T.R.O.N.
Transportional Regulation Obedient
Newbie
Dima Haddad
TAs:
William Dubel
Steven Pickles
Instructors:
A.A. Arroyo
E. M. Schwartz
Department of Electrical and Computer Engineering
EEL 5666
Intelligent Machines Design Laboratory
Table
of Contents
Abstract
Executive
Summary
Introduction
The project at hand is to build an autonomous intelligent machine that does a practical function. In this particular design the robot will have the ability to discern between the phases of a traffic signal. The main purpose of the design is to mimic a real-world application of a vehicle on a roadway.
There has been large scale and much more intricate projects that have successfully created an autonomous vehicle that can maneuver through signals and drive on freeways. This project will only concentrate on the aspect of recognizing and reacting to a traffic light.
The T.R.O.N.
project will entail several things of which the main parts will be the building
of the robot and its test environment.
The environment should be a miniature roadway model with a lane and
traffic signals. It will have a black
background representing the asphalt and white lines for the lane edges. The traffic signal should be overhead of the
vehicle as in the real world and will have the red, yellow, and green phase
lights. In the following paragraphs all
these aspects will be discussed in detail.
Platform
Platform Design
The platform will slightly resemble a vehicle. The platform consists of two main “T” shaped PVC board which is 6mm thick. The dimensions for these “T” boards are 6.75” in length by 5.125” in width. Figure 1 consists of AutoCAD drawings and renderings of the boards (Budget Robotics):
Figure 1: Robot Base Visual (Rigel, Budget Robotics)
Platform
Implementation
The T.R.ON. Robot will look like a miniature monster truck. The “T” base described would be the ideal platform due to the 60 sq inches of space that allows all the sensors and other equipment to be mounted on boards. Also since this is a vehicle robot the tires must be durable. The wheels that are mounted to the robot have the ability to traverse various terrains (such as, carpet, tile, grass, concrete, asphalt, and dirt.) The measurements for the complete body of the robot with the wheels attached are L: 6.75”, W: 6”, and H: 4.25”. Secondly since there are to main “T” section boards, the robot will have two levels, which will be separated by rises approximately 1 ¼” apart. The following figures are pictures of the assembly process for the robot body:
Figures 2, 3,
& 4: Periodic Pictures of robot
platform assembly.
Actuation
The robot platform will run on four wheels, each with its own individual servo. A view of the wheels and servos are shown below:
Figure 5: Bottom view of platform, with wheels and
servos (Budget Robotics).
The wheel and servo sets have the following specifications given by the retailer (Budget Robotics):
· Tire diameter: 65mm (2 1/2"); tread 7/8" wide
· Tire material: Medium-hardness treaded rubber (with "studs" for traction)
· Hub: Custom machined from PVC plastic
· Futaba-spline wheel hub, to match Futaba R/C servos
The R/C servos have been modified for continuous (360o) rotation and need any where between 4.8 to 6 volts of power to operate. In this project a regulated five volts will be used to power the servos as well as all other components on the robot. Another important point is that since the servos on opposite sides of the robot are mirror images of each other if the same pulse width is sent to both one side will go in one direction while the other side will go in the opposite direction, this should be taken into account when programming the servos for movement.
Other than the wheels the T.R.O.N. robot will not have other actuation components.
Sensors
There are several sensors that are need for the robot to accurately mimic a car on the roadway. The following is a list of sensors to be used along with some miscellaneous parts:
· Two IR Sensors
· One Sonar Sensor
· Four Bump Sensors
·
CMU
· Miscellaneous: buzzer and various leds
IR Sensors
There are two IR line tracking sensors to be used on the robot body. Each IR sensors will be placed on either side of the front of the robot about ½” off the ground. They will also be placed so the edge of the sensor board is further out than the robot body.
Figure 6: IR Sensor Mounting
The IR sensors will detect the difference between light and dark backgrounds. They go high for white surfaces and low for dark. The surfaces do not have to black and white. The IR sensors also work in various light conditions. They have been tested in only window light (daytime), window light plus intense light fixtures, and at night time with low intensity light bulbs that are not placed in the robots line of sight. The IR sensor has worked in all these situations. They also work if they are touching a surface to a little more than ½” away from a surface. The following are some schematics of the IR sensor used:
Figure 7: Diagram of IR sensor (Lynxmotion, Inc)
Figure 8: Circuit Schematic Diagram (Lynxmotion, Inc)
The IR sensor on the T.R.O.N. robot is used to avoid white lines on either side of the robot that represent a lane of traffic. When the right sensor detects the white line it causes the robot to turn left and vice versa for the left sensor. Secondly, since there are three IRs on each sensor, the closer the line is to the inner most sensor on one side, the more adjustment to the right or left the robot makes. After the adjustment is done the robot reverts back to a forward motion.
Sonar Sensor
The sonar sensor will be used for proximity detection so the robot does not collide with any other objects on the roadway. The robot only needs one sonar sensor due to the fact that the robot stops before an object and does not swerve into another lane or oncoming traffic. Once the sonar detects an object the robot stops and will not move until the object is removed.
The way the sonar works is it sends out a ping and waits for an echo to return and then measures the distance as the function of the time. The sonar used on the robot is a SRF04, the following list are the specifications of this sensor (Lynxmotion, Inc.):
· Sensor
type = Reflective Ultrasonic
·
Frequency = 40KHz
·
Ultrasonic sender = N1076
·
Ultrasonic receiver = N1081
·
I/O required = Two digital lines, 1 output, 1
input
·
Minimum range = Approximately 3cm
·
Maximum range = Approximately 3m
·
Sensitivity = Detects a 3cm diameter stick at
> 2m
·
Input trigger = 10uS Min. TTL level pulse
·
Echo Pulse = Positive TTL level signal, width
proportional to range
·
Input voltage = 5vdc regulated
·
Current requirements = 30mA Typ 50mA Max
·
PC board size = ~.75" x 1.75"
The sonar sensor on the T.R.O.N. robot is set to determine distance in inches. The minimum range for the sonar was 1” and the maximum range around 9 ft (108”).
The following charts are of the timing and beam pattern of the SRF04.
Figure 9: Timing Chart for Sonar SRF04
Figure 10: Beam Pattern for Sonar Sensor
Another collision avoidance tool will be the bump sensors which are used in the case that the robot does collide into an object or wall it will back up from that position.
CMU Camera
The CMU cam will be used for vision, so the robot can see the traffic light as well as which phase it is in (i.e. red, yellow, or green). Once the image is processed it will react accordingly, which would be stop for red, go for green, and double current speed for yellow. Below is a picture of the CMU cam from Seattle Robotics:
Figure 11:
CMU
The way the camera works: (Rowe, et al., 2002)
Upon completion of the
frame, it divides these accumulated values by the total number of pixels
returning the mean color. It also returns an approximation of the absolute
deviation from the mean of each color. This can be used like a variance measure
to quantify the spread of the colors about the mean. When used in conjunction
with other features such as windowing, described below, the color statistics
can be used as a building block for a motion detection algorithm or for
determining the color of an object at a specific location in the field of view.
Since the
robot is only going to be using the CMU cam to detect three colors the “GM\R”
command (or get mean color value) will be used.
The mean values produced by the camera are between the range of 16 to
240.
Miscellaneous
The miscellaneous section has parts in it that will allow the robot to look and be more car-like, such as having a horn and front and rear lights.
Conclusion
The T.R.O.N. robot will have the ability to drive in a lane of traffic and react to a traffic signal on the roadway in its environment. This behavior will be achieved through several sensors and a platform that is built for roadway maneuverability.
References
Budget
Robotic. Rigel Construction and
Operation. http://www.oricomtech.com/rigel/rgl-info.htm
Lynxmotion, Inc. Users Manual TRA-01 Version 5.0. & http://www.lynxmotion.com/Product.aspx?productID=57&CategoryID=8 &
http://www.lynxmotion.com/Product.aspx?productID=59&CategoryID=8
Rowe, et
al. A Low Cost Embedded Color Vision
System.