Cheezoid

Cheezoid is a robot which draws on a whiteboard. Think turtle graphics in meatspace. It was originally an Amplify Maker Days project but has since been open sourced under apache 2.0 license.

Overview

The idea is that cheezoid drives (vertically!) on a whiteboard drawing as it goes, with a whiteboard marker. There is no permanent attachment to the whiteboard, magnets are used to keep it in place. They are attached to the bottom of the cheezoid chassis, hover about 1mm from the board surface. Wheels are made or soft rubber, providing great traction while also soft enough to avoid damage to the board.

Cheezoid is has two controllers. One for the low level motor and encoder control, and the other for high level position tracking and drawing control. Low level controller is a PIC24H on a Microstick development board. It is hooked up to the motor driver directly, and uses a mixture of 3v (pwm/direction) and 5v(encoder feedback) signals. The low level controller receives commands over SPI protocol from the High Level Raspberry Pi B+ controller.

PIC development

Microchips issued tools are used for pic development. This includes:

• IDE / Programmer: MPLABX (v2.30)
• Compiler: xc16, aka c30 (v1.24)
• Part Support: xc16 part support (v1.24)

All of these can be downloaded from mplabx website.

After installing xc16 check out peripheral library documentation

RaspberryPi initialization

On RaspberryPI we're using Raspbian distro.

The following additional configuration has been performed through raspi-config:

• Enable SPI & drivers set to load on boot
• Enable I2C & drivers set to load on boot
• Enable Camera Module

Following packages have been installed:

• Creature comforts apt-get install vim git fish ipython python-setuptools python-dev
• SimpleCV & dependencies:
  • apt-get install ipython python-opencv python-scipy python-numpy python-setuptools python-pip
  • pip install https://github.com/sightmachine/SimpleCV/zipball/master
  • pip install svgwrite
• i2c requirements apt-get install python-smbus i2c-tools
• spi requirements pip install spidev
• RPIO pip install rpio

Additional OS config has been performed:

• /etc/hostname changed to cheezoid

• sshd password login disabled

• wpa_supplicant configured & wlan0 has been configured as follows:

  allow-hotplug wlan0
  iface wlan0 inet manual
  wpa-roam /etc/wpa_supplicant/wpa_supplicant.conf
  iface default inet dhcp
  

Cheezoid Hardware Platform V1

The first cheezoid hardware platform is a straight forward 4 wheel drive platform. Chassis is built from a .236" acrylic sheet cut to 4"x7.5", a hole is drilled in the middle for the whiteboard marker. And raspberry and motor driver are screwed down to standoffs made from the same acrylic sheet. All acrylic to acrylic joints are 'welded' with Weld-On 4. The rear standoffs are glued straight to the motor mounts using Devcon Plastic Welder (cream).

The tip of marker is attached to a sub-micro servo and is guided by a long acrylic standoff. The other end of the maker is zip tied to a rubber band which is attached to guide. This spring loads the marker in down position. Servo is use to lift the marker.

Initially the power was fed straight into raspberry pi, and in turn tapped out to motor driver over the breakout header. This turned out to be problematic Undoubtedly the raspberry regulator was not designed for upward of 2amp current draw. As a precautionary measure the power was then re-wired to go first into the motor driver, and raspberry tapped from there on. During normal operation PIC24H is powered from Raspberry's USB port, but during programming its connected to laptop. Programming PIC from raspberry, has not been investigated. Assumption is that it is not possible as microchip does not provide ARM compatible programming software.

Parts list:

• 4 drive units each made up of
  • microgear motor 50:1 later swapped for similar 210:1 microgear motors
  • dfrobot chassis encoder
  • wheels
  • mounts
• Dagu 4 channel motor controller driver
• Raspberry Pi B+
  • with wifi adapter
  • with camera (unused)
  • with prototype board
• Microstick development board.
  • PIC24HJ64GP502
• MPU-6050 3 axis gyroscope accelerometer
• 2.1Amp USB battery
• sub-micro servo

Lessons learned from platform 1:

• offloading motor control to pic is pretty good idea
  • Interrupt driven encoder feedback with painless time information from the input compare module is awesome!
  • straight forward PWM output with Output compare
  • Ran out of PWM output pins on pic, servo had to be driven from raspberry.
• Opto encoders
  • are rather annoying. Requiring constant adjusting.
  • If there is anything positive about this set up is that the motor driver comes with encoder mixing circuitry. Halving number on pins needed on pic for encoder feedback (unless you care about direction)
  • Since each encoder is tuned separately getting accurate timing information is impossible. Averaging could have helped but with just 48 events per wheel rotation would have made any corrective action too slow.
  • Ended up using only 2 of the 4 encoders. With the assumption that the 2 wheels on the same side behaved similarly.
  • Using the phase offset of encoders to determine direction proved to be unnecessary
• Drive system
  • 4 wheel drive skid steer is problematic in this application
    • Sideways friction is really high while cheezoid is attached to whiteboard easily 2-3x the force of gravity if its on horizontal surface
    • steering is not very accurate. We resorted to using the pen as a steering assist system. Push it down hard with the pen actuating servo to create a pivot point, and reduce wheel friction.
  • Gear reduced motors are quite awesome! Easy to change torque just swap the motor.
• Board attachment
  • Generally the board attachment using magnets works quite well. In particular
    • Having many small magnets is much better as it allows to modify magnetic force and distribute it differently.
    • gluing metal pads to the chassis to which the magnets were attached is much better than gluing magnets to chassis. Made it possible to adjust magnets in seconds.

Overall the biggest problem with chassis 1 was the drive train, the lack of precision in steering while attached to whiteboard made it impractical. All the other negatives were just small annoyances.

Next platform will be improved by:

• changing 4 wheel drive to 2 wheel drive with casters. The motors will be changed to higher power ones with even lower gear ratio. from 27 oz-in (108 oz-in total) at 60 RPM to 70 oz-in (140 oz-in total) at 100RPM. Biggest benefit will be to reduce friction during skid steer by not dragging the wheels sideways.
• Slightly smaller, skinnier wheels. Should increase torque and reduce smudging.
• swap opto encoders for hall-effect ones. The hall effect ones measure motor position rather than wheel position giving us many more events per wheel rotation. 48 events per wheel rotation to 7185 (most of these will be discarded by hardware and software tricks).
• replace big 4 channel motor driver board for a smaller package.