Performance of UTRA TDD Ad Hoc and IEEE b in Vehicular Environments

Transcription

1 Performance of UTRA TDD Ad Hoc and IEEE b in Vehicular Environments Andre Ebner, Hermann Rohling and Lars Wischhof Technical University of Hamburg-Harburg Department of Telecommunications Eissendorfer Strasse 40, Hamburg, Germany Rüdiger Halfmann and Matthias Lott Siemens AG Information and Communication Mobile Gustav-Heinemann-Ring 115, München, Germany Abstract The FleetNet project aims at the development of a wireless ad hoc network for Inter-Vehicle Communications (IVC). As a basis for the air-interface, the framework of the UMTS Terrestrial Radio Access Time Division Duplex (UTRA TDD) has been selected as the most promising candidate. Since UTRA TDD was developed for operation in a cellular network structure, modifications are required to enable mobile nodes to communicate in ad hoc mode without the existence of base stations. In particular, this comprises changes to physical (PHY) layer, Medium Access Control (MAC) and Radio Resource Management (RRM). The focus of this paper is the PHY layer of the air-interface and the resulting challenges in highly dynamic vehicular environments. The performance of the UTRA TDD ad hoc mode is assessed and compared to the IEEE b standard using linklevel simulations. Results indicate that the UTRA TDD ad hoc PHY outperforms IEEE b in a typical highway scenario with very large relative velocities. In urban traffic environments with strong multipath propagation, IEEE b is not able to meet the required performance in terms of packet losses. I. INTRODUCTION Communication into and between vehicles for Intelligent Transport Systems (ITS) has attracted major attention during the last few years. Typical applications range from emergency notifications in cases of accidents, the distribution of Decentralized Floating Car Data (DFCD) for a Self-Organizing Traffic Information System (SOTIS) [1] [2], to more advanced applications like co-operative driving. For these applications, time and safety requirements cannot be fulfilled by existing air interfaces relying on a cellular network structure (e.g. GSM). Furthermore, by using ad hoc wireless communication, typical established Internet applications like web browsing, or chat applications can be provided without being bound to a costly cellular wireless network infrastructure [3]. This paper is organized as follows: In Section II, the challenges and requirements to the physical layer in a highly mobile ad hoc network are explained. In Section III, the physical layer of UTRA TDD ad hoc and IEEE b are briefly described. In Section IV, the performance of both systems is compared using link-level simulations, followed by final conclusions in Section V. The project FleetNet-Internet on the Road is partly funded by the German Ministry of Education and Research under contract number 01AK025. II. CHALLENGES The FleetNet project aims at developing an ad hoc radio network for inter-vehicle communications. Therefore, the PHY layer of the system has to meet certain requirements to facilitate wireless communication in a highly dynamic ad hoc network: Support of various data bit rates: While emergency notifications require only a few bytes for signalization, services like Internet access may require bit rates of more than 100 kbit/s. Hence, the FleetNet PHY has to provide a large range of data bit rates. Robustness in difficult propagation situations: For vehicles driving in different directions in a highway scenario, very high relative velocities may occur ( 400 km/h). In urban traffic environments, the mobile radio channel may be characterized by a None-Line-Of- Sight (NLOS) situation with large channel tap delays. The air-interface should be able to perform a reliable data detection in both cases. Support of large transmission ranges: Especially in the beginning of market penetration only a small percentage of all vehicles will be equipped with the radio data system. Therefore, it is desirable to provide a transmission range of at least 1000 m for the realization of multihop communication [3]. Support of single burst transmission: The PHY should provide the functionality to transmit and receive data packets in a single burst to guarantee low latencies for high priority data packets (e.g. for emergency notification [3]). Beside these technical requirements, also some nontechnical requirements play an important role for the selection of a suitable air-interface. For instance the availability of an unlicensed frequency band is mandatory whereas the exclusiveness of the frequency band is optional. Exclusiveness avoids system failures (especially for safety-relevant services) due to non-interoperability of coexisting systems within the same frequency band. Additionally, FleetNet has to be based on an existing standard in order to guarantee system interoperability, to shorten time to market and hence to find market acceptance.

2 III. DESCRIPTION OF PHY LAYERS The PHY layer provides transparent transmission of data bits over physical communication resources; i.e. wireless link. For UTRA TDD, the 1.28 Mcps PHY option (Low Chip Rate, LCR) is considered for the performance comparison presented in this paper [4]. In case of IEEE b, different PHY modes with data rates of 1, 2, 5.5 and 11 Mbit/s are defined [5]. Since the transmission range of the system should be as large as possible, only the 1 Mbit/s option is used for comparison. The main PHY parameter of both systems are summarized in Table I. As a basis, a brief description of the relevant parts of the respective PHY layers is given in the following. A. UTRA TDD Ad Hoc PHY Layer 1) UTRA TDD Standard Burst Format: A traffic burst consists of two data symbol fields of 352 chips, a midamble of 144 chips and a guard period of 16 chips, see Figure 1. The Fig. 2. UTRA TDD ad hoc multiplexing example After convolutional coding, 2 zero-bits are inserted to meet the number of 704 bits, provided by a multicode transmission of 8 parallel codes like depicted in Figure 2. Fig. 1. UTRA TDD burst format (1.28 Mcps option) two data fields contain the useful data bits after multiplexing, interleaving, channel coding and spreading, thus increasing the bandwidth of the signal. The number of chips per data symbol is called the Spreading Factor (SF). Within each burst, a training sequence of 144 chips is transmitted in form of a midamble. Basically, it is used to estimate the channel impulse response of the mobile radio channel and is therefore chosen to have good autocorrelation properties. The guard interval at the end of each burst is used to cope with timing inaccuracies, power ramping and also with the propagation delay. 2) UTRA TDD Ad Hoc Modifications: Compared to the UTRA TDD standard PHY specifications, some modifications are required to meet the objectives described in Section II. It has been shown in [6] that the power-impairment problem occurring in direct sequence CDMA systems is of particular importance for the operation of an ad hoc network allowing communication of a node with arbitrary stations within its radio range. To resolve this problem and to reduce inter user interference, the data link layer controls the medium access in a way that within the local area, only a single node transmits in a specific time slot, exclusively. In the UTRA TDD standard, a PHY Service Data Unit (PSDU) is always interleaved and multiplexed over multiple bursts [7]. To meet the requirement of single burst transmission (see Section II), the multiplexing of a PSDU is modified. A multiplexing example of a PSDU for UTRA TDD ad hoc is depicted in Figure 2. Here, a PSDU contains 210 net bits that are protected by 16 bits Cyclic Redundancy Check (CRC- 16). Since a convolutional coder with a constraint length of 9 is used for channel coding, 8 bits are required for Trellis termination [7]. 3) Synchronization of UTRA TDD Ad Hoc: To facilitate the challenging task of synchronization in UTRA TDD ad hoc, it is suggested to beneficially exploit the existence of a globally known time information coming from the Global Positioning System (GPS) for a coarse time synchronization and to apply an additional one-shot fine synchronization to estimate and compensate residual time and frequency offsets of received bursts. The one-shot fine synchronization is carried out in three stages [8]: Fine Time Synchronization (Stage 1): The received signal is correlated with the commonly known midamble. By finding the position where the absolute value of the correlation is maximized, the exact start position of the received burst can be determined. Frequency Synchronization (Stage 2): The received midamble is cross-correlated with different modulated versions of the reference midamble. By finding the maximum absolute correlator output value, the frequency offset of a received burst can be estimated. Additional Frequency Refinement (AFR) (Stage 3): In case of strong multipath propagation, the received midamble will be affected by inter symbol interference and as a result, the frequency estimate of Stage 2 does not necessarily approach the actual frequency offset with increasing Signal-to-Noise-Ratio (SNR). Therefore, the reference midamble is filtered with an estimate of the channel impulse response before it is used for a second cross-correlation with the received midamble. Additionally, decentralized synchronization schemes are being developed to perform coarse time synchronization without being bound to an external time reference from GPS [9].

3 TABLE I PHY PARAMETER OF UTRA TDD AD HOC AND IEEE B Parameter UTRA TDD ad hoc IEEE b (1 Mbit/s option) carrier frequency 2 GHz 2.4 GHz system bandwidth 1.67 MHz 22 MHz chip rate 1.28 Mcps 11 Mcps spreading factor 16 (OVSF code) 11 (Barker code) multicode transmission 1-16 no modulation QPSK DBPSK pulse shaping root raised cosine (r=0.22) frequency mask code rate 1/2 or 1/3 (conv. code or turbo code) 1 (no FEC) max. output power (Europe) 30 dbm (EIRP) 20 dbm (EIRP) B. IEEE b PHY Layer 1) Burst Format: An IEEE b data burst with the mandatory long preamble is depicted in Figure 3. 5 ; + 5., H A = > A " " > E J I ' I 5 1 / ) A A H " , 7 ^ > K H I J ^ - / , 7 ] K I A = J = ^ Fig. 3. IEEE b burst (long preamble) [5] ( > E J I, * 2 5 ( # # H > E J I Each transmission frame is preceded by a Physical Layer Convergence Protocol (PLCP) preamble and a PLCP header, transmitted at 1 Mbit/s with a Differential Binary Phase Shift Keying (DBPSK) modulation and spread by a 11-chip Barker sequence. The PLCP preamble consists of the following fields: SYNC: The SYNC field consist of 128 bits of scrambled 1 bits. This field is provided so the receiver can perform gain setting, energy detection, antenna selection and frequency offset compensation. SFD: The Start Frame Delimiter shall be provided to indicate the start of PHY-dependent parameters within the PLCP preamble. It consist of a 16-bit field, where the rightmost bit (LSB) shall be transmitted first in time. The SFD is used for bit synchronization. All data bits (including PLCP preamble and header) are scrambled prior to DBPSK modulation and spreading. 2) Synchronization of IEEE b: Since the duration of the PSDU is variable, there is no common slotted time structure in the system. Therefore, synchronization of the terminals has to be realized in from of a one-shot procedure since there is no periodical frame structure that could be used in order to build an average over multiple frames to increase timing accuracy. A typical synchronization procedure is carried out in the following steps: Detect receive energy: As the received signal strength indication reaches a certain threshold, the next synchronization step is initiated. Initial code synchronization: In this step, the receiver reference Barker code has to be aligned with received code without any prior knowledge of the position of the received code. The simplest method is to examine all possible positions between the internal reference and received codes until a peak is detected in the autocorrelation function. Code tracking: Once synchronization has been achieved to within 1 chip, the receiver s tracking circuitry is relied upon to reduce and maintain code synchronization error to a value as small as possible. One possible form of code tracking is a delay locked loop [10]. Phase error compensation: Since the local oscillator of transmitter and receiver may differ in their respective output frequency, Carrier Frequency Offsets (CFO) of received bursts can occur. Using a differential modulation scheme like DBPSK, this leads to a constant phase error of demodulated symbols. One possibility is to use a coherent loop for phase error compensation. In this case, the phase error of each demodulated symbol is estimated and averaged. This average value controls the numerical controlled oscillator that produces a frequency to compensate the CFO after code tracking. Bit synchronization: After phase error compensation, the symbols are DBPSK demodulated and fed to the data descrambler. The IEEE b data descrambler is selfsynchronizing and therefore requires no prior knowledge of the transmitter initialization. The receiver should be capable of synchronization on a SYNC field derived from any non-zero scrambler initial state. Upon the reception of the SFD field at the output of the data descrambler, the exact position of the PLCP header is known and the receiver is therefore bit synchronized. IV. PHY PERFORMANCE COMPARISON To compare the performance of UTRA TDD ad hoc with the reference system IEEE b, the loss rate of received data bursts in highly mobile environments is considered as a criterion. In order to achieve a fair comparison, the number of useful bits per PSDU for both systems are harmonized. For IEEE b, the size of the PSDU is always a multiple of 8 bits ( octets ) and determined by the LENGTH field within the PLCP header (Figure 3). It indicates the number of microseconds required to transmit the PSDU and can have

4 A B? a value between 16 to Therefore, an IEEE b PSDU size of 26 octets = 208 bits is used for performance comparison. This value comes very close to the UTRA TDD ad hoc multiplexing example in Figure 2, where a PSDU size of 210 bits is assumed. One way to compare the systems would be to use the Signalto-Noise-Ratio (SNR) as the varied parameter. But since both systems have different values for bandwidth, chip and bit duration, respectively, this would not lead to a fair performance comparison. Therefore, we use the ratio of the energy per useful bit E b to the (single-sided) noise spectral density N 0 as the varied parameter. As a second possibility, we compare the systems with a fixed noise spectral density N 0 and a varied received signal power S. For our simulation, a noise spectral density of N 0 = kt and an environmental temperature of T = 291 K is assumed. A. Simulation Model For the link-level simulations described in this paper, a Wide Sense Stationary Uncorrelated Scattering (WSSUS) model of the time-variant mobile radio channel is used [11]. Two different scenarios are defined to characterize the FleetNet-specific radio channel behavior: A highway and an urban scenario, respectively. The channel parameters of the two scenarios used for the stochastic channel model are summarized in Table II. TABLE II CHANNEL PARAMETERS Parameter Highway Urban tap delay profile exponential max. tap delay 5 µs RMS delay spread 1.28 µs Doppler spectrum Jakes max. relative velocity 400 km/h 120 km/h Rice factor 10 db db The link-level simulator is divided into three main blocks: Transmitter, channel and receiver. A block diagram of the transmission chain is depicted in Figure 4. 6 H = I E J J A H + D = A 4 A? A E L A H 4 A? A E L A. E J A H Fig. 4. * E J / A A H = J H * K H I J / A A H = J H 6 E A 8 = H E = J + D = A D J J 5 6 H = I E J. E J A H F +. 6 I ) 9 / 5 O? D H E = J E, = J =, A J A? J E 2 A H B H =? A ) = O I E I Link-level simulation block diagram In the transmitter, a complex burst sequence is upsampled and then processed by a transmit filter. Within the channel 10 0 IEEE b "Urban" IEEE b "Highway" UTRA TDD(LCR) "Urban" UTRA TDD(LCR) "Highway" E b /N 0 in [db] Fig. 5. Loss rate of received data bursts over E b /N 0 main block, the transmitted sequence is modulated with a random frequency offset, which is assumed to be Gaussian distributed. Afterwards, the signal is affected by the timevariant mobile radio channel and Additive White Gaussian Noise (AWGN). In the receiver block, the symbol sequence is processed by the receive filter and then fed to the synchronization procedure. The synchronized complex symbol sequence can now be used for user data detection and a subsequent performance analysis. The system-dependent parameters used for simulation are shown in Table III. TABLE III SYSTEM-DEPENDENT SIMULATION PARAMETERS Parameter UTRA TDD ad hoc IEEE b code rate 1/3 1 active user codes 8 1 oversampling 4 Root Raised Cosine 5-pole Butterworth Tx / Rx filter with a rolloff lowpass with a cutoff factor of 0.22 frequency of 0.8/T c PSDU size 210 bits 208 bits equalizer ZF-BLE none burst duration 675 µs 400 µs system bandwidth 1.67 MHz 22 MHz B. Simulation Results Based on the assumptions described in the last sections, Figure 5 shows the loss rate of received data bursts depending on the energy per useful data bit divided by the noise spectral density. It can be observed in Figure 5 that an E b /N 0 of 8.5 db is required to achieve the target loss rate of 1% using UTRA TDD ad hoc in the Highway scenario. For comparison, this target loss rate can be achieved by IEEE b with an E b /N 0 of 11 db. In case of the Urban scenario, a target loss rate of 1% can be achieved by UTRA TDD ad hoc with 10 db.

5 IEEE b "Urban" IEEE b "Highway" UTRA TDD(LCR) "Urban" UTRA TDD(LCR) "Highway" Received Power in [dbm] =5 µs =3 µs =1 µs "Highway" AWGN, theoretical E /N in [db] b 0 Fig. 6. Loss rate of received data bursts over received signal power Fig. 7. IEEE b performance with variable max. channel tap delay Using IEEE b in the Urban scenario, the target loss rate of 1% cannot be achieved. This can be traced back to the fact that the Urban scenario is characterized by very large channel tap delays of up to 5 µs and therefore strong Inter Symbol Interference (ISI) of up to 5 adjacent symbols can occur. Since IEEE b does not use any equalization or channel coding, this leads directly to bit errors and consequently to packet losses. In Figure 6, the loss rate of received data bursts over received signal power is depicted assuming a fixed noise spectral density. It can be observed that the performance difference between both systems is even larger in this case. The reason for this large difference is the fact that IEEE b occupies a much larger bandwidth than UTRA TDD ad hoc and the useful signal is therefore interfered by a more powerful noise signal assuming a constant value for the noise spectral density. To analyze the influence of the maximum channel tap delay on the packet error rate of IEEE b, different values of τ max have been simulated. Additionally, a theoretical performance curve of IEEE b in AWGN conditions is depicted. For this curve, the packet error probability of a DBPSK modulated and spreaded data packet consisting of 400 bits is calculated [10]. Figure 7 shows that the packet error rate in the Highway scenario is very close to the analytically calculated performance in AWGN. In the Urban scenario, the packet error rate significantly increases with maximum channel tap delays that are larger than the bit duration of 1 µs. V. CONCLUSIONS For the air-interface of FleetNet, a modified version of the UMTS Terrestrial Radio Access Time Division Duplex (UTRA TDD) has been proposed. In this paper, the PHY layer performance of UTRA TDD ad hoc is assessed and compared to the well-known IEEE b standard. In order to achieve a fair comparison, the packet sizes of both systems are harmonized. Packet error rates are investigated in vehicular environments with very large relative velocities and strong multipath propagation, respectively. The proposed UTRA TDD ad hoc system outperforms IEEE b in both scenarios. In an urban scenario with strong multipath propagation, the IEEE b standard is not able to meet the required performance in terms of packet losses. REFERENCES [1] A. Ebner and H. Rohling, A Self-Organized Radio Network for Automotive Applications, Proceedings of the 8 th World Congress on Intelligent Transport Systems, Sydney, Australia, October [2] L. Wischhof, A. Ebner, H. Rohling, M. Lott, and R. Halfmann, SOTIS - A Self-Organizing Traffic Information System, Proceedings of the 57th IEEE Vehicular Technology Conference (VTC 03 Spring), Jeju, Korea, [3] H. Hartenstein, B. Bochow, M. Lott, A. Ebner, M. Radimirsch, and D. Vollmer, Position-Aware Ad Hoc Wireless Networks for Inter- Vehicle Communications: the Fleetnet Project, Proceedings of the ACM Symposium on Mobile Ad Hoc Networking & Computing, MobiHoc 2001, Long Beach, California, USA, October [4] 3GPP, Physical channels and mapping of transport channels onto physical channels (TDD), Technical Specification Group Radio Access Network, 3G TS V5.0.0, March [5] ANSI, Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications: Higher-Speed Physical Layer Extension in the 2.4 GHz Band, ANSI/IEEE Standard b-1999, [6] M. Lott, R. Halfmann, E. Schulz, and M. Radimirsch, Medium Access and Radio Resource Management for Ad hoc Networks based on UTRA TDD, Proceedings of the ACM Symposium on Mobile Ad Hoc Networking & Computing, MobiHoc 2001, Long Beach, California, USA, October [7] 3GPP, Multiplexing and channel coding (TDD), Technical Specification Group Radio Access Network, 3G TS V5.0.0, March [8] A. Ebner, H. Rohling, R. Halfmann, and M. Lott, Synchronization in Ad Hoc Networks based on UTRA TDD, Proceedings of the 13th IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC 2002), Lisbon, Portugal, [9] A. Ebner, H. Rohling, M. Lott, and R. Halfmann, Decentralized Slot Synchronization in Highly Dynamic Ad Hoc Networks, Proceedings of the 5th International Symposium on Wireless Personal Multimedia Communications (WPMC 02), Honolulu, Hawaii, [10] J. G. Proakis, Digital Communications, 3rd ed. McGraw Hill, [11] P. Höher, A statistical discrete-time model for the WSSUS multipath channel, IEEE Transactions on Vehicular Technology, vol. 41, p. 461, Nov