Keywords: Automated Highway Systems, Mobile Communications, Frequency Hopping.
This paper presents the development and results from field tests carried out on a frequency Hopping Spread Spectrum Wireless Communication system. This system is for real-time data transmission between moving vehicles and has been developed for the Ohio State University's Center for Intelligent Transportation Research (CITR) automated cars. A brief overview of Automated Highway Systems (AHS) and the role of wireless communication is presented in the beginning. This is followed by a description of the system requirements for AHS applications and factors leading to the selection of a particular commercial wireless modem. The hardware and software integration of the modem controller are discussed in the next section. The results of stationary and mobile tests of the final system, including the effect of frequency diversity with implications for AHS are presented in the end.
During the recent years a lot of research has been done on Automated Highway System (AHS). Automation involves operation of vehicles at freeway speeds without human intervention so as to optimize traffic throughput and decrease the frequency and intensity of accidents due to human errors. The AHS program at the Ohio State University, Columbus OH (OSU) way started in 1964. More recently, the Center of Intelligent Transportation Research (CITR) at the OSU is working towards the development of a prototype automated vehicle system. This prototype will be based on three autonomous vehicles that will be used in the 1997 Automated Highway System (AHS) Technology Demonstration and are general test-beds for safety, stability, and autonomy studies.
Wireless communication systems is a major sub-area of research for AHS applications. In order to achieve efficient and stable control, each vehicle must be able to exchange some of it's vital parameters with other vehicles operating autonomously nearby using a reliable wireless link. The variables of interest may include position coordinates, current speed, braking, acceleration, reading of various sensors and other information required for control. The proper selection of design factors such as data rate, modulation, range, and addressing are some of the important issues [1]. Furthermore, the spacing of vehicles on freeways for control in AHS has an impact on the design of the wireless system. For instance, autonomous vehicles may be grouped into platoons with small distances or individual vehicles may be uniformly spaced. The range specifications and network layer protocols are different for the two cases.
As mentioned before, the prototype system being developed at CITR-OSU would use three vehicles and requires a wireless link for data exchange. There is no specific grouping of these vehicles (no platooning). Data packets of a fixed length would have to be exchanged between each of these vehicles at a fixed update rate (figure 1a). The typical requirements, as specified by the control group include a data packet size of 100 bytes at an update frequency of 10Hz. Based on these specifications, we can calculate the minimum data rate or throughput required from the wireless link. Each of the three vehicles will transmit its data and then receive data from two other vehicles within its range in a TDMA (Time Division Multiple Access) fashion where a time slot is equal to a packet time (figure 1b). The minimum data rate for the system then comes out to be (3x100x10) 3000 bytes/sec or 24000 bps.
Figure 1a: 3 vehicles with communication links.
The AHS controls group at the Department of Electrical Engineering at OSU is using a dedicated computer for control and central data collection in every automated vehicle known as the ``AutoBox''. Within each car, the communication sub-system is required to interface to the Autobox using one of the available expansion slots and provide it with a send/receive type of transparent service.
For the application under consideration, this then suggests the use of a medium range mobile wireless system for each vehicle with omni-directional coverage. Each link would operate in a broadcast mode since the data is to be communicated to all receivers within range.
Figure 1b: Transmit and Receive cycles for 3 vehicles with TDMA.
Given the timing and financial constraints, it was decided that a commercial modem will be used to establish the wireless link. After a market survey, the vendor chosen was Pulse Inc. and the device selected is called the MTR2400PM modem. These wireless transceivers comply with FCC part 15 regulations by employing frequency hopping spread spectrum techniques. They can be made to operate on any 1 MHz channel from 2.400-2.498 GHz (Channel number: 0-98). This is done by loading the proper values into one of the PLL registers internal to the modem. The manufacturer specifies a data rate of 1Mbps with a Bit Error Rate (BER) of 1 in 100,000.
Frequency hopping Spread Spectrum (FHSS) systems change their carrier frequency (``hopping'') after fixed time intervals (``dwell-time''). The new carrier is selected from a set of randomly arranged carrier frequencies within a certain band. If the dwell time is smaller than the transmission symbol time (= 1/ Baud Rate), the system is classified as a fast hopping systems. This implies that the same symbol is transmitted over multiple carrier frequencies [4]. Figure 2a shows the channels vs time plot for a slow hopping device.
Figure 2a: Channel hopping vs Time
FHSS devices operating in 2.4-2.5 GHz ISM band must comply with FCC part 15.247 regulations. Some of these requirements have to be met by the user of the MTR2400PM modem. This includes the use of at least 75 hopping frequencies arranged pseudo-randomly, precise hopping synchronization between all the modems in the network and a maximum dwell time of 400 msec.
The MTR-2400PM modem is not a stand alone device. It requires a microprocessor based smart controller to implement the tasks of; Data Flow Control in order to exchange data with the modem in synchronous serial format, channel hopping by loading the proper values into a PLL register within the modem, hop synchronization of each modem in the network by maintaining and updating a high resolution hardware dwell timer within the respective microcontrollers etc. Figure 2b shows a functional block diagram for the controller.
Three devices available in the form of IC's play a key role in implementing the control functions. These include MC68H11F1 Micro-controller acting as a master, ESCC 85Z230 serial communications controller for handling data exchange with the modem and DS2010 First In First Out memory (FIFO) for data interface with the AutoBox. The software for performing the modem controller functions is programmed using MC68H11 assembly language. The use of a low level programming language helps in precise calculation of code execution delays and gives better control over the hardware. In addition to other tasks, software functions perform the tasks of initializing the ESCC controller chip for SDLC mode, loading appropriate control words into the PLL registers of MTR2400PM for initialization and channel hopping, acquiring and maintaining hop synchronization of the modems etc.
Figure 2b: MTR controller circuit board
Data is transmitted and received in the form of short packets consisting of 100 bytes of control information. In order to accomplish the tasks of data framing, SDLC (Synchronous Data Link Control) is used. SDLC protocol uses starting and ending flags to delineate the frame boundaries as well as a 16 bit CRC checksum for error detection. The MTR-2400PM modem also puts a header every time before transmitting a block of data in order to assist in data and clock recovery. The final frame format including the header appended by the modem as well as the SDLC frame headers and trailers, is shown in figure 3.
Figure 3: A MTR-2400PM modem frame enclosing a SDLC data packet.
The frequency hopping wireless system developed will be used to transfer data in a mobile environment. For the application at hand, we are interested in measuring the range as well as investigating the effect of multipath fading and time variation of the mobile channel with nearby traffic. We will treat the packet error rate as our figure of merit for performance characterization of the system. (For similar outdoor tests see [2] and [3]). The test equipment consisted of the following;
1. Two MTR-2400PM controller boards fitted with RF modems placed inside car #1 and car #2.
2. Two half wavelength monopole antennas mounted on the roof top of the test vehicles
3. Two 12V rechargeable batteries for powering the controller boards.
4. A pentium notebook computer with RS232 link to download the test program and upload the test data from the controller cards.
The test program when executed caused the two units to exchange N test data packets of 100 bytes. 75 frequencies were selected from the 99 available channels (2.400-2.498 GHz) and arranged in a pseudo-random sequence. After each Transmission or Reception the two units were made to hop in synchronization to the next channel in the sequence. This process was repeated until the required number of packets were exchanged. A packet was counted ``lost'' if the receiving modem failed to detect the frame header containing the synchronization and start word (see figure 3), whereas a packet with a CRC error in the payload was declared to be in error;
Packet Error rate = (lost + error)/N
The stationary tests consisted of parking one of the two vehicles at a fixed spot in an empty parking lot. The other car was moved away from the first car in steps of 7.4m and error rates were recorded by transmitting/receiving 2000 packets (N=2000). Line of sight (LOS) link was maintained between the two cars at all times.
Figure 4 shows the % Packet Error Rate plotted against distance. At each data point, the graph also shows the relative number of erroneous packets contributing to the total error. Taking error rate of 6% as a cut-off mark we can say that the range of the Pulse MTR-2400PM modem using an omni-directional antenna extends upto 140 meters. For the application at hand this is a reasonable value, since each car needs to maintain a link with at least the car immediately next to it.
Figure 4: Packet error rate vs distance
An important feature worth noting is that most packet lost results due to failure of the modem to synchronize. Once a valid frame is detected, very few packets show a CRC error in the data bytes. The low CRC error suggests that any post-processing on the data to extract valid data from incorrect packets will not give any appreciable gain in terms of the packet error rate.
Figure 4 suggests two fade points at a separation of roughly 35m and 75m. In order to investigate the effect of multipath fades another test was conducted. The separation was varied in steps of 5.0m, and decreased to 2.5m in areas of suspected fades.
Figure 5: Investigation of fades with FD=1 and 2.
At each step, in addition to the nominal error rate for 2000 packet as measured in the previous experiment, a frequency diversity factor (FD) of 2 was incorporated. This was done by transmitting the packet at two successive channel frequencies and recording an error if both attempts failed. The error rate with FD=2 was recorded for 1000 packets. Figure 5 shows the resulting plots with FD=1 & 2.
As expected, fades are recorded at a distance of 30.5m and 63.4m. The position of the first fade can be explained using a two path model as described in [5]. The earth is modeled as a perfect conductor acting as a good reflector for electromagnetic radiation and phase shifting by 180 degrees. Using the simple geometry of figure 6 and a wavelength of 12.24 cm corresponding to the center of the 2.4-2.5GHz band, we can determine the following fade positions;
1st fade: n = 1, d = 29.3m
2nd fade: n = 2, d = 14.5m
3rd fade: n = 3, d = 9.6m
It is seen that the first fade is very close to the predicted value. However packet error rates measured at a separation close to the 2nd and 3rd fades did not show any signs of fading. Both positions recorded an error rate of about 1%.
Figure 6: The two path model for Multipath Fading
This can be explained with reference to figure 7. The spread of the car roof is wide enough to block the reflected paths corresponding to the second, third and higher fades. Therefore there is only a line of sight path and no multipath fading is observed.
Figure 7: Absence of 2nd and 3rd fades
Another fade is seen at a distance of around 63.4m. This fade position cannot be explained using a two path model, however, this fade is also frequency selective because of the performance gain obtained using FD. The improvement resulting by using FD=2 is quite evident from figure 5; the error rate is less than 1.8% during the worst fades.
To investigate the effect of reflections from the roof top of a middle car, a simple experiment was conducted. A car was parked at about 10m from car #1 and perpendicular to the LOS link to eliminate the effect of reflection from the front hood and trunk. Car #2 was placed on the other side and moved away from the middle car. Packet error rate was monitored at different separations.
It was noted that when the distance of car #2 from the middle car was close to 10m, the packet error rates increased. The worst error rate of 10.6% (9.0% lost + 1.6% error) was recorded when the position of the middle car was equidistant from car #1 and car #2. Furthermore, it was noted that the use of frequency diversity did not result in any significant improvement. This suggests that this fade is non-frequency selective.
Figure 8: Fading due to a middle car
Using the two path model as discussed in [2] the car bodies are assumed to be good conductors, reflecting the incident microwaves with a phase shift of 180 degrees. Since the middle roof is at about the same height as the two antennas mounted on roof tops of the test cars, the reflected path has a path length very close to the LOS path (figure 8). When the middle car is in the center, the reflected radiation from its roof results in a strong destructive interference causing signal loss at the receiver. Moreover, using FD=2 does not help in improving the error rates, since all frequency components are reflected with a phase shift of 180 degrees which causes non-frequency selective fading.
The final application of the mobile wireless link developed involves operation at freeways and highways at speeds of upto 60mph. In order to fully test the performance an experiment was conducted in which both cars #1 and #2 were driven on highways in Columbus, Ohio.
Packet error rates were measured for 2000 packets for short, medium and long range. Error rates were also recorded for 1000 packets using a frequency diversity factor of 2. The results are presented below in Table 1. These tests were conducted in mild to medium volumes of traffic while maintaining a LOS path between the two cars.
| TRAFFIC | FD=1 | FD=2 | Short Range (20-30m) | None | 1.5 % | 1.0 % | Medium | 3.8 % | 3.4 % | Medium Range (30-60m) | None | 2.1 % | 1.4 % | Medium | 9.3 % | 5.1 % | Long Range (60-90m) | None | 6.6 % | 4.1 % | Medium | 10.2% | 6.6 % |
|---|
The following points are worth noting;
For a mobile transmitter (TX) and/or receiver (RX), the fade pattern changes over any given time interval. The extent of change depends on the relative motion of the TX/RX in terms of the channel wavelength in a packet duration. For a fixed packet size of 100 bytes, it is interesting to calculate the motion of the TX/RX during a packet time (~ 1ms) for the following two scenarios;
(lambda= 12.24cm, center channel 2.45GHz)
Vehicle Overtaking:
Relative motion at 10mph = 4.63 m/s
Motion during packet time = 4.63 x 0.001 = 1cm= 0.04 x lambda
Stationary Roadside Objects:
Relative motion at 60mph = 27.78 m/s
Motion during packet time = 27.78 x 0.001 = 6cm = 0.25 * lambda
The above calculations show that for maneuvers where one of the vehicles is overtaking the other, the fade pattern will not change much during a packet time. This is a case of slow fading [3] and we can incorporate frequency diversity by successively transmitting the same packet at two channels. However in the second case, the relative motion of the TX/RX in a packet time is a significant fraction of the wavelength and a single packet may be hit by multiple fades (``fast fading''). The implementation of frequency diversity in this case is not the same as transmitting the same packet simultaneously on two different channels.
Support for this research project was provided by the Center for Intelligent Transportation Research (CITR) at the Ohio State University. We would like to thank Prof. Umit Ozguner and Keith Redmill for their help and cooperation during the course of this project.
[1] Bret Foreman, A survey of wireless communications technologies for automated vehicle control, 1995 SAE Future Transportation Technology and Exposition, Systems and Issues in ITS, 400 Commonwealth Drive, Warrendale, PA 15096
[2] Chao Chen, Manjari Asawa and Bret Foreman, Outdoor Measurements on WaveLAN Radio, Path Lab, Department of EECS, U.C. Berkeley, December 15, 1995.
[3] J.S. Davis II and J.P.M.G. Linnartz, Vehicle to Vehicle RF Propagation Measurements, 28th Asilomar Conference on Signals, Systems and Computers, Pacific Grove, California, November 1994, pp 470-474.
[4] Ywh-Ren Tsai and Jin-Fu Chang, Using Frequency Hopping Spread Spectrum Technique to combat Multipath Interference in a Multiaccessing Environment, IEEE Transactions on Vehicular Technology, Volume 43, No. 2, May 1994.
[5] Chao Chen and Bret Foreman, A Discussion of the WaveLAN Radio As Relevant to Automated Vehicle Control Systems, Path Lab, Department of EECS, U.C. Berkeley, July 20, 1995.