly, also developed by Andrew Payne.  Among other things, Firefly enables the use of real-world data, acquired from various sensors or other input devices to explicitly define parametric relationships within a Grasshopper model.  In this project, sensor information is used to create a forward kinematic robotic simulation. Forward kinematics describes one type of solution for determining robotic positioning.  If given all of the relative angles of each joint and the lengths of each leg; the tool tip (also known as the end effector) can be found by performing a series of matrix transformations on each body in the robotic mechanism.  In this case, each of the potentiometers will return a 10-bit number between 0 and 1023.  These particular potentiometers were able to rotate up to 340º, so the angle between each joint can be found by simply multiplying the current sensor value by the sensor step size.  These angle values are used to perform a matrix transformation on each of the robotic legs, ultimately giving you the precise position of the tool center point. Once you know the location of the end effector, you can record this data over time to create real-time robotic tool paths.

In addition to the five high-precision potentiometers, the digitizing arm is equipped with a tool tip circuit board with two push button controls.  These allow the user to easily record or reset the digitized information.   A custom built circuit board processes all of the sensor information and sends a formatted string of information over the serial port to the virtual parametric interface.

The Grasshopper definition is relatively straight forward.  The Firefly Read component parses of the sensor information being sent directly from the microcontroller circuit board.  There is a custom component written in VB.NET (seen in item number 2), which creates the necessary tool data. The data from both of these components are fed into another custom component which calculates the forward kinematic solution and outputs the position of each leg, creating a real-time preview of the robot moving in the Rhino viewport.  In addition, the robotic simulator also returns the RAPID code, or the robotic programming language used by all of the ABB robots, to move the actual robot in the same manner as the forward kinematic preview.

The custom robotic simulation component written inside of Grasshopper outputs all of the necessary RAPID code to control the actual robot.  There are two methods by which this can be done.  First, all of the digitizing information is recorded and the movement commands are created.  Once the program has been written, it can then be saved to a file on disk and uploaded to the robotic controller to be played back.  An alternative method is to stream the angle information from the digitizing arm directly to the robot through a network cable.  In this method, a program is uploaded to the robot which tells it to sit and wait for any information being sent directly from the Grasshopper definition.