Adaptive schemes for packet data in a ds-cdma environment

ADAPTIVE SCHEMES FOR PACKET DATA IN A DS-CDMA ENVIRONMENT Oriol Sallent, Ramon Agusti Departament de Teoria del Senyal i Comunicacicons Universitat Polit6cnica de Catalunya C/ Gran Capitl s/n - 08034 Barcelona - Spain1 Tel. +34-3-401.71.97 Fax +34-3-401.72.00 E-mail : oriol @ xaloc.upc.es Abstract This paper presents a novel Adaptive DS-CDMA SlottedALOHA packet random access scheme for improving the throughput of the conventional DS-CDMA Slotted-ALOHA system. For this purpose a Mobile-assisted algorithm is envisaged to control the change of the transmission rate according to the traffic load. This algorithm revealed that the optimum behavior may be almost reached at a low complexity cost. Moreover, the proposed algorithm was found to be robust to intercell interference. I. INTRODUCTION Slotted ALOHA (S-ALOHA) has been widely recognized for packet radio applications when the data transmitted from the terminals are bursty or even recently for voice cellular communications [ 11. S-ALOHA DS-CDMA scheme allows several users to transmit at the same time using a different spreading code for each one [2], [3], so that unsuccessful transmissions are caused entirely by multiple access interference, provided negligible thermal noise effects are assumed. This paper is intended to improve the throughput achieved with a multi-receiver DS-CDMA S-ALOHA when a light load is offered to the system. For this purpose we will focus on the transmission rate used instead of the access protocol itself. In particular, different transmission rates are proposed to be used in order to utilize the most suitable rate according to the channel load at any time slot. Then, a fast and simple algorithm that command the processing gain of the Spread Spectrum DS scheme as a function of the channel load is envisaged. For low channel loads, a high CDMA processing gain is no longer necessary and it can be reduced so as to increase the actual bit rate. Therefore, in an extreme situation with a processing gain equal to one, all the users would access the same channel at their maximum bit rate according to a pure S-ALOHA. In a real cellular scenario, interference from neighbough cells must be taken into account. To keep these interference at a reasonable level, the concept of cell cluster arises in techniques with a FDMA component. However, it is the case neither for DS-CDMA based systems nor for cellular SALOHA [I], where a reuse factor of 1 is normally considered. Accepting that a throughput decrease will follow when considering ihe cellular structure, we are interested on assessing this decrease in terms of throughput as well as whether the proposed change of the transmission rate algorithm is still valid or not. We will show that no changes are required in the algorithm, despite the interference is not only dependent on the offered load but also on the propagation conditions (that strongly varies the interference level affecting to a reference user). Furthermore, some power control based policies are proposed to improve the system performance. The paper is organized as follows. In Section 2 an analytical model for the DS-CDIMA S-ALOHA system is presented, which is used to evaluate the optimum achievable throughput with an adaptive change of transmission rate algorithm. In Section 3 a particular S-ALOHA adaptive algorithm is proposed and assessed. In Section 4 a cellular model is considered and the impact of other-cell interference is attained. The change of the transmission rate algorithm is also studied under tlhese conditions and limited-power strategies are evaluated. Conclusions close the paper in Section 5. 11. MODEL FOR A S-ALOHA DS-CDMA ACCESS SYSTEM From now onwards N registered users will be considered. These users can be in two different operation modes: ‘idle mode’ and ‘backlogged mode’. In the former mode there is no packet to be retransmitted and new packets are generated with probability po. Terminals enter the backlogged mode when an attempt to transmit a new packet fails. In this mode, the retransmission of the backlogged packet occurs in any given slot with probability pr . While in the backlogged mode the user does not generate any new packet. The process defined by the number of backlogged users at the beginning of the k th slot is a Markov chain, whose performance has been studied among others in [4]. When considering DS-CDMA, the formulation must be modified as it was done in [5]. Analytical throughput measurements can be obtained from the markovian model. 0-7803-3659-3/97 $1 0.00 01 997 IEEE 1019 A. S-ALOHA DS-CDMA ACCESS SYSTEM A BPSK DS-CDMA access system with a processing gain given by Gp is considered. All users have been assigned random PN signature sequences. A perfect power control capable of mitigating fadings of the channel is introduced. With a view to achieving this, a continuous link between BS and MS is needed so that the mobile transmits at rate v when there is information to be sent and at rate v’ (in general much lower) when the terminal is not active. v’ should be high enough to allow the fading on the uplink path to be tracked by updating transmitted power in response to the BS commands. An instantaneous power control permits a huge improvement in the system performance when compared with an open loop power control, and it is, in fact, considered in the already operative [6] and the proposed CDMA systems [7] respectively. Although an instantaneous power control could be envisaged for packet radio on a packet by packet basis [8], we have retained the continuous link approach because in this case no synchronization overhead at the beginning of time slot would be required, since the mobile is already synchronized. Whether continuous power control is retained or not it is in any case irrelevant for the proposed S-ALOHA scheme taken for granted that a throughput decrease should be considered in the non-continuous packet based power control approach. B. S-ALOHA DS-CDMA PERFORMANCE By assuming an ideal instantaneous power control, the channel can be seen as an AWGN if we use the gaussian hypothesis to model the interference originated by other users [9], [2]. Under these conditions the following expressions hold for the evaluation of the BER where IZ is the number of simultaneous users. The probability of detecting correctly a packet containing aB bits is Keeping the transmission bandwidth constant, and taking as a reference a transmission rate of v bitds, packet length of B bits and a processing gain Gp, the transmission rate could be increased by a factor a ( a>l) at the expense of a reduction in the spread capacity. That is, transmitting at rate av bids would allow a processing gain of G,Ja . However, in exchange for this reduction in processing gain, aB bits per time slot can be allocated instead of only B. attempted transmissions per time slot -, and po is set equal to pr so that po=p,=p=G/N . G is assumed to vary slowly. Moreover, for numerical results v‘ has beem set to zero. Although there are no restrictions on the values of a, for the sake of brevity and clarity only three rates v, 2v and 4v bits/. (a=l,2,4) will be considered in the sequel. It is clear from Fig. 1 that the higher rate (4v bitds) is interesting when the system is lightly loaded: more bits per packet can be sent through the channel, since the little interference observed permits correct transmission. When the offered load increases so does the interference level, and therefore the rate 4v bits/s is no longer interesting because of the higher BER compared to the 2v bitsh rate. Using 2v bitds in this range is better than 4v bits/. in the sense that a higher throughput is achieved although the packet contains half the number of bits per packet (errors occur quite often with 4v bits/s, but 2v bits/s can still bear the interference because of the higher Gp compared to that of 4v bits/. rate). For higher offered loads the same trade-off appears between 2v bits/. and 1 v bitsh. 7000 1 I 6000 5 5000 r 5 4000 m ln r I 2 3000 Q g 2000 t 1000 Fig. 1. Throughput performance for different transmission rates. C. ADAPTIVE S-ALOHA DS-CDMA In the light of the above results, an algorithm able to change the transmission rate used by the MS as a function of the traffic load of the system could be foreseen so that the maximum possible throughput could always be obtained. The optimum throughput attained can be analytically obtained from the Markov model, and this result can be later used as a reference to assess the performance of the transmission rate change algorithm proposed below. In order to attain this optimum performance, the best combination of transmission rates should be obtained provided that n simultaneous users are present. The resulting optimal table would be reached after an exhaustive search, as it is further explained in [lo]. Fig. 2 shows this optimum throughput in comparison with the individual behaviors. The achievable global throughput with different transmission rates obtained with the Markov model can be seen in Fig. 1, where G stands for the offered load - average number of 1020 7000 l 7000 6000 - 5 5000 r L 4000 m cn c 3000 13 @ 2000 c 1000 Fig. 2. S-ALOHA DS-CDMA optimum throughput. 111. CHANGE OF THE TRANSMISSION RATE ALGORITHM In this section a proposal of an adaptive change of the transmission rate algorithm is addressed. The basic idea consist in sensing the traffic load through the channel in order to accommodate transmission rates accordingly. The proposed simple algorithm carried out by the mobile station (MS) works as follows: each terminal traces its own evolution during the transmission time, that is, terminals count their successful and erroneous packets. In the absence of errors the mobile will assume a low traffic load and tries to use a higher transmission rate. The throughput should be increased in this way. If errors occur, the mobile decides that the channel is too loaded and tries a lower transmission rate. In this case fewer bits per packet are transmitted, but a global improvement of the throughput should also follow because these bits can be now detected correctly since processing gain increases accordingly. Let us note that this decision is taken by the mobile without any exchange of information with the Base Station (BS) except for the packets acknowledgment. Even the MS does not need to indicate its choice of transmission rate before using it because the very same BS could be able to detect which one is arriving. Specifically, the MS only needs to establish two parameters: the number of consecutive packet failures before changing to a lower rate (max-tr) and the number of consecutive packet successes before trying a higher rate (min-suc). Fig. 3 shows the throughput attained for N=60 registered mobiles. Fig. 1 has been taken as a reference in order to appreciate whether mobiles choose the most suitable rate or not. In spite of the simplicity of the algorithm, the envelope of the three individual graphs is almost reached, and this is not far from the optimum behavior. Results have been obtained with m-t-1 and min_suc='l, which is the best choice, once other possibilities have been studied. 6000 - 5 5000 r 4000 m c 1 lz 3 2 3000 2000 c I000 0 0 10 20 30 40 50 60 G Fig. 3. Throughput performance for MS algorithm. IV. MS ALGORITHM IN A CELLULAR ENVIRONMENT When considering multiple-cell CDMA systems, which will normally be the case, the situation becomes more complicated. There are basically two new problems that must be taken into account: 1-. The interference level originating from users in the other cells. This interference varies not only according to the attenuation in the path to the subscriber's cell site, but also inversely to the attenuation from the interfering user to his own cell site, which through power control by that cell site may increase or decrease the interference to the desired cell site. 2-. The necessary handover procedures. In order to keep the general interference level as low as possible, it would be desirable for each mobile to be attached to the cell site for which the radio path attenuation is minimized. Handover decisions are usually based on pilot-strength measurements (radio distance measurements). However, some time is necessary for power measurement averaging and for handover execution. In order to evaluate the MS algorithm in a cellular environment several computer simulations have to be carried out, and the first requirement is to establish a channel model. The propagation attenuation is generally modeled as the product of a shadowing fading and a short-term fading. The former is due to the teirrain configuration, whereas the latter is largely due to inultipath reflections. An accepted representation for the received power is (3) where P, is the received power, Pt is the transmitted power, a2 is an exponential random variable, is the dB attenuation due to shadowing, with zero mean and standard deviation (3 1021 MAX-POW means reducing the interference caused to both his own and the other cell sites. However, it makes it more difficult for the power control to maintain the signal reference level at the base station for the single user. The reverse applies if MAX-POW is increased. Which of these phenomena has more impact on the global throughput depends on the level to which the power is limited as well as on the offered load, which determines the total interference level. Results for different values of MAX-POW are presented in Fig. 6. Only the performance of the MS algorithm is shown, and the gain obtained when limiting the maximum transmitted power can be noted. The throughput degradation for low loads when decreasing MAX-POW could be avoided by adaptively changing this parameter jointly with the transmission rate, according to the MS algorithm. Therefore, the interference effects of Rayleigh fading could in practice be strongly mitigated. 4500 I 4000 1 ......... ; ........ /.:?- ................... , ........... 1 .L. -._ ~ ,’, 3500 ........... ;.. ..... i. .: ........... ..:... ...... ..; .;Y:, .! ........... i/’: ,, . <. 1 2 3000 2500 bI c 3 8 2000 0 OL E 1500 1000 500 ....... ............... ..... ..... ........ .; .............. , .......... ...... .... .......... .; ........... .I . .... I .... 1.1.. I .... I .... I 0 10 20 30 40 50 60 G Fig. 6. A power limitation condition is imposed. V. CONCLUSIONS A new Adaptive DS-CDMA S-ALOHA technique for packet mobile communications access based on the choice of the most suitable transmission rate in any time slot has been addressed. The mobile assisted algorithm allows us to obtain almost the maximum attainable performance at a very low complexity cost, and by far outperforms those obtained with a conventional DS-CDMA S-ALOHA scheme. Simulations in a cellular environment show that the proposed algorithm is also robust to intercell interference. Moreover, limiting the maximum transmitted power proved to be a good policy for general throughput improvement. VI. ACKNOWLEDGMENTS The work described in this paper was carried out within the CICYT TIC94-0870-C02-01 project, in the framework of National Plan of Spain. VII. REFERENCES [l] M. ZORZI, S. PUPOLIN, “Slotted ALOHA for HighCapacity Cellular Communications”, IEEE Transactions on Vehicular Technology, Vol. 43, No. 4, November 1994, pp. 1011-1021. [21 D. RAYCHAUDHURI, “Performance Analysis of Random Access Packet-Switched Code Division Multiple Access Systems”, IEEE Transactions on Communications, Vol. Corn. 29, No. 6, June 1981, pp. 895-901. [3] R. K. MORROW, J.S. LEHNERT “Packet Throughput in Slotted ALOHA DSISSMA Radio Systems with Random Signature Sequences”, IEEE Transactions on Communications, Vol. Corn. 40, No. 4, July 1992, pp. 1223-30. [4] L. KLEINROCK, S.S. LAM, “Packet Switching in a Multi-access Broadcast Channel: Performance Evaluation ”, IEEE Transactions on Communications, Vol. Con-23, No. 4, April 1975, pp.410-422. [5] Z. LIU, M. EL ZA.RKI, “Performance Analysis of DSCDMA with Slotted ALOHA Random Access for Packet PCNs ”, Wireless Networks 1 (1995), pp. 1-16. [6] A. SALMASI, K. S. GILHOUSEN, “On the System Design Aspects of CDlMA Applied to Digital Cellular and Personal Communicatiions Networks”, Proc. 4lSt IEEE Vehicular Technology Conference, St. Louis, May 1991. [7] A. BAIER et al., “Design Study for a CDMA-Based Third-Generation Mobile Radio System ”, IEEE Journal on Selected Areas in Connmunications, Vol. 12, No. 4, May [8] M. J. McTIFFIN ‘et al., “Mobile Access to an ATh4 Network Using a CDMA Air Interface ”, IEEE Journal on Selected Areas in Conimunications, Vol. 12, No. 5, June [9] M. B. PURSLEY, “Performance Evaluation for PhaseCoded Spread-Spectrum Multiple-Access Communication - Part I: System Analysis ”, IEEE Transactions on Communications, Vol. Com.-25, No. 8, August 1977, pp.795-799. [lo] 0. SALLENT, F:. AGUSTI, “A Mobile Controlled Algorithm for Improving the Throughput in a S-ALOHA DSCDMA System”, Proceedings of the 7th Personal, Indoor and Mobile Radio Communilcations (PIMRCP6), pp. 1192-96. [ll] M. KWOK, H. WANG, “Adjacent Cell Interference Analysis of Reverse-Lin k in CDMA Cellular Radio Systems” Vehicular Technology Conference VTC’95, pp.446-450. [12] K. S. GILHOUSIZN et al. , “On the Capacity of a Cellular CDMA Systern”, IEEE Transactions on Vehicular Technology, Vol. 40, No. 2, May 1991, pp.303-312. [13] M. GUDMUNDSON, “Correlation Model for Shadow Fading in Mobile Radio Systems”, IEE Electronic Letters, Vol. 27, No. 23, 7th November 1991, pp. 2145-2146. 1994, pp. 733-743 . 1994, pp. 900-908 . 1023