An Architecture for Intelligent Automotive Collision Avoidance Systems

Size: px

Start display at page:

Download "An Architecture for Intelligent Automotive Collision Avoidance Systems"

Nathaniel West
6 years ago
Views:

1 IVSS-2003-UMS-07 An Architecture for Intelligent Automotive Collision Avoidance Systems Syed Masud Mahmud and Shobhit Shanker Department of Electrical and Computer Engineering, Wayne State University, Detroit, MI ABSTRACT Since 1960 s significant progress has been made with regard to vehicle safety. In spite of this progress, each year in the United States, motor vehicle crashes still account for approximately 40,000 deaths, more than three million injuries, and over $130 billion in financial losses [1]. Development of Automotive Collision Warning Systems (ACWS) is the next significant step towards vehicle safety technology. Such systems try to warn the driver about an impending collision. Most of the proposed ACWS are based on laser or radar technology, which requires line of sight communication. During the last several years, wireless technology has drawn significant interest among the people in both industry and academia. Interest in Wi-Fi and Bluetooth technology has also increased significantly during the last several years. GPS systems are increasingly becoming popular for vehicle systems. In this paper, we have proposed an architecture (using GPS, electronic compass and Bluetooth devices) for Automotive Collision Avoidance Systems (ACAS). Our main goal was to theoretically study the feasibility of using GPS, electronic compass and Bluetooth devices towards the development of ACAS. Our current work deals with keeping a vehicle within safe distances from other vehicles around it. Our theoretical study shows that it is doable INTRODUCTION During the last several decades, the improvements in seat belts, air bags, crash zones, and lighting have significantly reduced the rate of crashes, injuries and fatalities. The fatality rate per hundred million vehicle miles traveled has fallen from 5.5 to 1.7 in the period from the mid-1960s to 1994 [1]. In spite of these improvements, each year in the United States, motor vehicle crashes still account for about 40,000 deaths, more than three million injuries, and over $130 billion in financial losses [1]. By continuing with the passive safety technologies, it will be difficult to achieve significant further gains in reducing crash costs. Thus, the engineers and researchers have started developing active safety technologies to further reduce the costs involved with vehicle crashes. The introduction of automotive Collision Warning Systems represents the next significant leap in vehicle safety technology. Such systems try to warn drivers of an impending collision event, allowing the driver adequate time to take appropriate corrective actions. Crash statistics and numerical analysis strongly suggest that such collision warning systems will be effective [1]. Crash data collected by the U.S. National 183 Highway Traffic Safety Administration (NHTSA) show that approximately 88% of rear-end collisions are caused by driver inattention and following too closely. Modeling and simulation results show that head-way detection systems can theoretically prevent 37% to 74% of all police reported rearend crashes [1]. The goal of the Automotive Collision Avoidance Systems (ACAS) is to detect and warn the driver of potential hazard conditions in the forward, side, and rear regions of the vehicle. Most of the current work in ACAS are related to the use of: (1) long range radar or optical sensors to detect potential hazards in front of the vehicle, (2) short range sensors to warn the driver of nearby objects when changing traffic lanes or backing up, and (3) a lane detection system to alert the driver when the vehicle deviates from the intended traffic lane. The short and long-range radar and optical sensors require line of site communications. Thus, using radar and optical sensors, it would be difficult to detect another vehicle that is approaching a subject vehicle from behind an obstacle such as a tree, bush or building. But, if wireless communication technology is used, then a vehicle can easily detect all other surrounding vehicles regardless of the fact whether or not those surrounding vehicles are behind an obstacle. During the last several years, interest in using wireless communication technologies have grown significantly. Bluetooth features are becoming very common in cell phones, PDAs, laptops, etc. Automotive industry also has started introducing Bluetooth technologies in vehicles. The Bluetooth-enabled cellphone fitted in the 2003 Saab 9-3 car can access any other Bluetooth-enabled device in the car, such as a PDA [2]. Global Positioning System (GPS) technologies have also becoming popular in many applications including automotive applications. More and more vehicles are coming up with built-in GPS features, electronic compass, and other electronic devices. We hope that, in the future, Bluetooth enabled devices, GPS receivers, and electronic compass may become standard electronic devices for almost all the vehicles. With this hope in mind and the ability of a wireless device to setup omni-directional communications with other nearby wireless devices, we have proposed an architecture for Automotive Collision Avoidance Systems (ACAS). In our architecture, a vehicle collects data from the GPS receiver, electronic compass, accelerometer, speed sensor, etc., and then exchanges that set of data with other neighboring vehicles in order to warn drivers about an impending collision. The successful operation of an ACAS

2 will depend on how accurately the distance between vehicles can be measured and how fast the set of data can be exchanged among the vehicles. Some other researchers are also trying to use GPS and wireless technologies to develop collision warning and cooperative driving systems [3], [4]. In the following section of this paper we have presented some background materials on GPS and Bluetooth technologies. In the other sections of the paper we have presented our architecture, a description of the Protocol and Algorithm of our ACAS, and the conclusion. BACKGROUND MATERIALS ON GPS AND BLUETOOTH TECHNOLOGIES Standard GPS: GPS receivers use timing signals from at least four satellites to establish a position. Each timing signal indicates how much time the signal took to travel from the corresponding satellite to the receiver. The distance between the satellite and the GPS receiver is then determined by multiplying the travel time by the velocity of light. Each of those timing signals is going to have some error or delay due to various conditions such as satellite clocks, conditions of the earth s atmosphere, receiver noise, multi-path effect, etc, as shown in Table I. Since each of the timing signals that go into a position calculation has some error, the error in position is going to be a compounding of those errors. The total error in the position measurement can be as high as 100 meters [5]. Table I: Summary of GPS Error Sources [6]. Typical Error in Meters (per satellite) Standard GPS Differential GPS Satellite Clocks Orbit Errors Ionosphere Troposphere Receiver Noise Multipath Differential GPS: Errors introduced due to some of the sources can be eliminated using Differential GPS (DGPS) technique. Differential GPS involves the cooperation of two receivers. One receiver is located at a fixed point, and the other one is roving around making position measurements. The stationary receiver is the key. It compares all the satellite measurements with a fixed local reference in order to determine the timing error in the signals. These timing errors are then sent to the other receiver that is roving. The other receiver then compensates these errors from its own reading of the timing signals. As a result, the position measurement of the roving receiver becomes more accurate. Differential GPS can eliminate all errors that are common to both the reference receiver and the roving receiver. These include everything except multi-path errors (because they occur right around the receiver) and any receiver errors (because they're unique to the receiver). Bluetooth Technology: Bluetooth is a new technology standard using radio links. Unlike Infrared devices, Bluetooth devices do not require line of sight when transmitting. Bluetooth implementations support a range of roughly 10 meters, and throughput up to 721 Kbps. Long-range Bluetooth covers roughly 100 meters. Bluetooth enabled electronic devices connect and communicate wirelessly via networks called piconets. Each unit can simultaneously communicate with up to seven other units per piconet. One unit acts as the master of the piconet, whereas the other unit(s) acts as slave(s). Slaves can participate in different piconets on a time-division multiplex basis. A master in one piconet can be a slave in another piconet. The piconets are established dynamically and automatically as Bluetooth devices enter and leave the radio proximity. In the United States and Europe, the frequency range is 2,400 to 2,483.5 MHz, with 79 1-MHz RF channels. A data channel hops randomly 1,600 times per second. Each channel is divided into time slots 625 microseconds long. The master transmits in even time slots and the slaves transmit in odd time slots. THE PROPOSED ARCHITECTURE FOR AUTOMOTIVE COLLISION AVOIDANCE SYSTEMS In this section of the paper we have discussed our architecture for Automotive Collision Avoidance Systems (ACAS). In this architecture, we assumed that each vehicle is equipped with at least one Bluetooth device, a GPS receiver, an electronic compass, a wired network (e.g. CAN), and a number of sensors interfaced to the CAN bus. Though, currently every vehicle is not equipped with all the aforementioned devices, we hope that in the future almost all vehicles may have these devices built into their systems. Thus, our proposed architecture may be useful in the future. 184

3 the SV and OV. Though the directions of the vehicles can be determined using the GPS readings, we also decided to use the readings of the electronic compasses of both vehicles in order to accurately determine the directions of the vehicles. For a given OV, its RB around an SV is determined using speeds, accelerations, positions and directions of both the SV and OV. Figure 1: The Automotive Collision Avoidance Systems Architecture In our proposed architecture, a Bluetooth device is connected to the CAN bus of a vehicle using a CAN-BT Gateway interface. The vehicles that are close to each other on a highway or freeway will automatically form an ad hoc wireless network among themselves. A vehicle will automatically be disconnected from the network once it moves away from the other vehicles. Similarly when a new vehicle comes close to a group of vehicles, that vehicle will automatically be connected to this group through the wireless network. The Bluetooth device of a vehicle can collect (through the CAN bus) vehicle s speed, acceleration, direction, position and the status of the brake, gas paddle and steering wheel. The vehicle can then send this information to other vehicles through the wireless network. Since the success of our architecture depends on the accurate measurement of the distance between the subject vehicle and each one of the other object vehicles, we are presenting some ideas about how the distances can be measure with better accuracy. One option is to use differential GPS (DGPS) technique. However, in order to use DGPS we need to have many fixed ground stations distributed all over the highway and freeway systems. And, if a ground station is not available to cover a certain part of the highway and freeway systems, then DGPS technique can t be used. In this section we have proposed a technique that is similar to DGPS technique, but our technique doesn t require fixed ground stations. Figures 3-7 explain the technique of locating a GPS receiver on the earth. In order to make the illustrations simple, we are explaining the techniques for a 2-dimensional space. The same analogy applies for a 3-dimensional space. Figure 3 shows a GPS receiver and two satellites. By measuring the distance between the GPS receiver and the two satellites we can determine that there are two possible locations (points A and B shown in Figure 3) for the GPS receiver. These two locations are the intersection points of the two circles shown in Figure 3. But one of these two locations, point B, is at a ridiculously far distance. Thus, point A is selected as the location for the GPS receiver. Figure 2: An OV and its RB around an SV traveling in the same direction: (a) both vehicles are in the same lane, (b) the SV and OV are in adjacent lanes but close to each other. Our current work deals with keeping a subject vehicle (SV) within safe distances from other object vehicles (OV) around it, which are also moving in the same direction along with the SV. Our current work also deals with issuing a warning to the driver of the SV, if the SV tries to change lanes while there are OVs either in the blind spots or too close to the SV. For every OV within the neighborhood of the SV, we define a boundary around the SV called the Restricted Boundary (RB), as shown in Figure 2. The OV is not allowed to cross the RB around the SV. If an OV is detected inside the RB around the SV, then the driver of the SV is warned about the presence of an OV inside the RB. In order to determine whether or not an OV is inside the RB around an SV, we use readings from GPS receivers and electronic compasses of both 185 Figure 3: Two-dimensional analogy for determining the position of a GPS receiver. The distance between a GPS receiver and a satellite is measured by measuring the time of flight of the GPS signal from the satellite to the GPS receiver. But a number of errors are introduced in the time of flight due to various conditions shown in Table I. Due to the presence of errors in the time of flight of the signals from the satellites to the GPS receiver, the calculated position of the GPS receiver becomes different from the actual position as shown in Figure 4.

4 Figure 4: Error in the position of a GPS receiver due to errors in the time of flight of the signals from the satellites to the GPS receiver. Since we are showing examples in a 2-dimensional space, the distance between the GPS receiver and a third satellite can be used to detect the presence of measurement errors. Note that in the real world the position of a GPS receiver is determined by taking measurements from four satellites. Figure 5 shows that if there are errors in the time of flight of GPS signals, then there are three calculated positions for the GPS receiver. The processor in the GPS receiver then tries to minimize errors by making corrections in such as way that the three calculated positions come as close to each other as possible. Figure 6 shows the actual position, calculated positions before correction and calculated position after correction. identical. Using geometrical analysis we can show that the distance between the actual positions of the receivers will be almost same as the distance between the corresponding calculated positions of the two receivers. Even though the error (e 1 or e 2 shown in Figure 7) between the actual and calculated (after correction) positions of a GPS receiver can be very high (as high as 100 meters), we believe that due to the symmetry of the positions (actual and calculated) of the GPS receivers, the distance between the actual positions of the two receivers will be almost same as the distance between their calculated positions. As a result, the error e (shown in Figure 7) in distance measurement will be very small compared to e 1 and e 2, i.e e<<e 1 and e<<e 2. Thus, the key point that we have to pay attention to, in order to measure the distance between two receivers, is that both receivers must use the same set of satellites to determine their positions. For all the vehicles to use the same set of satellites, the vehicles should exchange information among themselves regarding which set of satellites they are using. Figure 6: Actual and calculated positions of the GPS receiver. Figure 5: Measurements from three satellites under error and error-free conditions. (CWE: Circle with errors. CWNE: Circle with no errors.) For Automotive Collision Avoidance Systems (ACAS), we are not interested in knowing the absolute positions of the vehicles. We are interested in knowing their relative positions. We believe that the relative locations can be determined more accurately than the absolute locations. Figure 7 shows the position measurements of two GPS receivers. For ACAS, since the distance between two GPS receivers (two vehicles) is very small (maximum feet) compared to the distance between the receivers and the satellites, the size of the measurement circles for both receivers will be almost Figure 7: Distance between two GPS receivers (e<<e1 and e<<e2). PROTOCOL AND ALGORITHM FOR ACAS Each vehicle can keep track of another seven vehicles, in realtime, using its Bluetooth device. The master device uses even numbered time slots and the slaves use odd numbered time 186

5 slots for communications. The master can exchange information with a particular slave at least once during every 14 time slots. The duration of a time slot is 625 microseconds. Thus, the master can exchange information with a particular slave at least once in every 8.75 milliseconds. Sometimes messages may need to be retransmitted due to noise or other kinds of interference. Probably we can safely assume that the vehicles will be able to exchange messages once every 10 msec, i.e. 100 times per second. Note that a vehicle traveling at 75 mph can move only 1.1 feet in 10 msec. Thus, the rate at which the vehicles can exchange information is fast enough for collision avoidance systems. The only bottleneck will be the rate at which data can be collected from GPS receivers. Most GPS receivers collect data at the rate of 1 Hz, but highspeed GPS receivers are also available that collect data at the rate of 10 Hz. If 10 Hz GPS receivers are used, then the vehicles can exchange position information once every 100ms. Between two successive GPS readings the vehicles can exchange speed, acceleration, direction, and the status of brake, gas paddle and steering wheel 10 times. After a new GPS reading is obtained, a vehicle can predict (using speed, acceleration and direction information) the positions of the neighboring object vehicles, every 10 msec. This way, the vehicles can also predict their locations at the time of the next GPS reading. If there is a difference between the predicted location of a vehicle, at the time of the next GPS reading, and the actual reading from the vehicle s GPS receiver, then some adjustment can be made to refine the prediction algorithm. Message Format: The message that will be exchanged among vehicles should contain the following parameters: Table II: Parameters of a Message Parameter Recommended Size Control Field 2 bytes Speed Acceleration Direction 2 bytes Brake Status Gas Paddle Status Steering Wheel Status Satellite Information 8 bytes Location in Longitude 5 bytes Location in Latitude 5 bytes Miscellaneous Data 23 bytes The basic ACAS will warn the driver of an impending collision. However, advanced ACAS can be designed by incorporating the basic ACAS with drive-by-wire technology. The advanced ACAS will warn the driver as well as help the driver in maneuvering the vehicle automatically. CONCLUSION In this paper we studied the feasibility of using Bluetooth devices along with GPS receivers to develop an Automotive Collision Avoidance System. We determined that the bandwidth available from Bluetooth devices is sufficient to exchange all vehicle specific information in real-time. We believe that if the subject vehicle and all object vehicles around it use the same set of satellites, then distance between the vehicles can be determined with a greater accuracy. Figure 8: Predicted values of a vehicle s position between successive GPS readings. Since a subject vehicle will have to keep track of several object vehicles in real-time, the vehicle s computer will have to do lots of computations. Every 10 msec the vehicle will have to work with new sets of data. However, we can efficiently use the vehicle s computation power if we ask the vehicle s processor to do computation on demand. If a subject vehicle detects that the relative speed of an object vehicle is almost zero and no other driving conditions is changing, then the subject vehicle can reduce the rate of computation for this object vehicle. But when certain driving conditions change in the object vehicle, such as pressing the brake paddle, removing foot from the gas paddle, significantly turning the steering wheel, etc., then the subject vehicle should resume processing the object vehicle s data at the normal computation rate. 187 ACKNOWLEDGMENTS We would like to thank Mr. Mohammed Tarique for helping us in doing literature search and participating with us during the initial discussions about various aspects of collision avoidance. REFERENCES 1. P.L. Zador; S.A. Krawchuk; R.B. Voas, "Final Report -- Automotive Collision Avoidance System (ACAS) Program", Performed by Delphi-Delco Electronic Systems, Contract No: DTNH22-95-H-07162, Washington, DC, August 2000, DOT HS Top 10 Techno-Cool Cars, IEEE Spectrum, February 2003, pp Ronald Miller and Qingfeng Huang, An Adaptive Peerto-Peer Collision Warning System, Proceedings of IEEE Vehicle Technology Conference (Spring), (Birmingham, Alabama), May Shin Kato, Sadayuki Tsugawa, Kiyohito Tokuda, Takeshi Matsui, and Haruki Fujii, Vehicle Control Algorithms for Cooperative Driving With Automated Vehicles and

6 Intervehicle Communications, IEEE Trans. on Intelligent Transportation Systems, Vol. 3, No. 3, September 2002, pp The Global Positioning System, Website of U.S. Geological Survey, 6. About GPS Technology, Website of Trimble, Dr. Syed Masud Mahmud has received his Ph.D. degree in Electrical Engineering from the University of Washington, Seattle. He published over 60 technical papers in refereed journals and conference. In 2002 he received the President s Teaching Excellence Award of Wayne State University. He has been listed in a number of Who s Who. Phone: (313) , smahmud@eng.wayne.edu. CONTACT 188

An Intelligent Architecture for Issuing Intersection Collision Warnings

IVSS-2004-UAS-03 An Intelligent Architecture for Issuing Intersection Collision Warnings Srinivas R Mosra, Shobhit Shanker and Syed Masud Mahmud Electrical and Computer Engineering Department, Wayne State