Student Projects
Contact Information
University and Department: University of California, Berkeley; Department of Mechanical Engineering
Team Members: Jiyang Liu, Hui Kuen Chan
Faculty Advisors: George Anwar
Primary Email Address: jiyangliu@berkeley.edu
Primary Telephone Number (include area and country code): (415) 350-1363
Project Information
Project Title: Robot Artist
List all parts (hardware, software, etc.) you used to design and complete your project:
NI LabView 2009
NI Vision Assistant 2009
NI Vision Acquisition Software 2009
NI Vision Development Module 2009
NI Control Design and Simulation Module 2009
NI Real-Time Module 2009
NI MathScript RT Module 2009
Describe the challenge your project is trying to solve.
Capturing image or a person’s portrait directly by using webcam
Resize the input image so that its size is within the drawable range limit of the XY stage
Using XY stage to draw a person’s portrait or any pictures with continuous lines
Constructing a algorithm that converts pxiels to continous lines
Precise position control of the XY stage
Describe how you addressed the challenge through your project.
The most challenge part is the algorithm that defines the continuous lines. In other word, teaching the XY stage how to move is the hardest part of this project. Initially, the team planed to use while loops to evaluate every pixel in an image to determine the possible drawing directions of every step. However, this method is very inefficient, and it requires multiple arrays to record the possible moves. Since the input image is 1600 by 1200 pixels, the while loops method took approximately 17 minutes to process an image in the simulation. Therefore, the team came up with another method.
In this method, the team first converts an image into binary array, and then uses comparison operation to find and record the x coordinate and y coordinate of each pixel only with color black. Afterward, starting from (0,0), a for loop is used to find the next closest coordinates with respect to the current coordinates. Since the length of the for loop is set to be the number of pixels with color black, this method doesn’t require using multiple arrays to record the possible directions of move, and it reduces the number of needed evaluations dramatically. As a result, depends on the complexity of the image, this new method now only takes 45 to 65 seconds to process an image.
Explain why you choose NI technology.
National Instruments technology has very good hardware compatibility. It enables us to interact with all kinds of hardware even though the hardware might not be built to be used with NI products. In our case, instead of using an expensive educational camera, we use a generic webcam along with the Vision Acquisition Software 2009 and Vision Assistant from NI. Since we are mechanical engineers and the semester is short, writing our own drivers for hardware is not an option. The compatibility of NI software saves our time from those issues, so we can focus on emphasizing our strengths. It enables us to spend more time on developing new features in our Graphical User Interface (GUI), improving the control algorithm for the robot, and modifying the system platform to achieve a better outcome. Moreover, the LABVIEW software also allows us to understand the data flow of our system visually. We can therefore easily determine the drawbacks in our codes, and thus improve the efficiency of our system. In fact, it would be much more challenge for our team to find the inefficient designs in our control algorithm if we used traditional textual programming. Also, the mathscript in LABVIEW allows us to perform trial and error, and thus determine the best set of control parameters for our robot.
As for hardware, the LABVIEW based real-time machine and FPGA module allowed us to receive signal from our robot and process them at the MHz level, which enable us for extremely precise control over our system, allowing our drawing to always have error of less than 0.5mm.
The mathscript feature in LABVIEW is one thing we really appreciate. As most programmers do, our team are very used to textual based programming, therefore we started our designs and simulations for the control algorithm in Matlab, and it was very challenging to convert all our Matlab codes into LABVIEW G language at once because we don’t know where we should start. With the help of mathscript, we can convert the textual codes gradually. Not only did it allow us to build our VI step-by-step, but it also gives us a systematic way to understand the G language.
The ability of running parallel loops also deeply impressed our team. A modification in GUI was done because of this feature. The user of our robot artist can now take as many picture shots as they want without restarting the entire VI. Our latest GUI design separates the picture capturing, the image processing, and the camera recording into three different loops. Another feature in LABVIEW that was unexpectedly useful was the network variable feature. At first we found it extremely tedious to be used, and thought it was a hassle that it must be used to communicate between the PC and the real-time machine. However, as we move along on the programming, we found that many variables must be used in many VIs simultaneously, and this is the moment that we found the network variable feature to be extremely useful. It helps us to solve a lot of problems. Last but not least, the rich GUI library in LABVIEW is the other design we appreciate. It enables us to create a nice and sophisticated GUI in a relatively simple way.
Here are some drawings of the robot artist:
.jpg)
