Bobette

Bobette must find a and extinguish a lit candle in a model home of 2,5 by 2,5 meters with four rooms. This robot must be able to navigate through an environment of known dimensions. It can use such information as distance to walls and white lines at door entrance to orient itself in the house. A digital compass may also be used to figure out the direction of the robot and possibly map the different rooms. The robot is equipped with several sensors such as infrared short range distance sensors, ultrasonic long range distance sensor, infrared sensors to detect white lines, flame detector, etc.

Bobette is an autonomous robot. Once activated, it does not require any human intervention. It uses the information provided by its sensors to navigate into its environment.

Bobette controls its two motors independently which allows it to steer itself in any direction. It can spin on itself by turning one wheel in one direction and the other wheel in the other direction at the same speed. To steer right, Bobette turns both motors forward, but activates the left motor faster than the right motor: the higher the difference between the speed of two motors, the sharper the angle of turn.

The program that controls Bobette is transferred from the PC to the microcontroller using an RS-232 interface (COM serial port). The program is "flashed" into memory, i.e. it is written in the microcontroller permanent memory, which is a special type of memory called flash memory. The microcontroller's flash memory is fairly small (8 Kb), so the program has to be fairly simple in complexity.

For its first version, Bobette used a finite state machine to navigate through the different rooms. Different landmarks were identified and the robot moved from one point to the other. There were also some parallel routines that performed a first order obstacle avoidance algorithm.

Here is a summary of the components used on the first version of Bobette:

Chassis	Old VCR frame used to form the base of the robot. Two 25 cm diameter discs were cut off the metal frame.
	Two 12V 100 RPM Hsiang Neng motors
	12V 4 Ah sealed lead acid battery
	Two wheels with a 7 cm diameter coming from an old Capsella kit
	One rollerball wheel placed in front
	Recycled CDROM reader motor and toy propeller fan to blow out candle
	Switch
	Nuts and bolts
	Meccano pieces
Sensors	Vector 2X digital compass
	Three Sharp GP2D12 infrared distance sensors
	Advantech SRF04 ultrasonic distance sensor
	Two QRB1114 infrared sensors to detect color variations on the floor
	OP805SL phototransitor to detect flame (by light intensity)
	Two thermistors to detect flame (by heat)
	Hitachi HD44780 LCD 2 rows X 40 characters
	Three pushbuttons (can be used as bumpers)
Main board	One circuit board (15 cm X 10 cm)
	5V 7805 1A voltage regulator
	PIC16F877 4 MHz microcontroller
	4 MHz crystal
	L293D H-bridge
	MAX232 RS232 Interface
	SN74LS05 hex inverter
	Two push buttons
	A dozen capacitors
	A dozen resistances
	Red LED to show that the baord is powered on
	4A fuse
	Phone plug (used as a RS-232 interface for serial communication between PC and PC)
	A dozen connectors to hook up sensors to the main board
Environmental board	Two LM741 opamp (one for temperature and one for sound detection)
	567 tone decoder (sound detector)
	Two 100K potentiometers
	LED to indicate detection of heat source
	1N4004 diode
	BS170 MOSFET to activate fan motor
	Half a dozen capacitors
	Ten resistance
	Two connectors
2-wire LCD interface	SN74LS174 Quad flip-flop memory
	20K potentiometer
	Diode
	Capacitor
	1K resistance

Bobette Details

Bobette Design Notes

• Mounting the wheels on the motor shaft: At first, I used a pressed wood support to mount the wheel on the motor shaft. Unfortunately, with friction, the wood around the shaft eroded and the wheels started slipping. Chris from ORE helped me design aluminum squares that I mounted on the wheels. The squares were attached to the motor shaft using a side screw.
• Sharp distance sensor: For the first 10 centimeters, the voltage goes up and then goes down. There is a band where the same voltage value gives two different readings. To circumvent this problem, the sensors where enclosed ten centimeters inside the robot.
• Ultrasonic distance sensor calibration: The version of the sensor that I had purchased had a manufacturing defect. After five minutes, the distance reported was not accurate anymore. Also, the width of the field of view was too large and the robot was picking up the side walls in addition to the front obstacles. To narrow the field of view, I installed sponges on each side of the sensor.
• Navigation: The robot was never able to follow the same series of events. Very often, it would get confused and lose track of where it was. Eventually, I added a digital compass (thanks to Mordechai) and was able to enhance the navigation aspects of the robot. The compass has its own problems (need for calibration, variable bearing reading in different places in the house), but was a great help.
• Size of program: The first software version reached the code space limit quite fast. I had to reconsider how I approached the problem because my first attempt required too much code space. I decided to implement a simple finite state machine infrastructure which reduced the code space and the complexity of the program.
• Circuit polarity: Two times I hooked up my circuit with reversed polarity. I lost several components this way and had to rebuild part of the main board once. I introduced some simple mechanisms that would prevent this from happening (add a switch to start the robot, add a fuse, add a special connector that forces to hook up the power plug the right way).
• Current management: The first time I hooked up all sensors, the robot died. The reason was that the power consumption was above the limit of the 7805. The robot used over 600 mA of current, yet the 7805 was rated for 0.5 A. I replaced the 7805 with one rated 1 A. I also noticed that my ground sensor was using way too much current. The reason was that I had used a 22 ohms resistance at the basis of the photdiode. This translated in the use of 250 mA per sensor. I replaced the resistance with a 100 ohms resistance which allowed me to save roughly 400 mA on my robot!
• Use of capacitors: It is recommended to add capacitors in strategic places in a circuit to avoid that a drop of current to cause an adverse effect like resetting the microcontroller. A 22 uF electrolytic capacitor is recommended at the base of the voltage regulator (between ground and Vcc). Tantalum capacitors are also recommended between ground and Vcc of different ICs, such as the microcontroller, the L293D, etc.
• Noisy motors: For my fan, I used a cheap Radio Shack motor at first. However, I discovered very early that this kind of motor can be very noisy and cause the microcontroller to reset. I ended up using a motor from an old CD-ROM drive. These motors are not noisy and won't cause interaction problems with other component.
• Tiredness: Long hours late at night took a toll on me. At the end of the project, this wasn't fun anymore. Unfortunately, there is no remedy for this condition. Patience and creativeness are key.

Bobette Bottom View

Events

Trinity College (2002)

Bobette participated in the 2002 Trinity College competition in Connecticut in April 2002. It was able to pass the qualification in the senior division. Only fifty of the hundred robots registered succeeded in the qualifications. To qualify, each team had to demonstrate that its robot could succeed blowing out a candle in any of the room of the model home.

In the competition, each robot had three runs. Bobette wasn't able to extinguish the candle in any of its three runs. In its first two runs, the candle was placed in the furthest room of the house. Bobette was able to visit each of the four rooms, but hit a wall before being able to see and/or detect the candle in the final room. In the third run, Bobette hit a wall head on halfway and wasn't able to resume its course afterward. Bobette ended thirty-first overall.

Eastern Canadian Robot Games (2002)

Here is a list of changes that were made to the robot in preparation for the ECRG competition held in Toronto in October 2002:

• Replaced the crystal and the microcontroller for 20 MHz versions
• Added an infrared distance sensor at the back of the robot
• Removed the digital compass
• Rewrote big chunks of the code

The results of this competition were much better than those at Trinity. Calibration of the robot was much easier than in Hartford. I had to tune the light intensity sensor (moving it up a couple of inches) and adjust the light intensity sensor and ground sensor thresholds, then fix a couple of small bugs and the robot was ready to roll. There were still some navigation problems that prevented the robot to locate the candle from time to time that I will need to fix before upcoming trials. Also, I had some problems with the light intensity sensor after I moved it up because the fan was blocking its field of view sometimes.

On the first day of the competition, Bobette was able to extinguish the candle flame on its first attempt. On its second attempt, Bobette went to the room where the candle was lit, but didn't find the candle. Either the light threshold was too low or the fan blocked the light intensity sensor. On the second day, on its third and final run, Bobette miserably crashed into a wall on its way into the first room (where the candle was). It could not move any further and thus did not succeed in extiguishing the fire.

Bobette finished fourth amongst the six robots that entered the competition.

Trinity College (2003)

• Replaces the propeller fan with a computer fan (with fancy LEDs patterns when spinning).
• Improvements in the navigation algorithm, including algorithm for return path.
• Addition of four LEDs to indicate status of sensors.
• Changed power feed. Now using 6 'AA's. Robot is much lighter.
• Added activation circuit for the ground sensors. This reduced the amount of current required for these sensors (from 200 mA to a few mA).
• Added a circuit to detect sound of an alarm.

Bobette was outstanding in this event. The qualification was much more difficult than the previous years, but this time around, Bobette qualified in its first attempt. From the sixty entries in the senior category, only thirty-two qualified for the event. The entries that qualified seemed stronger than the year before. In the actual competition, Bobette was able to detect and blow out the candle on its second run (in the most difficult room). In its third attempt, it was able to find the candle, but was never able to put it out because of a reflection problem (it focused on the reflection of the candle and could not position the fan to effciently blow away the candle). There is still more work to do on navigation, but Bobette behaves well in the maze. Bobette reached 23rd place in its category and was probably the lowest cost entry in the competition.

Notwithstanding the sad break in of my car, in which I lost all my tools and components, this event would have been a resounding success.

Future Projects

Lack of time has delayed upgrades to Bobette, but I shall soon resume its quest to stardom!

To reach me, send an email to: martin.dubuc@rogers.com.

Last updated: March 14, 2004