当前位置:首页 >> >>

Adaptive radio resource sharing for collocated Wireless Personal Area Networks


Adaptive Radio Resource Sharing for Collocated Wireless Personal Area Networks
Hiroyuki Yomo
1

, Petar Popovski, and Ramjee Prasad

Department of Communication Technology, Aalborg University Niels Jernes Vej 12, 9220 Aalborg, Denmark TEL/FAX: +45-96359897/+45-98151583 e-mail: yomo, petarp, prasad @kom.auc.dk

Abstract— A Wireless Personal Area Network (WPAN) provides wireless networking among proximate devices, usually carried by an individual. As the WPAN gains momentum in ubiquitous usage, the interference that collocated WPANs cause to each other, termed self-interference, will be one of the major sources that degrade the communication performance of WPAN. In this paper we propose Dynamic Adaptive Frequency Hopping (DAFH) mechanisms that are concurrently employed by collocated WPANs in order to avoid the self-interference. With DAFH, WPAN adaptively self-allocates a subset of frequency channels to be hopped, such as to minimize the experienced interference. The packet error rate is the only input to the proposed mechanisms, which enables DAFH to also avoid interference from static, nonFH interferer. However, achieving higher throughput is not the sole optimization target and DAFH is designed such as to cause minimal increase of the interference to the collocated non-WPAN entities that utilize the same unlicensed spectrum. The operation of DAFH is robust and adaptive to the dynamic changes in the environment and to the noise errors in the channel. Simulation results show that DAFH signi?cantly increases the throughput of the WPANs, while the WPANs employ best effort to minimize changes in the overall interference pattern.

I. I NTRODUCTION An important recent development in the wireless technology is Wireless Personal Area Network (WPAN), which provides a network among user’s wearable devices, as well as wireless links within short range (e.g. 10 meters) to access points, sensors, actuators etc. WPAN is an ad hoc network, being established and maintained without preexisting networking infrastructure. Bluetooth [1] is a ?rst approximation of the WPAN concept and it has been a basis for standardization within the IEEE 802.15 Working Group (WG) [2]. The ubiquitous usage scenarios are setting WPAN to operate in unlicensed spectrum, such as Bluetooth in the 2.4 GHz ISM band. Due to the open access, a WPAN may experience/cause high interference from/to the transmissions/receptions of proximate entities that utilize the same spectrum. Bluetooth uses slow pseudorandom frequency hopping (FH) over 79 channels, but this does not solve the interference problem, since a serious throughput degradation occurs when e.g. a WPAN is settled close to IEEE 802.11 Wireless Local Area Network (WLAN) [3]. In the future, many WPANs will occur to be in close proximity at the spots with high user density, such as airports, conference rooms, etc. Such collocated WPANs cause severe interference to each other [4] [5] since they are not coordinated in the usage of the radio resources. Since different WPANs are identical from a system viewpoint, the mutual interference among collocated WPANs is termed self-interference. To alleviate the interference with proximate entities, a WPAN should apply certain coexistence mechanism. Since the
1 Hiroyuki Yomo is working for Future Adaptive Communication Environment (FACE) project.

orthogonal pre-allocation of the radio resources in unlicensed spectrum is impossible, the coexistence mechanisms should adapt to the actual interference pattern in an ad hoc manner. There has been a considerable work regarding the coexistence between WPAN and WLAN and two classes of mechanisms have been de?ned [3] [6]: collaborative and non-collaborative. In collaborative mechanisms the interfering entities exchange data to achieve mutual coordination, while non-collaborative mechanisms consider that data exchange is not available. In this paper we are focusing on the non-collaborative mechanisms. IEEE 802.15.2 proposes two generic types of noncollaborative techniques: adaptive packet selection/scheduling (APS) and adaptive frequency hopping (AFH). Both noncollaborative techniques perform classi?cation of channels as “good” or “bad” (low or high interference). In APS the packet type and the transmission timing are adapted to the channel conditions [6]. AFH dynamically changes the hopping set, shortly hopset, such that the WPAN preferably hops over the set of good channels [7]. The AFH proposed in IEEE 802.15.2 is restricted to coexistence of WPAN with frequency static devices in the 2.4 GHz ISM band, such as IEEE 802.11b [7]. To tackle the self-interference from collocated WPANs, the non-collaborative coexistence mechanisms require different approach than the one used against non-WPAN interferer. The essential observation is that the WPANs will run the identical algorithm in a concurrent manner. Listen-BeforeTalk (LBT) [8] has been proposed as a non-collaborative mechanism for avoiding the dynamic interference among collocated WPANs. LBT introduces a carrier sense, such that the packet transmission in WPAN is suppressed if the signal at the chosen frequency is above a threshold, which in case of many collocated WPANs yields large throughput decrease. The AFH proposal by IEEE 802.15.2 [7] can not be used to mitigate the self-interference, since its design does not account that the interfering entity may apply the same algorithm. There are two reasons which render the heuristics for the frequency static interferer inapplicable for AFH to combat the self-interference. First, the collocated WPANs will induce/receive approximately the same error rate at each channel. As such, the simple rule “remove bad channels from the hopset” will drive the WPANs towards removal of the same channels and the WPANs will self-interfere over the remaining channels. Second, when a WPAN hops at a channel, the error probability is proportional to the probability that at least one collocated WPANs visits that channel. This is much less than a probability of packet error at a frequency with static interferer. Then, the time to gather reliable statistics at the individual channels is highly increased. This suggests that, when AFH is applied concurrently by the interfering entities, the cumulative error statistics for channel groups should be used. In this paper we propose dynamic adaptive frequency hop-

ping (DAFH) for avoidance of a frequency-dynamic selfinterference among collocated WPANs. To the best of our knowledge, there has not been a proposal so far for AFH that is concurrently applied by the mutually interfering entities. The algorithm uses the realistic assumption that a WPAN can not differentiate whether the interferer is (are) another WPAN or a non-WPAN entity. Thus, being able to avoid the dynamic interference, DAFH is inherently able to combat a frequency static interference. The price paid for the throughput improvement in individual WPANs is the possible increase of interference caused to other collocated devices that operate in the unlicensed spectrum. Therefore, the design of our DAFH tries to exhibit fair behavior towards the other collocated nonWPAN entities, while signi?cantly increasing the throughput of the individual WPANs by avoiding self-interference. II. S YSTEM M ODEL The focus in this work is put on the self-interference among the collocated WPANs. Therefore, we will abstract the internal structure of a WPAN. The abstraction is derived from the actual networking structure in Bluetooth (IEEE 802.15.1) and we brie?y outline some facts pertaining to this technology. A basic networking entity in Bluetooth is piconet—a star topology with a master and up to seven active slaves. The communication channel in a piconet is slotted with a nominal slot value 625 [ s]. Piconet applys per-packet frequency hopping. The hop selection in each slot is based on a pseudorandom generator determined by the master. The slaves are timeand hop-synchronized to the master. The master uses polling to schedule the packet transmissions, such that, within the piconet, the communication is collision-free and at each slot only one device transmits. A WPAN corresponds to a piconet and we will use the terms WPAN and piconet interchangeably. The ?th piconet . The distance between is denoted by , where devices within a piconet is small, and if at least one device in ? is in transmission range of a device in ? , then each device in ? is in transmission range of any device in ? , and vice versa. This allows us to consistently de?ne that two piconets ? and ? are collocated if the devices in the piconets are in each other’s transmission range. Therefore, a receiving device in piconet ? perceives the collocated piconet ? as a single transmitting entity, such that we can represent a piconet as a transmitter with slotted channel, where each packet has a duration of single slot. The piconets are asynchronous in a sense that their slot-starts are not coinciding, making a slot in a piconet overlap with two slots in another piconet. We model as a transmitter for which the probability that the piconet packet occurs in a slot is [4] [9] [10]. In this paper, we only )[5] [11]. consider fully loaded case ( Two collocated piconets interfere with each other if they simultaneously transmit packet at the same frequency. In this paper, a collision is always considered destructive, resulting in a packet error with probability one [4] [12]. IEEE 802.15.2 considers monitoring of the Packet Error Rate (PER) as a primary method to assess the channel quality, due to its simplicity. Our algorithm also uses the PER measurements as an input. We do not assume any special mechanism for collision detection—the collided packet is detected to be erroneous by an error detection code. Therefore, the receiver

Level 0 Level 1 Level 2

??? ??? ??? ??? ??? ??? ???

?? ?? ?? ?? ?? ?? ??

11111111 11110000 00001111 11000000 00110000 00001100 00000011

Fig. 1.

Admissible utilization vectors for

?

and

? ?

cannot distinguish between the error due to collision with other piconet and error due to other channel impairments. III. DYNAMIC A DAPTIVE F REQUENCY H OPPING (DAFH) The goal of DAFH is to allocate non-intersecting hopsets to the proximate WPANs such as to avoid the self-interference among the WPANs. This allocation is distributed, since each WPAN self-allocates a hopset based on the monitoring of the PER. However, the throughput optimization due to the avoidance of the mutual interference should not be the sole target of the DAFH algorithm. The smaller the hopset of a WPAN, the higher the interference that WPAN causes at the frequencies from the hopset, which can have devastating effect towards proximate non-WPAN devices, operating in the unlicensed band. Therefore, each WPAN should attempt to select a hopset that is non-intersecting with the hopsets of the proximate WPANs and in the same time is as large as possible, in order to achieve good etiquette i.e. fair behavior towards the collocated devices that operate in the same unlicensed band. In the sequel, we describe the four building blocks of the proposed DAFH algorithm that aims to achieve this behavior: (1) De?nition of the superset of possible hopsets; (2) Reduction of the hopset based on PER; (3) Hopset doubling; and (4) Choice of the threshold for the PER measurements. 1) Superset of possible hopsets: The hopset used by the is represented by a utilization vector ? piconet ?? ?? ? ? , where ? ? ? for ? ? , then the frequency ? is in ? . If ? ? is ? ? the hopset. The size of the hopset of ? ? ? . The frequencies belonging to the hopset are ? ? ? chosen uniformly with probability ? . The reason to de?ne a superset of possible hopsets is to speed up the orthogonalization of the hopsets and to have controlled interference patterns among the piconets. The superset is represented by the . The admissible set of admissible utilization vectors ? ? ? ? represents the -th hopset utilization vector ? ? ? ? and ? . If at level ?, where ? uses ? ? , then we write: piconet

??

?

??

?

??

?

If

?

??

? ? ?? ?? ? ? ??? ? then ? ? ? ·? ? otherwise
? ?

? ?

??

? ?

?

? · ?? ?

? ?

(1)

, The full hopset with ? frequencies corresponds to ? and an admissible utilization vector at level ? represents a ? hopset of size ? ?? . The maximal level ? is determined as

?? ??

?

??

?

? ?

? the WPAN technology and we consider that ? ?? ? is an integer. Fig. 1 gives an example of admissible utilization

??

, where

? and ?? ? are speci?ed for

??

vectors for ? and ? . If the hopset of is a subset ? . From Fig. 1 and (1), of hopset of , then we write ? it can be seen that if two admissible utilization vectors ? and ? are interfering, then either ? ? or ? ? . 2) Reduction of the hopset based on PER: If a piconet experiences PER that is higher than a threshold , then the piconet is triggered and it chooses the new hopset randomly, between the left and right half of the current hopset. To give a complete description of the randomized reduction, let have a hopset with ? ? ? , where ? ?. If is triggered, then it ?ips a fair coin to choose the new hopset or the right subset to be either the left subset ? ? ?? . However, this reduction by a binary splitting ? ? , since ? ?? ? and is inapplicable when ? the hopset can not be reduced further. Therefore, if triggered, ? ? , will select ? ? ? , where ? a piconet with ? ? is chosen uniformly from the set ? . runs the protocol relying on the obEach piconet served PER. The piconets are not explicitly coordinated in any manner and the only way by which they “communicate” with each other is by causing interference, which is the essence of the non-collaborative coexistence mechanisms. ? , then the If the number of collocated piconets ? randomized reduction by binary splitting relatively quickly produces orthogonalization of the radio resources among the counts its total number of collocated piconets. Piconet packets as ? ?? ? ???? and the number of erroneous packets ???? ????. When ???? ???? reaches a prede?ned value ????? , calculates: of ?

?

? ?

? ·? ? ·??

? ·? ? ? ?

?

?

?

??

? ?

? ?

?

? ?? ?

?

. If ? ? and resets ? ?? ? ???? , the piconet is triggered. By limiting the ???? ???? instead of ? ?? ? ????, the PER is always estimated with the same reliability, while the piconets tend to faster leave the states that are characterized with high PER. In a Bluetooth piconet, the DAFH algorithm should be run by the master in a centralized manner and the master should also account for the packet errors at the slaves. 3) Hopset doubling: The desired behavior of DAFH is that a WPAN should work with reduced hopset only if there is interference at the other frequencies, either from static interferer or from other collocated piconets. Hence, a piconet should have a mechanism for periodical probing in order to regain the frequencies that were excluded during the reduction process. The same argument is used in the adaptive frequency hopping in [7], where channels that were claimed “bad” are rechecked periodically to assess whether the interferer had disappeared. To cope with such situation, we introduce the randomized set doubling, a mechanism which is complementary to the have ? ? ?? ? , with ?? randomized reduction. Let . If has not been triggered for ? slots, then doubles its ? ?? ? , where ?? ?? ? and ? hopset by selecting ? ?? is chosen uniformly from ? . ? is referred to as doubling period. After doubling, resets the ? ?? ? ???? and ???? ???? and starts to monitor PER. remembers the set ? ?? ? after doubling from level ?? to ?? . If is triggered when ???? ???? reaches ? ????? for the ?rst time, then sets again ? ? ?? ? . If

????? ? ?? ? ???? ???? ????

(2)

?

? ?

not triggered at that time, resets the counters, “forgets” ? ?? ? and proceeds with the usual randomized reduction, is not triggered at level ?? described above. Further on, if for ? slots and if ?? , will double the set to level ?? ?? ? . should again forget the previously used subset at level ?? and remember only the selection at level ?? . This “forgetting” is introduced in order to avoid the deadlocks, which occur by repetitive unfavorable selection of the same hopset. Regarding the selection of doubling period, if ? is too short, then the effect of the orthogonalization by reduction will be hindered and a large overhead will be introduced. On the other hand, if ? is too long, the interferers that forced the piconet to reduce the hopping may not be collocated anymore, such that the piconet induces unnecessarily large occupancy at the frequencies of its hopset. 4) Choice of the adaptive threshold: The choice of the PER threshold used for triggering the piconet is in?uenced by several factors. First, the threshold should be higher than the maximal expected PER due to the channel noise, in order to avoid unnecessary triggering. This is because the hopset change cannot improve the PER when the source of error is the noise that is uniform over the frequency channels. Hence, ?? , where ?? the minimal value of the threshold for is is the estimated maximal probability of packet error produced by channel noise. The second condition for choosing the threshold is implied by the dynamics scenarios of self-interference, where the set of collocated piconets is time-variant. Consider a case when and become collocated, while for the two piconets ? i.e. ? ? and ? utilization vectors we have ? ? . According to the discussions of the desired behavior for DAFH, we aim to choose the threshold such that only the piconet with larger hopset is triggered. Therefore, we make is using hopset at level the following threshold choice: If ?, then ? and we set the condition:

?

?

?

?

??

If ??

?? then

??? ? ?

??? ?

(3)

such that, a piconet with smaller hopset needs higher PER to be triggered. In order to determine the concrete value ? , observe that there are ? hopsets at the level ?. for has ? Then, it is reasonable to choose the criterion: If of level ?, then it should not be triggered by less than ? fully loaded piconets with full hopsets. This criterion results in the following threshold at level ?:

??

?

? ??? ? ? ?? ? ? ? ? ? ?
?

???? ??

?

(4)

?

?

?

? ? ?? ?

?

?

?

?

?

?

which can be derived by using the approach from [4] for estimating the interference from a FH interferer and the assumption that the maximal estimated PER due to channel noise ?? is constant over all frequencies. Once the adaptive threshold according to (4) is adopted, the following observation leads to additional acceleration of have ? the resource division among the piconets. Let ? ? . Let have separate error counters for the left and right half of its current hopset. Denote by ? the number of errors that occurred at frequencies in the hopset ? ? and by ? the errors in ? ? . If there is a piconet with ? ? and ? ?? , then gets ? ? . This should force to choose the right subset

? ?

? · ? ? · ?? ? ·?? ?

? ·? ? ?

in the reduction, such that and become non-interfering is triggered. However, the rule that a piconet should after ? and vice versa does always choose the left subset if ? not work if the interfering piconets have identical hopsets. In ? , but the deterministic such case we can also have ? selection drives the piconets towards the selection of the same hopset. Clearly, the randomization is needed to break the symmetry. Therefore, we use the following combined strategy: The probability to select the left half of the current hopset upon triggering is: ?

1 0.95 0.9 0.85 Average Throughput 0.8 0.75 0.7 0.65 0.6 0.55 0.5 2 3 4 5 6 7 Average Number of Piconets (N) 8 9 10 DAFH-AT DAFH-CT W/O AFH

??

??

? ?

(5)

IV. N UMERICAL R ESULTS A. Simulation Model We consider a quasi-static scenario where WPAN users arrive at a hotspot, and leave the hotspot after a random dwell time. We assume a Poisson process with generation rate for the piconet arrivals and exponential distribution with the average of ? [sec] for the dwell time at the hotspot. In this case, the average number of piconets at the hotspot, denoted ? , can be obtained from Little’s formula as ? ? ? [13]. In the simulation, we ?x ? at 60 [sec], and change the average number of piconets by giving different value of . We set the minimal dwell time to be 20 [sec] (see [14]), which makes the exponential distribution shifted for 20 [sec]. We set the in order to keep system total number of frequencies ? parameters close to those of the Bluetooth technology. The . This minimal allowable size of a hopping set is ?? ? choice is made so that we can accommodate as many noninterfering piconets as possible, following the condition that ? ? ? ?? ? is an integer. The number of packet errors used for the calculation of PER is ? ????? , and we set the maximal estimated PER due to noise errors to [5]. be ?? We mainly investigate the following two performance measures of WPAN systems with DAFH and without AFH: a) Average Throughput: The throughput is de?ned as the fraction of time that is used by successful packet transmissions. For fair comparison, we account for the overhead packets that are used to convey the information about the change of hopsets among piconet members in DAFH. Considering Bluetooth technology with at most 7 slaves in a piconet, we assume that 14 slots (a packet plus acknowledgement per each slave) are required to convey the hopset information reliably. This assumption re?ects the worst case for the overhead with respect to the number of piconet members. The throughput of DAFH is calculated by subtracting the overhead packets from total number of successful packets. Average Throughput is the throughput averaged among all the piconets that have been at the hotspot. b) Frequency Occupancy: In each time slot the probability that a frequency is used at the hotspot is calculated for each frequency, by using the information about the hopset of each piconet. At each time slot we pick the value of this probability that is maximal among all the frequencies. The Frequency Occupancy is calculated as an average of these maximal values over all time slots in the simulation. Thus, the higher Frequency Occupancy has the potential to cause larger interference at a speci?c frequency as a collective entity and has worse performance in terms of etiquette.
Fig. 2.
0.5 0.45 0.4 Frequency Occupancy 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0

Average throughput against average number of piconets.

DAFH-AT DAFH-CT W/O AFH

?

2

3

4

5 6 7 Average Number of Piconets (N)

8

9

10

Fig. 3.

Frequency occupancy against average number of piconets.

??

??

B. Simulation Results In the following simulation results, we adopt the minimal acceptable value of doubling period ? [slots] that was obtained by the preliminary simulations for DAFH. This minimized value of ? re?ects better conformance to the etiquette when a piconet leaves the hotspot. Fig. 2 shows the average throughput versus the average number of piconets. Here, in addition to the proposed DAFH with adaptive threshold (DAFH-AT), we show the perfor? , of PER mance of DAFH where the threshold value, for triggering is ?xed at the maximal estimated PER due to the noise in the channel (we call this DAFH-CT (Constant Threshold)). From this ?gure, we can see that, if no AFH is applied, the throughput is severely degraded as the number of collocated piconets increases. This is due to the fact that each piconet uses the complete sets of frequencies, such that the probability of collision increases almost linearly with the number of collocated piconets. On the other hand, DAFH signi?cantly improves the average throughput as compared to the system without AFH. This is because DAFH succeeds in the orthogonal assignment of hopset to each piconet for most of the staying period in the hotspot. Furthermore, we can see that DAFH-AT has higher average throughput than DAFH-CT for larger average number of piconets. The number of triggerings in DAFH-CT is much larger than in DAFHAT due to the constant threshold determined irrespective of the interference conditions, resulting in the larger number of overhead packets out of the total successful packets. Fig. 3 shows frequency occupancy against the average num-



???

??

1 0.95 0.9 Average Throughput 0.85 0.8 0.75 0.7 0.65 0.6 0.55 0.5 2 3 4 5 6 7 Average Number of Piconets (N) 8 9 10 PER due to noise errors = 1 % DAFH-AT DAFH-CT W/O AFH PER due to noise errors = 0.1 % Frequency Occupancy

0.5 0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 2 3 4 5 6 7 Average Number of Piconets (N) 8 9 10 PER due to noise errors = 0.1 % DAFH-AT DAFH-CT W/O AFH PER due to noise errors = 1 %

Fig. 4. Average throughput against average number of piconets with packet errors due to noise errors.

Fig. 5. Frequency occupancy against average number of piconets with packet errors due to noise errors.

ber of piconets. Here, the frequency occupancy of the system without AFH represents a lower bound to the frequency occupancy that can be achieved by system with DAFH. From the ?gure, we can see that DAFH-AT has lower frequency occupancy than DAFH-CT, especially when the average number of piconets is small. This is because DAFH-AT attempts to keep the hopsets as large as possible and triggers the piconets with larger hopsets with higher probability. The adaptive threshold prevents piconets with smaller hopsets (i.e. higher interference at the used frequencies) from further hopset reduction. On the other hand, for the larger average number of piconets, there is no signi?cant difference between DAFH-CT and DAFH-AT since both systems have to use hop sets with minimum number of frequencies (?? ? ) in order to accommodate all present piconets while keeping the orthogonal hopset assignment. Figs. 4 and 5 respectively show the average throughput and frequency occupancy against average number of piconets with packet errors due to channel noise. From these ?gures, we can see that the performance of DAFH-CT is largely degraded %) is much even when PER due to noise errors (?? %). For smaller than the maximal estimated value (?? ?? %, piconets with any hopset DAFH-CT, if ?? are always triggered to reduce the hopset, irrespective of the number of interfering piconets. This results in even worse average throughput than the system without AFH. On the other hand, DAFH-AT accounts for errors due to channel noise independently of the errors due to self-interference in the threshold design, and can adjust the threshold according to the number of interfering piconets. That is why DAFHAT is more robust to packet errors due to channel noise than DAFH-CT.

DAFH-AT can adaptively self-allocate appropriate frequency hopsets according to the number of interfering piconets at the hotspot. Also, the DAFH-AT shows the robustness to the channel noise, and exhibits a best-effort behavior from the perspective of the interference toward the entities operating in the same unlicensed band. R EFERENCES
[1] Bluetooth SIG, “Bluetooth Core Speci?cation,” Feb. 2001. [Online]. Available: http://www.bluetooth.com [2] IEEE 802.15 Working Group for WPAN . [Online]. Available: http://grouper.ieee.org/groups/802/15/ [3] IEEE 802.15 Coexistence Task Group. [Online]. Available: http://www.ieee802.org/15/pub/TG2.html [4] A. El-Hoiydi, “Interference between Bluetooth networks - upper bound on the packet error rate,” IEEE Commun. Lett., vol. 5, no. 6, pp. 245–247, June 2001. [5] F. Mazzenga, D. Cassioli, P. Loreti, and F. Vatalaro, “Evaluation of packet loss probability in Bluetooth networks,” in Proc. IEEE Intl. Conf. Commun (ICC’02), vol. 1, New York, Apr. 2002, pp. 313 –317. [6] C. F. Chiasserini and R. Rao, “Coexistence mechanisms for interference mitigation between IEEE 802.11 WLANs and Bluetooth,” in Proc. 21st IEEE INFOCOM, vol. 2, New York, June 2002, pp. 590 –598. [7] B. Treister, K. C. Chen, and A. Batra, “Clause 14.3: Adaptive frequency hopping,” IEEE P802.11 Working Group Contribution, IEEE P802.15-TG2-366r1, July 2001. [8] B. Zhen, Y. Kim, and K. Jang;, “The analysis of coexistence mechanisms of Bluetooth,” in Proc. IEEE VTC Spring ’02, Birmingham, Alabama, May 2002. [9] C. de M. Cordeiro and D. P. Agrawal, “Mitigating the effects of intermittent interference on Bluetooth ad hoc networks,” in Proc. 13th IEEE Intl. Symp. PIMRC’02, vol. 1, Lisbon, Sept. 2002, pp. 496 –500. [10] A. Zanella, A. M. Tonello, and S. Pupolin, “On the impact of fading and inter-piconet interference on Bluetooth performance,” in Proc. IEEE WPMC ’02, vol. 1, Honolulu, Hawaii, Oct. 2002, pp. 218–222. [11] G. Pasolini, “Bluetooth piconets coexistence: Analytical investigation on the optimal operating conditions,” in Proc. IEEE ICC ’03, vol. 1, Anchorage, Alaska, May 2003, pp. 198–202. [12] T.-Y. Lin and Y.-C. Tseng, “Collision analysis for a multiBluetooth picocells environment,” IEEE Commun. Lett., 2003, to appear. [13] L. Kleinrock, Queueing Systems Volume 1: Theory. John Wiley & Sons Inc., 1975. [14] T. Camp, J. Boleng, and V. Davies., “A survey of mobility models for ad hoc network research,” Wireless Comm. and Mobile Computing (WCMC): Special Issue on Mobile Ad Hoc Networking: Research, Trends, and Applications, vol. 2, no. 5, pp. 483–502, 2002.

??

?

?

V. C ONCLUSIONS In this paper, we have addressed the problem of selfinterference among collocated WPANs based on frequency hopping, and proposed Dynamic Adaptive Frequency Hopping (DAFH) mechanisms to avoid the throughput decrease of individual WPANs due to self-interference. The throughput optimization is not the sole target of the proposed DAFH, because WPAN operates in unlicensed spectrum and arbitrary adaptation of the FH pattern may be harmful to proximate nonWPAN devices. This has motivated us to embed etiquette rule into the DAFH design. The simulation results have shown that


相关文章:
更多相关标签: