Squirt is a small, autonomous plant watering robot, designed to operate in an indoor environment, locating plants and bringing water to them. This design was created for EEL 5666: Intelligent Machine Design Laboratory. More information about the course can be found at the course homepage. All source code can be found in the form of snapshots throughout the weekly updates, or within the attachments section at the bottom of the page. Schematics, artwork prints, and a BOM are included for each of the custom printed circuits. By the end of 2009, a formal write-up will be posted that integrates the information here, as well as extends it with more pictures, flowcharts, and algorithmic explanations. Week 16: December 10, 2009 Tuesday, December 8. Final (for the course) code snapshots uploaded, along with formal report (which includes same). Today is Media day - Monday, December 7 I have a functional water squirting system, as well as a water guard for the camera, and face-paint. Here are the latest photos.   I have tweaked the control system parameters for a smooth floor, as well as taken the training wheels up ~0.5mm. The front panel is now attached with three thumb nuts for quick access to the guts. Wires that were interacting with the drive train have been re-routed and removed. The squirt nozzle and reservoir used to be attached in the form of a 250ml wash bottle. Now, an automotive windshield washer pump pushes a ~20mL, one-second stream of water ~2 feet in front of squirt, while strobing blue and white lights at 16Hz. Squirt will operate for around 1 hour of testing on the 1350mAh 3S Lipo battery. Week 15: December 3, 2009 An initial target has been fabricated for the vision system. It consists of a jar of peanut butter wrapped with two layers of bright pink post-it notes. The target is illuminated with high lumen white and blue LEDs. The color is such that false positives are unusual and can be eliminated from a test environment. The new illuminators can be seen below on either side of the avrCam imager being fed by yellow and black wires. Also, a wider angle lens was added in order to aid in target acquisition and approach. A new behavior was generated that gradually approaches the jar of peanut butter. The robot then aligns itself with the jar and triggers a water dispensing operation. The water pump has not yet been received, so the operation is currently reflected symbolically by flickering of the illuminators. The object avoidance and pink approach behaviors have been fused. The robot now wanders about, until the pink jar is spotted. The jar is approached, and symbolically sprayed with water. The robot then proceeds past the jar, avoiding collision and searching for another target.   Week 14: November 26, 2009 Double-sampling plant detection working. Code snapshot below. Additional illumination added to assist with plant identification. Autonomous recognition of plants with color blob tracking on RGB555 image data does not appear to be feasible within the timeline. This method has been abandoned, in lieu of tracking a bright pink color. Week 13: November 19, 2009 Website generated, and fuse installed. All electronics and mechanical structures mounted. Power switch and fuse installed. Object avoidance behavior tweaked.   Week 12: November 12, 2009 Object avoidance initial code in place. First Object Avoidance Test Week 11: November 5, 2009 Initial recognition of plants. Proposal of more advanced plant detection scheme, and an overall progress update. Week 10: October 29, 2009 IR Range sensors calibrated. Overview of the firmware framework. Snapshot of the early firmware Week 9: October 22, 2009 Working processor and motor controller assemblies. Powertrain functioning with shaft stall initialization and an angular velocity control loop. IR range sensor coprocessor sampling range sensors. AvrCam constructed, tested, and mechanically mounted. IR Range co-processor source code Week 8: October 15, 2009 Custom printed circuit boards designed and ordered. Schematic, artwork, and BOM for motor controller assembly Week 7: October 8, 2009 Initial work on control logic for powertrain. PCB design initiated. Week 6: October 1, 2009 First revision mechanical assembly completed Note: Please check the G code files. I wrote these quickly and often interacted with the CNC mill, overriding the programs Week 5: September 24, 2009 Mechanical design photos, and picture of one completed powertrain.  Week 4: September 17, 2009 Updates to many small tasks, including mechanical design, parts orders, and more MATLAB modelling. Project Proposal for the "Plant Watering Robot" First formal report Week 3: September 10, 2009 Control system modelling in MATLAB. Full source code is included in the report. Week 2: September 3, 2009 Early Project Ideas, initial SolidWorks mechanical design, and initial budget   |