Difference between revisions of "The Gibbot"
Nelson Rosa (talk | contribs) |
Nelson Rosa (talk | contribs) |
||
| Line 39: | Line 39: | ||
<!-- Milestone: The receiver should be a function on the PIC that returns an (x, y) location using a unit of metric length. Camera should be chosen and calibrated. Give PCB redesigner (see below) the inputs and outputs of module. |
<!-- Milestone: The receiver should be a function on the PIC that returns an (x, y) location using a unit of metric length. Camera should be chosen and calibrated. Give PCB redesigner (see below) the inputs and outputs of module. |
||
Deliverables: For the camera, a Gibbot tracking algorithm as well as a calibrated lens for picking up the robot’s location. --> |
Deliverables: For the camera, a Gibbot tracking algorithm as well as a calibrated lens for picking up the robot’s location. --> |
||
| ⚫ | |||
Mechanical: For prototyping any size wall and material is fine. |
|||
Electrical: Wire up the LED sensors (1- or 2-inch spacing). For both run a calibration routine that can pick up the location of the robot. How large of a PCB can we order of the LED array? Build a PIC32 library that can take in the image and LED data and produce the robot’s location. The idea here is to build a prototype that is easy to scale up for spring term. |
|||
Deliverables: A mockup showing a gridded wall with an emitter/detector pair inside each grid pulsing its location (let’s call these guys the transmitters). Communications protocol should be a simple IR protocol (copy same one for TV remotes). Demo must show onboard Gibbot circuitry (the receiver) picking up the transmitter signal. Should also output data to PC via XBee. Must make sure that work can be scaled up to an 8ft. x 6 ft. wall. What are the power and component requirements as we scale up? |
|||
Milestone: Idea of artistic use for wall. small scale PCB circuitry showing receiver and transmitter talking to each other. The receiver should be a function on the PIC that returns an (x, y) location. Give PCB redesigner (see below) the inputs and outputs of module. |
|||
==Main PCB Design== |
==Main PCB Design== |
||
| Line 61: | Line 52: | ||
<!-- ==Wheel Carriage (1 person)== |
<!-- ==Wheel Carriage (1 person)== |
||
Build wheel carriage prototype. If it works well think of integrating it into the gbot design for spring term. --> |
Build wheel carriage prototype. If it works well think of integrating it into the gbot design for spring term. --> |
||
| ⚫ | |||
We need an alternative form of position sensing for the robot beyond a camera. An alternative that we believe would work is to wire LEDs (1- or 2-inches apart) on the steel sheet. The robot would then pick up a pulse-encoded signal from the LED telling the robot where it is. Constructing the smart wall isn't a problem, it's maintaining it. If we look at 2-inch spacing, then for a 48 sq. ft. wall we are looking at 1,728 LEDs. If one burns out, how will an MSI employee find the LED and replace it with ease? Can you come up with a better design for a backup sensor that does not rely on conventional wireless communication? |
|||
Revision as of 01:53, 16 January 2013
Questions for the group: Where should we host the project? (Google Docs, DropBox, GitHub, Google code, SourceForge, etc.)
Below is a list of projects. Each project will have a team lead responsible for keeping everyone on track. We also ask that everyone is constantly documenting their work and that they develop a test suite along with their deliverables. The tests should include operation under “normal” operation, bad operating conditions (how do you gracefully recover?), and really bad operating conditions (how will the system persist without your module? What are the redundancies in the system?).
Power Module
Build the charging station consisting of two metal pads, electrically isolated, with a smart charger on the back side of the wall. For now assume you are charging the 11.1 V batteries we have in the lab. The design challenge is in robustly delivering the power to the robot. A few ideas are:
- the wall has telescoping power leads to reach the robot.
- the robot has two spring-loaded contacts (springs, ball bearings?) to slide along the wall and make contact with the pads.
- add mechanical magnetic clamps to the two pads to be able to support the weight of the gibbot.
- inductive charging (replace batteries with supercapacitors?)
Robot Frame
motor frame magnet must play nice with other modules
Mechanical: Select a material that can withstand falls of ~15 ft of a 10 lb robot. Select a motor that will allow the robot go uphill with weight of frame + batteries + electronics. Ideally the motor would be backdrivable. Is it possible to design a clutch mechanism? The links should be between 12-18 inches. The frame will most likely undergo a minor revision in the spring term once all other components are finalized. For the winter term, a base frame should be constructed that can test the motor such that the robot is capable of moving uphill. We can add weights inside the frame to test the expected payload.
Electrical: A motor model + motor control PCB module should be built. The module should be able to sense current and perform simple PID tracking.
Deliverables: Rugged frame (with padding inside for components?) that can withstand falling from sheet. With one link clamped (can use a vice with a rod sticking out to attach frame), frame should be able to use its motor to reach heights above the pivot link. PCB motor module with h-bridge, encoder circuit, and current sensing. PIC32 should have FF and FB model running (should have motor constant, Coulomb + viscous friction model, etc.).
Milestone: Frame (being) built and motor modeled. Give PCB redesigner (see below) the inputs and outputs of module.
Side Projects: Can frame deploy parachute for soft landing? Can a clutch be added? Should we pursue pancake motors? Use two motors at joint, one on each link, but with shafts coupled so only one needs to be active. The other is a generator while inactive, can we reuse the power generated to recharge the batteries?
Vision System
Build a vision system that can track the robot at all times. Assuming a steel sheet that is about 6 ft x 8 ft, the camera needs to be calibrated so that it outputs real-world coordinates. The major components of this project are:
- calibration of a wide-angle lens, which can be fitted with a special color filter
- robust wireless communication (IR, XBee, other)
- robust tracking
The calibration of the camera will involve selecting a camera and calibrating it as well as purchasing a PC to read in and transmit the camera images. The wireless communication should be a reliable link between the camera and robot. The vision tracking software should contain a suite of algorithms to track the robot on the wall. What is the best design that includes as many redundancies as possible? While the camera is being selected the tracking and communicating can be done with a webcam as a stand-in.
Main PCB Design
This project involves improving upon the existing PCB, for example, better connectors between components, moving the H-bridge onto the actual PCB board, and integration of modules from other groups. There are various important control and data signals running across the PCB, but the signals are not easily accessible. The redesign should include breaking out certain signals (e.g., sensors) for easy debugging on an oscilloscope. We also need a good characterization of the on-board sensors. How long can we rely on the sensors for tracking the robot before there is too much error accumulation. The PCB currently has a current sensor, position sensor, gyro, accelerometers, and an XBee onboard. Are these parts easy to maintain and replace? Are there better variants out there? We also want to record data to an external Flash card for data logging and send info back to PC through wireless. When transmitting to the PC, the user should have a way of (securely) controlling the robot remotely and querying it for useful information.
Electropermanent magnets
Can we design and build a rotary magnet similar to what is in the current Gibbot, but with electropermanent (EP) magnets? It would have to consume under 16 W and have a holding force in the ballpark of 300 N for the magnet to be a worthwhile replacement to the electromagnet.
Smart Wall
We need an alternative form of position sensing for the robot beyond a camera. An alternative that we believe would work is to wire LEDs (1- or 2-inches apart) on the steel sheet. The robot would then pick up a pulse-encoded signal from the LED telling the robot where it is. Constructing the smart wall isn't a problem, it's maintaining it. If we look at 2-inch spacing, then for a 48 sq. ft. wall we are looking at 1,728 LEDs. If one burns out, how will an MSI employee find the LED and replace it with ease? Can you come up with a better design for a backup sensor that does not rely on conventional wireless communication?
