2
Abstract-- A car that controls its self is a marvel. Naturally, one of the best ways to control the car is with a control system. The car is driven by differential steering meaning that there is a motor on each of the two driving wheels. The first step to analyze what the system needs to control the car; research was done on sensors, cameras, and power systems. Then research was done to determine how various sensors, cameras, and motor controllers worked and which combination would work best for the requirements of the project. After thorough research and a few design matrixes the best setup turned out to be ultrasonic sensors and cameras combined with an expensive motor controller. This gave us the best response time limiting the amount of overshoot occurs during turning.
Index Terms--The author shall provide up to 10 keywords (in alphabetical order) to help identify the major topics of the paper. The thesaurus of IEEE indexing keywords should be referenced prior to selecting the keywords to ensure that the words selected are acceptable. The thesaurus is posted at http://www.ieee.org/organizations/pubs/ani_prod/keywrd98.txt. It is also available by sending an e-mail (with no message necessary) to <mailto:>. A large ASCII file of the keywords (220 kb, approximately 55 pages) will be returned to you via e-mail. Before using this option, be sure your local e-mail has the capacity to receive a file as large as this one.
I. Nomenclature
A.D.D. – Automated Designated Driver
ABS – Antilock Braking System
II. Introduction
A
team of four electrical engineers and one computer science major are designing and building an automated vehicle. The name of their group is A.D.D. (Automated Designated Driver). In this paper they will be referred to as either A.D.D. or “the design team.” The Team goal is to enter into the IGVC (Intelligent Ground Vehicle Competition) and win. The paper contains how the team determined the design and parts of their control system that would control their automated car. The team has a budget that is near to $4500 dollars. Fig. 1 shows the team budget and how much money is available for some of the parts that are introduced in this paper. However, some of the items on the budget will not be talked about in this paper, but they give a general idea of where the money for the team is going to go. The first section will consist of a general analysis of how the main control system will work and what parts are necessary. The second part will consist of research done on the various parts needed and a comparison between the same groups of parts. The third section will be comparisons between the various combinations of parts from varying categories and the determination of the parts used. The fourth section will draw the selected parts together and explain in detail how they relate and how the system will interact.
III. System Overview
The control system used to design the car is fairly simple. The loop that I am concerned with in this paper is how the camera, proximity sensor, and PC will be combined to determine the output to the motor controller. This control system has some hardware, but the output is essentially derived from software driven PC. Fig. 1 shows the basic control system for the project. This control system will be expanded later in order to provide a higher level of detail.
Fig. 1. Original block diagram for the control system.
IV. Proximity Sensors
A. Proximity Sensor Uses
Proximity Sensors are needed for the automated car design because they will allow the car to detect obstacles. The price of these sensors are usually proportional to the distance they detect obstacles. Therefore, the sensors for this project will be quite expensive because moving at five mph the sensor must detect obstacles at a minimum of twenty feet in order for the car to have an adequate amount of time to react or turn.
B. Proximity Sensor Types
The search for a proximity sensor turned up three major results. The first result was the laser proximity sensor. These sensors are very effective. Right now they are made so that when someone buys them the software actually can show you a 3D image of what is all around it. This is very effective in determining where obstacles are and what the best path is between them. The downfall of these sensors is they are extremely expensive. They are on the order of 10’s of thousands of dollars. Due to the fact that the project does not have that big of a budget this sensor is excluded.
The next sensor that was researched was an infrared proximity sensor. These sensors are very good. Most of these sensors can only detect obstacles at a maximum of five to ten feet away. However, this is enough space to meet the requirements of our project. The cost of the PIR (Proximity Infrared) sensors is also within the allotted budget. This makes the PIR’s a competitor for the spot on the car.
The last sensor that was researched was an ultrasonic sensor. These sensors are very good at longer distances. They reach maximum distances of 50 to 100 feet. These sensors are acceptable for the range specified in the requirements of our project. The cost of the ultrasonic sensors is also within the allotted budget. This makes the ultrasonic sensors a viable alternative to the PIR sensors.
C. Proximity Sensor Comparisons
The laser sensors are not even going to be compared to the PIR and ultrasonic sensors because the laser sensor is automatically rejected due to its extreme costs that do not meet the allotted money for the project. The PIR sensor has a lower cost compared to the ultrasonic sensor. However, the ultrasonic sensor has a wider range of detection area then the PIR sensor. The PIR sensor essentially shoots one beam of infrared light like a common laser pointer, whereas the ultrasonic sensors shoot a sound wave comparable to the one shown in Fig. 1. The PIR laser is shown in Fig. 2. During testing if the PIR sensor was moved or touched the PIR would trigger detections. The ultrasonic sensor did not incorrectly trigger detections as it moved, only when there was something to detect did it trigger detections.
V. Cameras
A. Camera Uses
The automated car uses cameras as an integral part of the control system. This type of sensor can and will be used in obstacle detection. Two cameras will be used. They will be positioned above the car at approximately five to seven feet high. They will be positioned on the sides of the car with at the same angle in degrees so they both point to the same spot in front of the car. This will allow the cameras to view the whole track in which the car will drive.
A PC will be used to filter the color out of these pictures leaving only lines of obstacles that need to be avoided. These filtered pictures are shown below in Fig. 2. As you can see, the color picture is of the track and the filtered pictures show precisely where the obstacles are. The software that is implementing the filtering of the pictures will only be able to filter a certain number of images a second. Therefore, a camera for this project does not necessarily have to take over five or six pictures per second. Given the overview of how the cameras will be used the next step is to determine the types of cameras that the team may possibly use.
Fig. 2. Pre-filtered color pictures are shown and their corresponding filtered pictures.
B. Camera Types
The team did extensive research on many different types of digital cameras and concluded that there are three viable options that need to be compared and researched further. The three types are security, commercial, and web cameras.
The team researched security cameras. The security cameras are fairly large. They are also durable. Most of the security cameras can be dropped from varying distances without breaking. The cost of these cameras can range anywhere from 200 dollars to 2000 dollars. The refresh rate is proportional to the cost of the camera. If the camera costs 200 dollars then the refresh rate is slow. If the cost is around 2000 dollars then the camera refreshes fairly quickly.
Research was done on commercial digital cameras. The cameras that meet the requirements of the project without going excessively over cost at most 200 dollars. These cameras have varying picture qualities. If the car would need a higher quality picture this is one way to go. However, the car does not need that. The commercial cameras have a very slow refresh rate. They can literally only take a picture a second at the fastest times. This is mainly due to the auto focus and replay features of the digital camera. While these two features are activated the camera does not take a new picture.
Research was also done on web cameras. It was determined that the web cameras cost approximately 50 dollars. These cameras have a refresh rate of approximately 10 frames per second, which meets the requirement of the project. The speed does not need to be any faster because the computer can not process more then five two or three per second. These cameras are very small, but they are not very durable. If this camera is dropped there is a good possibility that it will break.
Seeing as all these cameras have benefits and drawbacks it is time to determine which camera is the best fit for the current project. The team does not want to get a camera with functionality they will not use, but they want to get a camera with all of the features they will need.
C. Camera Comparison
Looking back at the three cameras above we know that we can exclude the security cameras because in order to meet the requirements and meet the budget the least expensive camera with the least amount of capability would need to be chosen. With the least expensive set of security cameras the purchaser would need to build a case for it and also write some extra software in order for it to work. The purchaser would also have to build or order the wires and connectors in order to connect it to anything. This seems like excessive work considering that if the team gets any of the other two their labor hours will be cut down dramatically. So, security cameras are excluded from the list. Commercial cameras are also excluded automatically because there are no cameras that have a quick enough refresh rate that can run continuously. The quality of the camera creates too much of a delay between pictures at this time. That will change over the next few years so they might be compatible, but for now they are not. So, looking back over the comparisons it is obvious that the choice of cameras will be two web cameras. They are relatively inexpensive. They have a quick enough but do not have an excessive refresh rate. They are also durable enough to withstand the shaking of the car.
VI. Processor
A. Processor Uses
A processor is needed in our design because the signals coming from the proximity sensor and the cameras will not go directly to the motor control to determine the speed. The processor will accept these sensors as inputs and produce an output that the motor controller will be able to interpret. The design calls for a processor that will be capable of filtering the images, doing the triangulation of images to determine the distance of obstacles, and convert the input signals to the appropriate output signals for motor controls all in real time. The processor must also have many serial connectors available because that is how the data will be passed from the sensors to the PC and then from the PC to the motor controllers.
B. Processor Types
The first type of processor the team researched was a prototype board processor. This process gave the team the accurate speed of processing that was needed, but did not allow the team to easily receive drivers to install and use the cameras effectively. The amount of time the team would need to spend on recreate drivers that were already created was considered a waste of man hours and deemed unnecessary. This option also does not allow for a sturdy connection of the serial cables running from the various devices.
The next option is to use an older laptop with a minimum of a 2 GHz processor. This option would work effectively too. The laptop would be approximately eight hundred dollars. The laptop would allow the camera drivers to be installed and interfacing the laptop and the camera would take hardly any man hour at all. There are at most two serial connections on the back of a laptop.
The final option is for the team to build a desktop computer. This option gives the team great flexibility because they can purchase the exact components and pieces that they need in order to complete the project. With the team building their own desktop they can put as many serial connections in as needed.
C. Processor Comparison
Looking back at the three processor options above it is easy to exclude the prototype board processor due to the lack of expansion slots available. The man hours required to design this system and actually incorporate it into the overall system would be hard to justify. The cost of man hours exceeds the costs saved by buying this type of processor.
The closest comparison is between using a laptop or a self-built desktop. The laptop has the advantage of containing its own battery, which would mean that no AC voltage would be needed on the moving car. For a desktop PC the team would either need to buy a generator or a power supply with a back up voltage supply. The back up voltage supply is relatively inexpensive compared to the generator. The back up voltage supply also weighs less, which is also a consideration for this project since the heavier the car is the more powerful the motors need to be.
The laptop also lacks in the ability to be configured exactly the way the team wants it to be done. The team needs to have many serial connections, which can easily and affordably be incorporated into the design of their desktop PC.