ME 333 Suggested Final Projects
(→High-speed Motor Control from Matlab)
Revision as of 22:31, 9 January 2010
Students working on the projects listed below may begin their work right away, upon approval. Students wishing to propose their own project must write a proposal of approximately 3-5 pages, with at least one drawing (hand drawing OK) showing the whole device, a paragraph or two discussing the overall function and goal of the project, as well as discussions of the sensors and actuators you will use, the computation, and the mechanical design. Although you do not have to have worked out all the details, the proposal should show that you've thought about how the whole project will work. It can be a fun whimsical project, or it can solve a practical problem. Your project should creatively use simple sensors and actuation, but your proposal should be beyond simply applying what we do in lab. Previous projects are a good indicator of what's possible. Your proposal must also include milestones to be met during week 8.
Your final project cannot be a robot for DC.
Projects will be judged on functionality (does it do what it's supposed to do? does it serve a useful function?), reliability (does it do it every time?), ambitiousness (is the problem challenging? did you contribute a new capability to the wiki?), and aesthetic appeal (is it packaged nicely? is it pleasing to watch or fun to interact with?).
High-speed Motor Control from Matlab
You will make motor control PICs that make it easy for anyone to do high-speed encoder-based feedback control of brushed DC motors with only a few dollars of hardware and a cable to connect to matlab on your PC. Your project will build on this project, see also this project, and will consist of a "master" PIC that communicates with the PC plus any number of "slave" PICs, one for each motor to be controlled. The job of the master PIC is to take the program written in matlab and communicate it to the slave PICs, and to coordinate the initiation of moves by the slave PICs, which implement PID controllers at rates up to 2 kHz. Your final project will demonstrate high-speed trajectory following by a 2-DOF parallelogram linkage that can throw and catch objects in a vertical plane. Optional: use an EEPROM to perform data logging at each slave PIC, and send this data back to matlab upon request.
Week 8 milestone: Demonstrate controlled moves by the motors following a simple program in matlab, and show a design of the parallelogram linkage.
Portable 6-DOF Vibratory Platform for Parts Manipulation
The PPOD is a 6-degree-of-freedom vibratory platform that manipulates multiple parts simultaneously. See this video, for example. You will make a small, portable version (PPOD-Lite) that uses 6 small speakers arranged with 3 acting in the horizontal plane and 3 acting vertically. This version will be no larger than 8"x8"x6". It will have 6 axes of accelerometer feedback on the motion of the plate, and it will use FFTs of the accelerometer data to update the periodic speaker control signals so that the plate learns to follow the desired periodic motion profile. Desired motion profiles will be sent to the plate via a USB connection with a PC. Power for the device will come from a power adapter plugged into the wall.
Week 8 milestone:
Make a planar puppet actuated by strings pulled by robot arms out of view of the viewer. The robot arms could be actuated by RC servos. See this video (warning! it's 33 MB) of another automated marionette system.
Week 8 milestone: Demonstrate controlled 2-DOF motion, e.g., to make the arm of the puppet wave.
Moving Fish Refuge
(Client: Prof. Malcolm MacIver.) The weakly electric knifefish often prefers to hide in tight spaces rather than to swim in the open. See this video of this fish attempting to stay hidden in a tube as it is moved back and forth. Prof. MacIver would like to build a tube that can move under computer control, at different angles (other than purely horizontally), to see what motions of the tube the fish will track. You will build a 1-DOF prismatic device that slides a tube back and forth, and design a mechanism so that the angle of the tube can be changed relative to the linear slide.
Week 8 milestone: Demonstrate trajectory following of sinusoidal motions by a carriage along a linear rail using either a stepper motor or a motor and encoder.
Whisker-Based Tactile Profilometer
(Client: Prof. Mitra Hartmann.) Use a whisker-like sensor to measure object shapes in 3D, without use of a laser scanner. A simple length of wire (a "whisker") can be tapped along the contours of an object to extract the object's complete 3D shape. You will build a whisker/follicle combination actuated at the proximal end by a motor. The proximal end of the whisker will also be outfitted with strain gauges to measure bending moments and axial force. By moving the wire over an object and sensing these forces and moments, the object's shape can be reconstructed. Find some cool videos of rat whisking here (scroll down to the eighth paper or so).
Experiment in Nonlinear Dynamics: Ball Bouncing
You will build a classic experimental apparatus in nonlinear dynamics. The PIC controls the oscillation of a speaker in the vertical direction. A concave lens is attached to the top of the speaker, and a ball bearing bounces on the lens as the speaker moves in the vertical direction. The concavity of the lens keeps the ball centered. For some frequencies and amplitudes, the ball will bounce in a steady cycle, bouncing to approximately the same height after each impact. As you increase the amplitude, however, the ball may go through a period-two phase, where it first flies high, then low, then high, then low, etc. If you continue to increase the amplitude, you may see higher-period behavior, and finally the motion of the ball will be hard to predict or "chaotic." You will create a microphone circuit to listen for the impact between the ball and the lens, and the PIC will calculate the time between impacts and send the data back to a matlab program. The PIC will also monitor the acceleration of the speaker using an accelerometer and send that data back to the matlab program. The matlab program controls the amplitude of the speaker oscillation by communicating with the PIC. The matlab program plots the experimental results, possibly as a "bifurcation diagram." See this site and this article for more information on this experiment.
You can also use the fact that you have a PIC microcontroller to do real-time feedback control, modulating the control signal to the speaker as a function of the duration between ball impacts.
Week 8 milestone: Demonstrate successful control of the amplitude of the speaker motion, reading of the accelerometer, and bidirectional communication between the PIC and matlab.
Three-Speaker Chladni Pattern Generator
See videos on youtube of "Chladni Patterns." Now imagine making one actuated by three separate speakers spaced at 120 degrees on a round plate. Signal generator chips, controlled by the PIC, will be used to generate the input signals to the speaker amplifiers. Provide knobs to allow the user to set the vibration patterns, or switch to stored motions on the PIC. See what kinds of patterns you can create.
Week 8 milestone: Demonstrate control of one speaker using a signal generator chip and analog (knob) inputs to the PIC. Show initial (traditional) Chladni patterns.
Furuta Pendulum (Inverted Pendulum)
Build an inverted pendulum system to experiment with balancing control algorithms. This kind of inverted pendulum is called a Furuta pendulum. A motor controls the rotation of a horizontal beam about a stationary vertical pivot axis, and a second beam rotates freely about an axis along the horizontal beam. The goal is to take the second beam from the hanging down configuration to balanced in the up configuration. An encoder measures the angle of the second beam. The motor pumps energy into the beam, then (probably) switches to a stable linearizing controller when the beam is vertical. See this page, for example. This problem is a challenging dynamic task, and is impressive when it works.
Week 8 milestone: Have the mechanical system nearly finalized, with controlled motion of the motor. The swing-up and balancing algorithms do not have to be implemented yet.
- Claude Shannon, known as the "Father of Information Theory," made a bounce juggling machine many years ago, perhaps the first to make a real juggling machine. It uses no sensor feedback. See the video at . Can you build something similar?
- This would make a great demo if you had a dog! See .
- Build a 2-DOF pizza manipulator. See .
The '*' indicates that a proposal is due.
- Team 11: Three-speaker Chladni patterns
- Team 12: Juggling*
- Team 13: High-speed motor control
- Team 14: Automated fish refuge
- Team 15: Rock-paper-scissors*
- Team 16: TBD musical instrument*
- Team 21: Marionette
- Team 22: Swinging robot*
- Team 23: Portable 6-DOF PPOD
- Team 24: Automated xylophone*
- Team 25: Vision-based cannon*
- Team 26: Persistence-of-vision display*