Capture Aware Channel Access Protocol in Wireless Network

: Spatial reuse in wireless networks is limited by the SINR threshold and it might be feasible to capture a packet in the presence of ongoing foreign transmission. This work considers a new capture aware channel access protocol by incorporating global channel state information in the decision making process for the channel access jointly with adaptive power framework. The protocol employs power heterogeneous ad-hoc networks; it assigns different transmission power level to individual nodes based on dynamic observation of the network traffic. It exploits spatial heterogeneity of flows at a given channel allocation and sets up either symmetric or asymmetric carrier sensing which in turn schedules the data packets transmission to maintain adequate service quality and fairness enjoyed by a user. It stands atop capture capable PHY to leverage the channel reuse which is of paramount importance in the design of high capacity ad-hoc networks. Through extensive simulations, the paper demonstrates the efficacy of the new protocol. It delivers high network utilization and also provides fair access to the media.


Introduction
A key problem in ad-hoc wireless networking is the design of distributed media access control (MAC) mechanism to discover and negotiate the access to the reusable physical channels such that the transmission can occur without collision.A representative mechanism of MAC is the Carrier Sense Multiple Access/Collision Avoidance (CSMA/CA) algorithm that forms the basis of the dominant wireless multiple access protocol Distributed Coordination Function (DCF) in 802.11 (Kurth and Redlich 2009).Physically a channel is sensed busy either when the detected energy level is above a certain threshold ζED, when a signal with the same PHY characteristics is detected, or a combination of both (Larroca and Rodríguez 2014).
Carrier sensing is the main interference mitigation mechanisms used in the PHY/MAC layers of 802.11WLANs, and its efficacy becomes a key factor in determining network capacity.A station defers its transmission for a random period upon detecting a carrier on the channel.However, the absence of a carrier does not mean a transmission will succeed, nor does the presence of a carrier mean that a transmission will interfere.Hereby carrier sense approach can unnecessarily reduce network throughput and strongly affect the delay characteristics at a station (Jamieson et al. 2005;Zhu et al. 2004;Ma et al. 2009).Many studies have been done to illustrate that physical carrier sense adversely limits the effective bandwidth utilization in the network because of fixed carrier sense threshold and inadequately chosen carrier sense range.Extensive research efforts advocate the tuning of carrier sense threshold for utilization improvement (Ma et al. 2009;Zhu et al. 2004;Zhu e et al. 2006;Haghani et al. 2010); however, many of today's 802.11MAC implementations do not allow the threshold to be independently tunable or even accessible (Zhu et al. 2004).Moreover, (Pelechrinis 2009) claims that tuning carrier sense threshold opens the door for selfish users with lower back-off times to ignore other transmission and therefore obtain a higher unfair share of the spectrum.
There is a clear trade-off between the probability of interference and the utilization in setting carrier sense range.Larger sense range decreases both the probability of interference and the bandwidth utilization, whereas small carrier sense range will be more subject to interference.
Exposed node prevents a successful transmission while hidden node wastes a transmission opportunity due to a collision.As has been pointed out in (Ma et al. 2009), the optimum value of carrier sense threshold is that at which the carrier sense range of the transmitter just covers the interference range of the receiver.
The setting of interference range is rather heuristic and remains an open problem.Many studies on wireless networks have largely considered an interference range as twice the transmission range, and others rely on equal interference and transmission ranges.Physically the interference range is defined to denote the range within which an interfering node will lower Signal-to-Interference ratio (SIR) below the threshold for successful decoding of a transmission (Maheshwari et al. 2008).It has been depicted in (Chen et al. 2007 andLee et al. 2010), an interferer's impact becomes serious when its preamble is detectable such that the intended end-user device could engage with, even if it is not able to decode the frame correctly.Hereby the author recommends characterizing the interference range within which the preamble of a frame is detectable and the carrier sense range should be equal to this range.
Recall that spatial channel reuse relies on the carrier sense threshold and the transmit power each station uses.MAC like 802.11 DCF uses the same transmission power at all nodes regardless the link distance between transmitter-receiver pair as well as interference level at the receiver (Li et al. 2009).Moreover, it always transmits at full power which in turn potentially maximize the link utility from the medium.However, this selfish behavior leads to excessive mutual interference and degrades the network performance.In contrast, some proposed MAC mechanisms (Gomez et al. 2001;Lim and Yoshida 2005) transmit at minimum possible power to leverage the spatial reuse hence maximizing the interference probability.
The SINR threshold for successful reception depends on the timing and the relative order of signal and the interference at the receiver (Chen et al. 2007;Lee et al. 2010;and Mustafa 2015).The radio captures a new stronger frame above a predefined preamble capture threshold γP, if it arrives during the first frame's preamble reception.802.11'sPHY layer continuously monitors the received signal strength even the layer in data reception state.It enables a radio to correctly receive a second frame, even after it has already synchronized with the first one, if the ratio of powers is sufficiently high.This is called Message in Message MIM mode which is not exploited before in capture aware mechanisms.A typical ratio of powers required for body capture is γD=10dB (Chen et al. 2007 andLee et al. 2010).Thus, the SINR at the intended receiver and transmission order are crucial factors for successful delivery.
This work presents an approach to leverage ad-hoc network's capacity relying on a distributed capture aware framework to coordinate the transmit power and time schedule the traffics to exploit the PHY capture effect with ultimate goals of improving utility and spectrum fairness.Time schedule the traffic is done by assigning transmit power level to set up either symmetric or asymmetric carrier sense scenario that can potentially schedule the data packets transmission.Asymmetric carrier sense situation is neither actually introduced nor applied in previous works.It is known that a hidden link would starve in presence of sensed link with high load traffic.The simulation study offers a new prospective regarding the asymmetric carrier sense scenario and demonstrates its benefit to time schedule the hidden traffic without an extra timing signal.
In this work, Capture Aware Multiple Access Protocol (CAMA) is proposed.It requires disseminating the locally measured link gain at individual stations throughout the network neighborhood for identifying links that can be concurrent and allowing them to coexist, thereby enhanceing the effectiveness of channel reservation.It doesn't modify the carrier sense threshold.CAMA exploits spatial heterogeneity of flows at a given channel allocation by means of dynamic power control to manage the cochannel interference and allow for higher and fair utilization of resources.
The rest of the paper is organized as follows: section 2 goes through works related to this paper.Section 3 describes CAMA mechanism that adjusts the transmission power of the nodes to maximize the benefits of the capture effect.Section 4 presents the simulation results that describe the behavior of the default and CAMA mechanisms, and demonstrate the benefits of the proposed protocol.Finally, Section 5 concludes the paper with general observations and suggestions for future research.

Related Work
Extensive works have subjected different aspects of 802.11MAC protocol to enhance perflow and aggregate throughput.This section presents an overview of some existing work that aims to improve network capacity by spatial diversity through tuning transmit power.Ding et al. 2005 proposes transmit power control protocol to adjust the transmit power for each frame under the same decision rule as IEEE 802.11 which in turn achieves limited improvement.Mhatre et al. 2007 claim that different power levels at transmitters use the same carrier sense threshold in 802.11 networks introduces asymmetric links which in turn lead to throughput starvation.They demonstrate that cross layer approach by joint tuning of the clear channel assessment CCA threshold and transmit power ensures starvation free control system.Furthermore, the clients use the same transmit power and CCA threshold as their associated access point regardless of different channel state each sees.The current work demonstrates that asymmetric link do not necessarily introduce starving nodes.
The traditional access control cannot handle multiple packets without declaring a collision.A capture aware MAC mechanism was proposed in (Santhapuri et al. 2007), where primary node transmits Request-To-Send (RTS)/Clear-To-Send (CTS) and waits for one preamble time interval before transmitting data while secondary node(s) that overhears RTS will initiate its own transmission either during wait time for the primary user or during primary's DATA transmission depending on the estimated SINR at its receiver and at the primary user.However, this scheme doesn't address the synchronization issue due to the existence of propagation delay, and it incorporates that receiving the preamble of the interference frame helps to detect the frame of interest that arrives later while in contrast has been proved in (Lee et al. 2010).
In (Li et al. 2009), nodes transmit RTS/CTS at the same maximum power, and transmit DATA/ACK at the power based on the link distance, as well as the interference level at the receiver.Based on the overhead RTS or CTS, the protocol enables a concurrent transmission at a possible highest power level as long as it does not interfere with the ongoing ones.Although, the protocol considers the aggregate interference level, but it assumes equal contribution from the neighboring nodes, no matter how close they are as well as the transmit power they use.
There are some limitations on these mechanisms: hidden terminal problem still exists with RTS/CTS, and possible loss in ACK packets is ignored to benefit from fewer collisions and build inaccurate results.
Various Further, these mechanisms don't exploit MIM mode and the assignment of concurrent links set satisfies a single SINR constraint.The study in (Mustafa 2015) demonstrates the inaccuracy of the well-known interference physical model.

Capture Aware Multiple Access CAMA Framework
The 802.11 DCF (Li et al. 2009) is strong transmit power MAC design wherein all nodes transmit with the same power no matter how close the sender and receiver are.The current work here claims that by moving beyond the maximum fixed transmit power, the fairness and channel utilization can be improved.Further, successful unicast transmission depends on channel condition at both the receiver and sender due to downstream data traffic and upstream MAC layer acknowledgment ACK traffic.However, suggestion of simply extending the carrier sensing mechanism to the receiver does not address the MAC limitations of inefficient and unfair spectrum utilization.A capture aware multiple access mechanism is proposed based on broad channel observation to tune power level of each node to successfully retrieve the transmitted signal and make a decision to access the channel, while carrier sense range is fixed and set equal to interference range.In environment with users of homogenous technology competing for resources, the interference range in this work is characterized within which the preamble of a frame is detectable.Depending on propagation paths between the four stations, link(s) may or may not be established between the nodes.

Two Sender-receiver Pairs
Psi denotes transmit power level at Si and η is the noise floor. ,  denotes Si-Rj link gain where a propagation model suited for the surrounding environment is used to describe the fluctuation in the received signal strength such that the mean received power falls off as (1/d) α .α is a path loss exponent.A receiver can engage with a delivered packet if the received signal power is greater than the receiver sensitivity threshold ζS, i.e.
Whether parallel data transmissions of data packets are successful or failure depends on the observed SINR basically at receivers that should be higher or equal to γD.It enables the radio to correctly receive an intended frame irrespective of the timing relation between overlapped transmissions.γD is usually set to 10 that map to ten times the signal power over interference power, as stated earlier in first section.This condition holds if Ri is closer to Si than to both Sj and Rj by a sufficient factor, and vice versa.It may happen that this condition holds even though CSMA/CA would prevent the second pair from transmitting.
To illustrate this possibility, consider the following example.The four nodes are on a line in the order, from left-to-right R1; S1; S2; R2.Their locations on the x-axis are 0; 1; d+1; d+2, respectively.Assume all the nodes transmit with a unit power and ( exceed γD and ζD, respectively.Further constraint should be followed since it is also possible that a data packet is received correctly, but the ACK packet is lost Briefly, the channel is utilized efficiently only if S1 commence transmission before S2.
Flow of higher transmit power can afford to start later.It is important to highlight that we are not using a special timing signal to schedule the data transmission; we only depend on carrier sense scenario to time order the traffic.
It is well understood that asymmetric links in power heterogeneous network unfairly degrades the utility experienced by low power links (Mhatre et al. 2007, Shah et al. 2007).However, our simulation study confirms that diversity in transmit power does not necessarily starve low power users if some constraints are followed.

N Sender-receiver Pairs
The intended stations are responsible for estimating and disseminating the channel quality around to make a joint prediction about whether more than a flow can transmit simultaneously through minimum possible transmit power at each station to mitigate the interference and optimize the per-flow and overall network throughput on a collision free channel.If the prediction result is positive, the stations should not block their own transmissions, if any.Otherwise, an approach that considers tuning of transmit power as a PHY parameter, should be applied to set a suitable sense interaction and subsequently schedule the access to a shared medium.
Simulation results in section 4 illustrate that such an access scheme performs better than the CSMA/CA even at high traffic loads, and applying it has the benefit of improving perflow and aggregate utility in wireless networks.
To estimate the complexity of such a protocol, assume there are N sender/receiver pairs willing to access the media.The nodes in pair i should estimate the channel gain over link(s) that may establish;  ,  ,  ,  and  ,  ∀ neighbor pair j, either periodically or whenever it has data to transmit and/or node(s) observes degradation in the performance.They then post these values to a distributed control system to coordinate the channel access and transmit power of the nodes according to the specified policy.The stations may exchange simple control signaling to initiate the process and run the optimization algorithm.Despite that the optimal algorithm provides optimal utility of the channel, it is rather computationally expensive for a large network.
If the simultaneous transmissions could ban the capture on a subset of M links, higher transmit power should be assigned at the senders of these links that being oblivious of lower power flows within their sensing range.SINR constraints should satisfy the earlier inequalities at the receivers of the M links.SINR experienced at the receivers of the remaining N-M links should exceed ζD as given by Further, SINR experienced at the senders should exceed ζD for successful delivery of ACK packets, as given by CAMA determines the appropriate minimum transmit power that ensures that the sender can sustain a data rate to reach the intended receipt and the interference level perceived at other nodes can be mitigated thereby higher number of concurrent transmissions can be achieved.

Simulation Results
In this section, we present simulation-based studies using NS-2.34 which incorporates the modeling details of the IEEE 802.11MAC and PHY modules (Chen et al. 2007).The PHY module includes cumulative SINR computation, preamble and PLCP header processing and capture, and frame body capture.The MAC accurately models transmission and reception coordination, back off management and channel state monitoring in a structured and modular manner to presents the CSMA/CA mechanism.
All nodes implement the 802.11g technology.The senders and receivers are placed in an indoor environment, and the radio propagation reflects shadowing model with path loss exponent of 4 over the distance (Srinivasan and Table I lists the main parameters in our simulation study.The minimum PHY bit rate δ=6Mbps is considered in the simulation.Hence the minimum SINR which guarantees a reliable packet communication ζD=ζP=5dB, as a hardware defined threshold (IEEE Std. 2007;Chen et al. 2007;and Lee et al. 2010).The simulation is restricted to one-hop unicast UDP traffic.Each link carries constant bit rate (CBR) traffic.Each simulation runs for 60 seconds with disabled RTS/CTS virtual carrier sensing.The nodes are configured to be always backlogged with packets to send and each MAC data frame to be 1028 bytes long.
During the simulation we consider utility in terms of MAC layer goodput at the receiver and Channel Usage Efficiency CUE in terms of the ratio of MAC layer goodput at the receiver to PHY layer throughput at the transmitter over a given channel as two distinct metrics.We first have activated one link to find the maximum achievable throughput, goodput and CUE of unicast traffic that are approximately 5.19 Mbps, 5.05 Mbps and 0.973 respectively due to MAC/PHY layers overhead.Low CUE indicates that a node has gained access with failed transmission which consumes energy and may cause interference to neighbors over the same channel.
Fairly Shared Spectrum Efficiency (FSSE) index is considered to measure the portion of System Spectral Efficiency SSE that is shared equally among all active users.In case of scheduling starvation, FSSE would be zero during certain time intervals.In case of equally shared resources, FS SE would be equal to SSE.If FSSE is maximized, the max-min fairness can be achieved (Eriksson 2001).

Performance Evaluation of Default DCF
Determining how well carrier sense works in wireless networks is the focus of the simulation studies in this section.We show few examples where the default bandwidth utilization by the 802.11MAC is far from optimal setting.Carrier sense may sometimes make incorrect access decisions thereby leading to inefficient channel usage, as explained before when exposed and/or hidden terminals are present.Utility in terms of goodput and CUE are our basic metrics for comparison of link performance across various carrier sensing settings and for measuring the efficiency of default channel access scheme.
Consider a scenario of a simple four nodes topology in a 100x100m2 indoor environment to compose two wireless links, with symmetric incomplete view of channel state in Fig. 2a.The senders are within sense range of each other and within the capture range of their intended receiver.Hence, if S1 commences a data transmission, S2 should not block its own transmissions, if any, and vice-versa despite the presence of the ongoing data delivery.They could simultaneously transmit without causing excessive interference at the receivers to disrupt  successful communication.However, the default MAC mechanism based on local channel assessment may mispredict the channel state and consequently waste the transmission opportunity.
Figure 3 illustrates the normalized utility on two communication links L1 and L2 when the senders depend on local channel observation only or extend the observation to the intended receiver.Having a single exposed node potentially drops the overall throughput to 50% of the optimal setting.
Further complications arise in cases where hidden nodes are likely, scenario (b).Suppose S1 is oblivious of ongoing flow from S2 and viceversa.The receivers are in close proximity such that S1 (S2) data transmission and/or R1 (R2) ACK transmission could corrupt R2 (R1) reception.The incomplete view of channel state possessed by the sender causes hidden terminal where the absence of a carrier at the sender does not mean a transmission will not collide.Both links experience packet loss due to the concurrent transmissions from the hidden terminals.The senders double their contention window size for each failure of data transmission; hereby an obvious drop in the channel utilization is illustrated in Fig. 3. Misprediction of channel state can drop the utility significantly than if an access decision is extended to the intended receiver.
It is obvious that exposed and hidden terminals in the previous scenarios affect the two flows equally.The third scenario in Fig. 2 considers a topology of hidden terminals with heterogeneous link quality.S1 and S2 are out of each other's sensing range, and S2 cannot corrupt R1 reception.However, S1 may initiate a transmission while S2 is transmitting, and disrupt the reception at R2.If S1 occupies the channel for a long time, S2 will starve.
The simulation in Fig. 3 shows extreme unfairness due to asymmetric interference and diverse transmission opportunity the senders have, which cannot be addressed in this scenario with extending the channel observation to the intended receiver.The unfairness in access opportunity which is undesirable as it can adversely affect delay sensitive applications cannot be addressed without coordination with the surrounding nodes.
Next, we have evaluated the performance of hidden terminals in the topology of Fig. 2c with different traffic load on the wireless links and default MAC mechanism.S1 is located within the capture range of its intended receiver while S2 is not always depending on the traffic rate on the interfering link L1.
Regarding the simulation results in Fig. 4 in terms of normalized utility on L2, R2's default MAC layer has recorded nothing at traffic load of 3Mbps.It could work better at low traffic load wherein the target utility of 1Mbps is attained.However, CUE metric considered here is not satisfying even with low traffic load.Failed transmissions not only alleviate CUE and waste energy but also have the potential to corrupt other transmissions.
Generally, the overall performance claim that the default mechanism doesn't work efficiently even with low traffic load, and simply extending the observation to the intended receiver don't address the limitation.An intuition to maximize the utility and channel usage efficiency is to set up capture aware MAC scheme jointly with adaptive transmit power as has been described in section 3.

Performance Evaluation of CAMA
In this section we assess the performance of our proposed framework to mitigate the interference and leverage the spatial reuse via a broad picture of the channel state.In this paper, we assume that information exchange works perfectly as well as the coordination.We conducted three different sets of simulations using different number of sender-receiver pairs use CBR applications on the same channel.

First Scenario
First, we have conducted a number of simulations using two links for illustrative purpose.The position of nodes is sampled from a two-dimensional distribution in 100x100 m 2 indoor environment, as shown in Fig. 5.The receivers are deployed in overlapping area of the senders and each tolerates a certain level of interference depending on channel gain and the transmit power.The unicast traffic considered is generated and flowed at a rate sufficient to saturate the medium.The link gain is assumed as being fixed for the duration at which power updates are performed in CAMA framework.
Fig. 6a shows the normalized utility experienced by users to illustrate the performance improvement compared to the default DCF at different power levels.The default MAC mechanism depends on local channel assessment to decide whether to commence a transmission or to defer it.With default DCF and transmit power level of 19dBm (Case1), the senders are hidden and transmit independently.L2 hits the maximum utility with sender in capture range of the intended receiver, whereas L1 is seriously affected by co-channel interference.Fig. 6b shows inefficient channel usage by S1, in contrast to S2 that being oblivious of the collisions experienced by R1.Normalized Utility Higher transmit power can be used to set symmetric carrier sensing between the senders, for example 27dBm (Case2).The goal is to silence a sender when the other initiates a transmission.Slight different performance: 0.48 and 0.59 of the maximum utility is hit, respectively over the transmission links.Collisions occur at R1 when two senders choose the same slot in the contention window to transmit their frames since L1 observes less degree of channel quality; however, this case doesn't happen as often compared to the case when the senders are hidden from each other.
While the default solution for sender out of the capture range of the intended receiver, is to preclude the concurrent transmissions, CAMA framework shows the existence of optimal transmit power for each node in this scenario to set up asymmetric carrier sense and satisfy SINR constraints, where an obvious improvement in per-flow and aggregate utility is resulted (Case 3).By switching for example to transmit power 27dBm at S1 and 19dBm at S2, R1 and R2; S2 is able to carrier sense data transmissions from S1 but not vice-versa.
The sender S2 can access the channel during time interval between successive transmitted frames of high power sender within its sensing range.S1 doesn't block any transmission request from an upper layer that could overlap with existing flow.Moreover, if S1's frame arrives when S2 is in transmission state, S2 will not be able to hear this frame even if its transmission would end very soon after, since the received energy from S1 is below energy detection threshold and S2 would have missed the preamble and PLCP header.We confirmed the above argument by tracing the transmission order and inter-transmission time NS-2.34 trace file.
Albeit the transmission time of the terminals could overlap and eventually the frames based on the traffic size, the concurrent transmissions with the new setting of transmit power don't introduce enough interference to deprive the reception.Low power data flow survives the collision with the early interference frame from S2 with | −  1 ≥10dB and hits the maximum utility.The senders successfully and fairly utilize the channel wherein normalized utility and CUE results in Fig. 6 are almost one.The results demonstrate no starving with asymmetric links in power heterogeneous network, and confirm the efficacy of CAMA framework to provide collision free channel.Our simulation results with unicast traffic in contrast to the results have been observed in (Rao and Stoica 2005) with broadcast traffic over asymmetric sensing links.Extended Inter Frame Space (EIFS) is supposed to be the main reason for the unfairness in the transmission opportunity.
Figure 7 shows FSSE and SSE results wherein CAMA mechanism obviously has much better spectral efficiency and fairness since it gets off hidden and exposed node scenarios, which are the well known sources for the limitation of the default mechanism.The results shows that coordinated and collaborative framework enables efficient use of the wireless resource and achieves potential improvement over DCF.

Second Scenario
We substantiated the proposed CAMA mechanism on larger networks of senderreceiver pairs distributed in three-dimensional distribution in 100x100x100m 3 indoor environment with disparity in link quality.The probability with which a given frame could be captured by intended receipt is a function of the number of ongoing transmission.The energy received from the intended packet should be higher than total accumulated energy received from the remainder parallel transmission over the capture interval, as defined in the equations ( 5)-( 8).The optimal solution consists of defining a set of willing end-users to transmit, and apply CAMA mechanism to enable possible concurrent transmissions over the channel.
To portray the impact of the aggregate interference, consider a set of four senderreceiver pairs (Fig. 8).The distance between the senders S1, S2, S3 and S4 is roughly 40m except the distance between S3 and S4 is 70m.The position of the receivers is sampled such that R3 is within the interference range of S1 and S2, i.e. either R3 is able to decode the foreign frame correctly or not, it could engage with overheard S1 or S2's preamble.R4 is also within the interference range of S1.Each flow is sending separate CBR flows to the corresponding receiver on the same channel.
In a scenario where the nodes have no knowledge of ongoing hidden transmission (Case1), S1 and S2 monopolize the channel and achieve the maximum utility at the expense of the starving nodes S3 and S4, as simulation results in terms of utility and FSSE demonstrate in Figs. 9 and 10, respectively.Average CUE in terms of the ratio of the aggregate MAC layer good put at the receivers to the aggregate PHY layer throughput at the corresponding transmitters, is also presented in Fig. 10.We have recorded 10.4Mbps an aggregate utility.In the next case (Case2), mutual sensing leverages FSSE and drops SSE with an aggregate utility reaches roughly 7.2Mbps.Capture Aware Channel Access Protocol in Wireless Network Later, we have applied CAMA mechanism in Case 3 to find the appropriate transmit power for each node depending on the interfering environment it sees, such that parallel uplink and downlink transmissions cannot introduce enough interference to corrupt the reception at neighbor send-receive pairs.
Higher transmit power of 23dBm and 21dBm are set at S3 and S4, respectively due to the link quality each sees.Lower power level 13dBm is set at the remaining nodes such that:  S1 and S2 are sensing S3 and S4's traffic but not vice versa. S1 and S2 transmit independently as well as S3 and S4 since they don't experience peer interference.
Power heterogeneous network is required here to schedule the possible largest number of successful concurrent transmissions in a single time slot on a single wireless channel.Various power levels are set at the nodes to drive a sender in or out of sensing range of others herby schedule the time to transmit, and to satisfy the SINR constraints in equation (5-8).An obvious improvement is portrayed in the simulation based results.
Energy detection threshold here is set at the default value.Individual transmission of S3 and S4 results in SNR higher than ζED and ζP, respectively at S1 and S2 that enables either S3's preamble or energy detection, and S4's energy detection if S1 or S2 are listening.It is obvious that their concurrent transmission is also detectable.As for individual links, we observe that power adaptation according to CAMA maximizes S3 and S4's utility only, and their saturated traffic especially of S3 reduces obviously other's opportunity to access the channel.Nevertheless, obvious improvement in FSSE and SSE is attained.
Later in Case4, higher ζED is configured such that S1 and S2 do not sense a delivered energy of ongoing transmission when the corresponding preamble is missed.Preamble detection only is required to schedule the instantaneous transmissions of senders with homogenous technology.S1 and S2 knock the channel frequently to achieve the maximum utility and subsequently lead to the maximum fairness, though the power levels are not changed.Fairly and efficiently channel utilization demonstrates the efficacy of CAMA framework.

Third Scenario
We next randomly distribute 20 and 50 nodes in a 100x100 m 2 indoor environment, respectively.Each sender has its own receiver chosen randomly from the set of immediate neighbors.A random traffic model at each sender is used with a sufficiently high load such that a sender to be always backlogged with packets to send.A high ζED is configured at each node such that they do not sense a delivered energy of ongoing transmission when the corresponding preamble is missed.In this case study, CAMA uses a genetic algorithm GA approach for optimal power distribution in a network.The GA is applied to find the optimal power level for each user under different operating conditions based on the constraints defined earlier.The problem is formulated as a multi-objective optimization problem which aims at maximizing the utilization and fairness.GA can work even when the objective function is not exactly known since it relies only on an objective function's evaluation.It is well-known for their remarkable generality and versatility, and has been applied in a wide variety of settings in wireless networks.The author skipped the details of GA since it is not in the scope of the work rather than for applying an optimization algorithm in CAMA framework.For details on GA, the author recommends (Sivanandam and Deepa 2007).
Figure 11 illustrates the aggregate one-hope utility achieved using 20 and 50 nodes in a random topology using the default DCF and CAMA as a channel access mechanism.The aggregate utility is found as the aggregate throughput of concurrent one hop communication in the network.The results are averaged over three independent simulation runs.Our approach offers clear improvement against DCF.
Figure 12 illustrates the average value of the FSSE and CUE using the default DCF and CAMA in random topology of 50 nodes.Performance evaluation in terms of fairness and the channel usage demonstrates the significant improvement using CAMA based on a broad observation of the channel state.The results are achieved using default DCF and CAMA mechanism.Through the investigated use cases, asymmetric carrier sense is introduced and applied in CAMA.The simulation results portray the benefit of asymmetric sense scenario to time schedule the overlapped traffic without an extra timing signal.CAMA employs power heterogeneous adhoc network to optimize criteria involving throughput and fairness.This requires obtaining effective spatial reuse while satisfying the interference constraints.

Conclusion and Future Work
Based on local channel assessment, the default DCF mispredicts transmission failures and wastes potential transmission opportunities when they exist.It ignores the possibility of parallel transmissions that make use of the capture effect.It can be ineffective even at low traffic load.
This paper proposes a new captureaware mechanism for ad-hoc networks while their benefit is amplified by properly adjusting the transmission power of the nodes based on broad observation of the channel.Through simulations, such a mechanism is shown to deliver high network utilization and provide fair access to the media than the standard DCF that these networks use.Using the simulator trace files, we have found that tuning transmit power is helpful in setting specific sense scenario and subsequently schedule the access to a shared medium without modifying carrier sense threshold, as has been exploited in CAMA framework in an interference-free fashion.
The nodes cooperate in power updating process via exchanging simple control signaling in power heterogeneous network.Further, based on the simulation results, we have found that throwing energy detection carrier sense line offers higher network utility.
In environment with heterogeneous technologies, foreign preambles are undetectable and energy detection may results in lower per-flow utility.However, the target SINR by CAMA at the corresponding heterogeneous nodes is expected to be lower since a receiver would not engage to a first coming foreign preamble, which in turn enables more concurrent transmissions and records higher aggregate utility and overall spectral efficiency.
Our future work will demonstrate it.We also propose to explore how a similar mechanism can be used to adjust more PHY parameters like the PHY rate or/and contention window size as MAC parameter.

𝑆𝐼𝑁𝑅|Figure 1 .
Figure 1.Two flows sharing a wireless channel.Depending on propagation paths between the four stations, link(s) may or may not be established between the nodes.
. Parameters as antenna gain and system loss are assumed to be fixed.The default values of configuration parameters are used.Compared to preamble detection threshold ζP where preamble detection starts to work, energy detection threshold ζED in terms of SNR is higher(Lee et al. 2010).

Figure 2 .
Figure 2. Three scenarios of two flows sharing a wireless channel.In scenario (a) the intended links could always present, given that senders are in capture range of their respective receiver.Scenario (b) portrays hidden terminal case where the intended links could not present simultaneously.Scenario (c) shows a topology of two heterogeneous wireless links in terms of channel gain, given that the senders are hidden terminals and S1-R1 link only is always present.

Figure 3 .
Figure 3. Normalized utility for the scenarios in Fig. 2, in terms of MAC's layer good put with default channel assessment at the senders and extended assessment to the intended receiver.

Figure 5 .
Figure 5. Network model and topology as used in the simulation.The nodes disseminate the estimated link gain to get broader view and perform joint optimization.

Figure 6 .
Figure 6.Performance evaluation in terms of a) Normalized utility, and b) CUE of possible scenarios with default DCF and CAMA.

Figure 7 .
Figure 7. Performance evaluation in terms of fairness and overall spectral efficiency of the scenarios in Fig. 6.

Figure 8 .
Figure 8. Four sender-receiver pairs distributed in three-dimensional distribution in 100x100x100m3 indoor environments with disparity in link quality.

Figure 9 .
Figure 9. Performance evaluation of different scenarios of four heterogeneous wireless links in terms of channel gain, and distributed in 3D indoor environment.

Figure 10 .
Figure 10.Performance evaluation in terms of fairness, overall spectral efficiency and average channel usage efficiency.

Figure 11 .
Figure 11.The average of the aggregate one-hope utility in random topology of 20 and 50 nodes.The results are achieved using default DCF and CAMA mechanism.

Figure 12 .
Figure 12.Performance evaluation of 50 nodes in a random topology in terms of fairness FSSE and average channel usage efficiency CUE, using the default DCF and CAMA mechanism.

Table 1 .
Configuration parameters in the simulation study.