UGCSbot technical description

UGCSbot will be an autonomous robot that among other things will be able to maneuver around the UGCS lab, acquire a printout from the printer, and deliver the printout to the workstation that requested it.

Mechanics
Mechanically, the robot will have two driving wheels in the front and two casters in the back. The 5cm radius wheels are coupled to PK-244AB stepper motors with 0.23Nm of torque through 3.3:1 planetary gear boxes. The robot can go forward, backward or turn by running the motors at different speeds.

It has space for mounting a battery, 6 circuit boards and a laptop, as well as the drive components. The front portion of the robot is dedicated to drive components and the back portion to electronics. There will be a partial Ferriday wall between the motors and electronics.

The base and structural components of the robot will be made from 1/4 inch polycarbonate plastic which is very strong, and the case and other cosmetic parts will be made from acrylic plastic which is more scratch-resistant. The robot will have a passive paper receptacle mounted on the top which can hold printouts it receives from the printer. We will couple (non-destructively) a special paper tray to the printer which can dump printouts onto the robot's passive tray when the robot is positioned correctly.

Electronics

The custom electronics for the system consist of the computer interface board, the motor controller board, the sensors board and a power supply board. We may have to have a separate board to handle the systems user interface.

• The computer interface board implements the ECP parallel port protocol and provides a simple I/O interface for other circuits. It also has an 8 channel, 8 bit, 0 to 2.5v ADC for sensory input. This board has been completed and thoroughly tested.
• The motor control board implements two 6 amp microcontroller-based stepper motor controllers with optical feedback. Feedback is implemented using sine/cosine optical encoders for each motor. These allow the microcontrollers to determine each motor's speed and direction and report that data back to the host computer. The board also has a current sensor for each coil which connects to the ADC on the computer interface board. This board has been completed, but the software for the microcontrollers does not implement feedback yet.
• The power supply board implements 3 functions.
  • It delivers primary power to the system from the battery when not connected to an external power supply, and delivers primary power from external sources when they are present. External power will be 16-20vDC.
  • When under external power, it charges the internal lead-acid battery and reports the status of the battery to the system. The system can enable or disable the battery charger and find the current voltage of the battery and external power.
  • The power supply circuit also contains a switching regulator to generate negative voltages (approx -9v) needed by some analog circuitry
  This circuit has been built and tested.
• The sensors board has circuitry for the other on-board sensors. This board has been build, but net you debugged.
• The miscilaneous board will contain reset circuitry, a watchdog timer, a servo motor controller, and possibly part or all of the user interface. The circuit will have direct control over the motors and other power equipment. It disables these devices on reset or when the watchdog timer expires. They can be re-enabled by the on-board computer. The user interface electronics will also be on this board if there is sufficient room.

  This circuit is still in the early design phase.

• We plan to use an IBM ThinkPad 570E with a wireless Ethernet card as the systems computer. The robot has a bay large enough to support a wide variety of laptops. The laptop will charge its battery from external power

• Optical encoders on each wheel to detect velocity.
• Accelorometers in all 3 axis. The accelorometers will use a hybrid analog/digital integrater to produce velocity with extreamly low drift.
• A Camera mounted on two servo motors giving it a 360x180 degree view of its environment. The robot will also use a low-power laser to project a grid that may be used for 3-d vision.
• Possibly a sonor system to detect objects very close to the robot.
• 64 sub-sub-miniature switches mounted in a ring around the robot's exterior to detect collisions.
• 4 internal thermometers.
• Microswitches to detect the status of the paper tray.

The major parts of the software are the vision, navigation and control code. We already have the kinematic analysis for a two-wheeled robot. I plan to write the control code during the first half of the term while I am working on the electronics.

The camera will be used for obstacle detection and landmark-based positioning. I plan to use the vision code as my CS148 project this year. We have not completed plans for the vision system and are currently considering different types of artificial landmarks that we could add to the lab to aid in landmark based positioning.

Regardless of the landmark system we use, the vision software will use will first convert the image from RGB color space to IHS color space. This separates intensity and and color information in a nonlinear way which is more sophisticaled than the YUV system. Next, it will use a standard gradient mechanism to find edges points then use the Hough transform together with a clustering alrorithm to find line segments. (Clusters of points in Hough space correspond to line segments in cartesian space.)

Navigation will be done using approximate cell decomposition on a global scale, and potential field methods on a local scale. The approximate cell decomposition allows global path planning from the robot's current position to the goal position taking into account static obstacles and dynamic obstacles recently encountered, while the potential field allows immediate obstacle avoidance and is used to travel between cells. The initial cell decomposition can be such that every point of interest (workstations, printer, etc.) has an associated cell in the workspace, and the decomposition can be refined as changes in the workspace are encountered.

references
C. Guerra. "Vision and Image Processing Algorith." Algorithms and Theory of Computation Handbook. Ed. Mikhail J. Atallah. CRC press, 1999. Chapter 22.
"RGB-IHS Conversion."  < http://www.pcigeomatics.com/cgi-bin/pcihlp/IHS|ALGORITHM >