Indonesian J our nal of Electrical Engineering and Computer Science V ol. 22, No. 1, April 2021, pp. 315 325 ISSN: 2502-4752, DOI: 10.11591/ijeecs.v22i1.pp315-325 r 315 Mobility-pr ediction and ener gy optimization f or multi-channel multi-interface ad hoc netw orks in the pr esence of location err ors Hassan F aouzi 1 , Mohammed Boutalline 2 1 Sultan Moulay Slimane Uni v ersity , Morocco 2 National Schools of Applied Sciences of Beni Mellal, Sultan Moulay Slimane Uni v ersity , Morocco Article Inf o Article history: Recei v ed Oct 11, 2020 Re vised Jan 27, 2021 Accepted Mar 2, 2021 K eyw ords: A OD V End to end delay Ener gy consumption Kalman filter Mobile ad-hoc NS2 (Simulator) Routing protocols ABSTRA CT W e present a mobility-prediction and ener gy optimization solution for multi-channel multi-interf ace (MCMI) ad hoc netw orks in the presence of location errors. This solu- tion incl udes routing of the MCMI communication links that adapt to dynamic chan- nel, traf fic conditions, interference and mobility of nodes. W e start first with imple- menting a no v el cross-layer routing solution in order to share information between netw ork and MA C layer , the benefit of t his technique is to collect information about the channel quality and residual ener gy of the nodes and send them directly to the netw ork layer . Ne xt, we present a mobility-predi ction model using Kalman filter to predict accurate locations and enhance routing performance, through estimating link duration and selecting reliable routes. The performance of proposed mechanism is measured using NS2.35 simulations with dif ferent scenarios and v arying load in a net- w ork. Comparati v e analysis of simulation results sho ws better performance of our protocol (ME-MCMI A OD V) in terms of reducing end-to-end delay , total dropped pack ets and increasing netw ork lifetime and pack et deli v ery ratio (PDR). This is an open access article under the CC BY -SA license . Corresponding A uthor: Hassan F aouzi Sultan Moulay Slimane Uni v ersity Beni Mellal, Morocco Email: f aouzi.hassan.mi@gmail.com 1. INTR ODUCTION W ireless netw ork w orks under tw o modes, infrastructure and the other without the aid of central ized administration. Mobile ad hoc netw orks (MANET) does not ha v e a fix ed topology and include a set of wireless mobile nodes which transfer data dynamically among themselv es. Unlik e V ANET netw orks where we can control traf fic by reducing the speed of mo v ement of v ehicles either by the dri v er or by utilizing mechanisms lik e the ones described in the papers [1, 2] which the authors proposed a ne w approach to estimate, track, and control users mo ving abo v e speed limits in L TE-Adv anced (L TE-A) netw orks. T o reach these objecti v es the y use mapping of the uplink CQI inde x of the UE since the CQI range can pro vide an indication to the system re g arding the mo v ement and the speed of the UE. Ho we v er , MANET netw orks do not ha v e this specificity because nodes can mo v e freely in space. The main objecti v e of mobile ad hoc netw orks (MANET) is to e xtend the concepts of mobility to enable access to information and communication “an ywhere and an ytime” using routing protocols as ad-hoc on-demand distance v ector (A OD V) [3], dynamic source routing (DSR) [4], destination sequenced distance- J ournal homepage: http://ijeecs.iaescor e .com Evaluation Warning : The document was created with Spire.PDF for Python.
316 r ISSN: 2502-4752 v ector (DSD V) [5], and optimized link state routing (OLSR) [6] and temporally ordered routing algorithm (T ORA) [7] to transfer data. So, it is important to increase netw ork lifetime by minimizing the control pack ets, because the mobile nodes are po wered by independent po wer sources such as batteries or other consumable ones. In addition, since nodes ha v e limited battery po wer and mo v e frequently in the netw ork, the links between them are frequently brok en and causes a loss of data pack ets. Therefore, the node should send only necessary pack ets and select the path with lo wer link f ailure based on the localization of neighboring nodes to sa v e its ener gy resource. Ho we v er , dynamic topology of MANET mak es it harder to determine the location of nodes in re al time. Hence, our proposed method uses a mobility prediction to measure the link duration time, which can reduce o v erhead messages and impro v e routing performance. Most of the e xisting researches mainly concentrate on the influence of node mobility on link rel iability , the y ha v e ignored the residual ener gy of nodes in their methods and instead of w orking on multi-channel multi- interf ace, the y adapt their mechanisms to use single interf ace single channel (SISC) en vironment. In this paper , ener gy and mobility are introduced as a ne w routing metric to select links i n terms of reliability in a multi- channel multi-interf ace ad hoc netw orks. Firstly , the node in our approach uses mobility to predict the link duration time. Then, it combines this duration with the residual ener gy to find a stable route that has a long lifetime. Finally , it searches a good mapping between channels and interf aces to send data to destination nodes. The remainder of this paper i s or g anized as follo ws: Section 2 introduces the related w ork and discuss some details about Mobility-prediction and ener gy opt imization in MANET . Section 3 describes functional details of the proposed ME-MCMI A OD V to impro v e the A OD V protocol. Section 4 presents the e xperimental modeling and results of our proposed protocol using netw ork simulator NS2. Finally , conclusions and some plans for future de v elopment in this field are gi v en in Section 5. 2. RELA TED W ORK In literature, a lot of research has been done to impro v e the performance of mobile Ad hoc netw orks. These searches seek to find routes satisfying certain constraints. Some ha v e used probabilistic approaches to limit the number of routing pack ets while others ha v e used queue length, bandwidth, mobility , ener gy and hop count that separates the source and the destination. Rare researchers who ha v e tak en into account the location errors in their models. In [8], the authors implemented a mobility-assisted using A OD V protocol and taking into consideration location errors. T o that end, the y implement Kalman filter to predict accurate locations and tak e for granted le v el confidence in disco v ering routes to choose the best route. The authors in [9] proposed tw o e xtensions of A OD V protocol [3] to find routes based on residual ener gy and hop-count, the first uses Flo yd W arshall and the other Bellman-F ord algorithm. These protocols implemented in the netw ork in which, each node equipped with a multiple netw ork interf ace [10] to o v ercome the problems of SISC. Their idea is to add a no v el cross-layer routing solution which allo ws communication between ph ysical and netw ork layers. The authors of [11] ha v e de v eloped tw o topologies namely chain and grid, in which the y further w ork ed on directional and Omni-directional antenna. After simulation, the results sho w that directional antenna is more ef ficient in enhancing the spatial di v ersity and reducing collisions. In an another w ork [12], the authors de v eloped tw o e xtension of A OD V protocol, the first is A OD VEA proto- col, which incorporates local forw arding decision based on max min ener gy of nodes in order to increase the lifetime of the netw ork. The second (A OD VM) combines the same local forw arding decision parameters used in A OD VEA protocol and shortest distance. Instead of using hop count as a parameter to calculate the best routes from source to destination, the authors of [13-15] utilize the ener gy and po wer le v el of the nodes. Simu- lations ha v e sho wn that these impro v ements gi v e better results compared to other algorithms in the same field of research. In [16] recei v ed signal strength from the (MA C) layer is used to estimate the stability of the radio connection. Their objecti v e is to select the stable route to a v oid paths which ha v e a higher probability to be brok en. The authors propose another solution to impro v e Quality Of Servic e by incorporating residual ener gies of source and destination nodes to calculate a v ailable link bandwidth. A recent study [17] proposed a m ethod to frame up a stable link netw ork using a temporal data anal ysis model. In this model, the authors analyzed the mobi lity , position of neighbor nodes and used the statistical model auto re gressi v e mo ving a v erage (ARMA) to predict the stable neighbors of each node in a future time frame. The y applied a Biogeographic -based optimization (BBO) technique to estimate rele v ant parameters in the ideal path from source to destination nodes. According to them, this optimal link of fers a stable and reliable connection for the remaining lifetime of the data transfer in the netw ork. Indonesian J Elec Eng & Comp Sci, V ol. 22, No. 1, April 2021 : 315 325 Evaluation Warning : The document was created with Spire.PDF for Python.
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752 r 317 A dynamic po wer ad hoc on-demand di stance v ector (DP-A OD V) protocol which is an impro v em ent of the e xisting A OD V routing protocol w as applied in [18]. In this e xtension, the authors modified the pack ets headers in the routing layer to include the distance i nformation of the destination and neighbors count. The y also modified hello pack ets of A OD V to carry the “x-y position” coordinate field information in order to obtain the e xact location of a node, which used by the routing protocol to determine the route. At the wireless ph ysi- cal layer , the algorithm selects the po wer required to k eep the connecti vity among the nodes and consequently reduce the o v erall po wer . The authors of [19] use a routing conditioned upon the achie v ement of man y require- ments. Among these, is the intermediate U A V able to respond with a throughput requested by t he source U A V? Does the speed not e xceed a predefined threshold? Simulation results sho w that their protocol (MDRMA) gi v es better results in terms of speed of establishment and stability of routes. In [20], the authors proposed impro v ements of ad hoc on demand distance v ector . These i mpro v em ents tak e into account a metric based on ener gy consumption during route disco v ery in order to decrease load of control pack ets and increase both the netw ork lifetime and pack et deli v ery ratio. The y did the simulation in an en vironment close to realit y by using the Gilbert- Elliot model. T o reduce the route-establishment o v erhead in A OD V , the authors of [21, 22] attempt to minimize the number of intermediate nodes that participate in the route disco v ery process. This is achie v ed by reducing the number of route request (RREQ) depending on the length queue and ener gy of nodes or count of RREQ (nodes stop transfer the requests if the count of RREQ e xceeds a threshold). Additionally , [ 23 ] presented tw o techniques for computing the link a v ailabil ity and decreasing the broadcast of RREQ pack ets. In t h e first technique, the l ink a v ailability ratio (LAR) for all neighboring links is calculated using the present position of the neighbor and its angular sector in the transmission range. In the second technique, the transmission range of each node is di vided into the outer , inner and middle zone. So, based on the recei v ed signal strength and tw o predefined thresholds only the nodes in the middle zone participate in the route disco v ery process. The Multi-path routing allo ws data to be sent o v er a set of paths leading from source to destinat ion. This is wh y other authors choose to w ork on this component [24-26] so as to limit the problem of road disruption and dis trib ute the traf fic between source as well as destination. In summary , the techniques used to decrease the dropped pack ets, end-to-end delay and increase the netw ork lifetime and pack et deli v ery ratio (PDR) in ad hoc netw orks focused only on one performance parameter or based on a single-interf ace single-channel en vironment, and in order to b uild a better routing protocol, we must satisfy all quality services. This is what we tried to implement in this paper . 3. PR OPOSED APPR O A CH AND FUNCTION AL DET AILS In this section, we present our model to impro v e A OD V protocol by using link duration time and residual ener gy as a metric of routing in multi-channel multi-interf ace communications in mobile Ad Hoc netw orks. 3.1. Cr oss-lay er T o tak e benefit of information about the channel quality and residual ener gy of the nodes, we de- v eloped a cross-layer to share this information between netw ork, MA C and ph ysic layer . Although se v eral methods using single channel single interf ace schemes tried to achie v e a high quality of service scheme, most of them, if not all, were not successful due to intra-flo w interference and inter -flo w interference. So, in our w ork we used the multichannel en vironment to solv e t hese problems and pro viding a more reliable MA C protocol for the users. In MANET , channels are separated in frequenc y , so to use the dif ferent channels of fered by the ad hoc netw ork we need to de v elop a channel assignment approaches which allo w coordination between nodes [27]. These approaches classified into three cate gories: Static, dynamic and h ybrid channel assignment. In this w ork, we focus on h ybrid channel strate gy to benefit from the adv antages of static and dynamic assignment. In this strate gy , each node has a multiple interf ace, only one is designed to be fix ed and the others become switchable. When a source node needs to communicate with a destination, it will switch its switchable interf ace to the same channel as pointed by fix ed interf ace of the tar get node. Figure 1 illustrates an e xample of communication between nodes when using fixed and switc hable interfaces ”. Assuming that node X has a data to be sent to node Y . The fix ed interf aces of nodes X and Y are assigning to channels 3 and 1 respecti v ely . T o ensure this communication, the switchable interf ace of node X is assigning to channel 1, before transmitting the pack et, because channel 1 is the fix ed channel of node Y . Mobility-pr ediction and ener gy optimization for multi-c hannel multi-interface ... (Hassan F aouzi) Evaluation Warning : The document was created with Spire.PDF for Python.
318 r ISSN: 2502-4752 So, node Y c an recei v e the pack et since its fix ed interf ace is listening to channel 1. In the replay step, node Y switches its switchable interf ace to channel 3 and send a replay request, which is recei v ed by node X using its fix ed interf ace on channel 3. N o d e   X   F i xe d   i n t e r f a c e   ( C h a nne l   3 )   S w i t c h a bl e   I n t e r f a c e s   C h a nne l     C h a nne l   C h a nne l   N o d e   Y   F i xe d   i n t e r f a c e   ( C h a nne l   1)   S w i t c h a bl e   I n t e r f a c e s   C h a nne l   C h a nne l   C h a nne l   Figure 1. Communication between tw o nodes using fix ed and switchable interf aces 3.2. Residual ener gy The nodes consume the ener gy during transmission and reception acti vities . Therefore, ener gy is one of the actual considerable constrained in MANET . When a node participates in route establishment in man y times, it may run of f its po wer in later stages resulting in the breakdo wn of the link. So, our approach is ener gy a w are reacti v e protocol which considers the nodes residual ener gy to select path to the destination, by applying this method, nodes can select paths with maximum lifetime, thus achie ving considerable ener gy sa vings. T o attain this objecti v e, we modified the route disco v ery process to select the path that consists of nodes with higher remaining ener gy . In this method, when a RREQ message is transmitted in the netw ork, not e v ery node, which recei v es the message, will dif fuse it. If the residual ener gy of intermediate node is lo wer than a predefined threshold v alue, the RREQ is dropped, otherwise, the message is forw arded Figure 2. S   E =   8 j   E =   1 j   5   E =   1 0 j   4   E =   1 0 j   E =   6 j   D   E =   8 j   3   E =   7 j   Figure 2. Route disco v ery process in ME-MCMI A OD V 3.3. Location corr ection In reality , the location measurement tools do not gi v e the e xact location of the nodes. So, to gi v e credibility to our approach, we ha v e introduced a model that tak es into account the presence of location errors. The measurement error (also called Observ ational Error) is the dif ference between a measured quantity and its true v alue. Indonesian J Elec Eng & Comp Sci, V ol. 22, No. 1, April 2021 : 315 325 Evaluation Warning : The document was created with Spire.PDF for Python.
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752 r 319 In our model, we used the Kalman filter to estimate the internal state location of nodes i n the netw ork. Each node runs a Kalman filter predictor , which is em plo yed to predict the node’ s o wn position (x, y) and v elocity (vx, vy). The position v ector at moment t is measured by: X t = [ x ( t ) ; y ( t ) ; v x ( t ) ; v y ( t )] T (1) T o find the best estimate of the current state in re gular interv als, we apply the time and measur ement update mechanisms of Kalman filter as sho wn in (2) and (7). The steps in v olv ed in state estimation of our system are described as follo ws: T ime update (Prediction) Location Prediction : X 0 t = AX t 1 + W t (2) Error Co v ariance : P 0 t = AP t 1 A T + Q t (3) Measurement update (Correct) Measurement of state : Y t = C X 0 t (4) Kalman Gain : K = P 0 k H . H P 0 t H T + R (5) Update Prediction Measurement : X t = X 0 t + K Y t H X 0 t (6) Update Error Co v ariance : P t = ( I K H ) P 0 t (7) In abo v e equations: A : the state transition matrix; Q : the noise co v ariance matrix; H : the observ ation matrix; R : the noise co v ariance matrix of the observ ation. Y t : the observ ation v ector achie v ed from the current node, namely , the node’ s current position. The figure as sho wn in Figure 3 the general flo w and o v ervie w of our system model. In itial  lo catio n   Pr evio u lo catio n   New   lo catio n   (p r ed icted ,   b ased   o n   p h ysical  m o d el  an d   p r evio u lo catio n )   Resu lt   of    u p d ated   lo catio n   State  an d   co varian ce  u p d ate   Kalm an   gain   Meas u r em en t   from   sen so r   I nit ial   s tat e   be c om e s   pr e v ious   =   S t at e   m at r ix   P   =   P r oc e s s   c ovar ian c e   m at r ix   ( r e p r e s e n t s   e r r or   in   t he  e s t im at e )   I   =   I dent it m at r ix   K   =   K al m an   gain   =   S e n s or   n ois e /m e as u r e m e n t   c ovar ian c e   m at r ix   =   C onve r s ion   m at r ix   ( t m ak e   s iz e s   c ons is t e nt )   Y   =   M e as ur e m e nt   of   s t at e     =  P r e d i c te d   s tate   n oi s e   m atr i x   Q   =   P r oc e s s  n oi s e   c ova r i an c e   m atr i x.  K e e p s  th e   s tate  c ovar i an c e   m atr i x fr om   b e c om i n g too  s m al l  or    goi n to 0.     A C  =  A d ap tati on  m atr i c e s , to c on ve r i n p u s tate   to p r oc e s s  s tate   X 0   P 0   X t   P t   X t - 1   P t - 1   = +   =   = 1 +   =   1 +   = + ( + )   = (  )   Figure 3. Ov ervie w of our system model Mobility-pr ediction and ener gy optimization for multi-c hannel multi-interface ... (Hassan F aouzi) Evaluation Warning : The document was created with Spire.PDF for Python.
320 r ISSN: 2502-4752 3.4. Pr edicting link expiration time W e shall ensure that the select ed route wil l be the stable one by calculati ng its link e xpiration time. Only if this time is greater than t he predefined threshold v alue, then that node will t ak e part in route disco v ery . The link e xpiration time is calc ulated at each hop of the route and taking into consideration the mobility of the nodes in v olv ed in a netw ork. So, the node computes the link duration time, after reception the RREQ, between itself and the sender which implies the predicted lifetime of the link using the real location calculated in the pre vious section as sho wn in (6) instead of the measured location information. Let us assume tw o nodes i and j are within the transmission range r of each other . Let: ( x i , y i ): The coordinate of node i . ( x j , y j ): The coordinate of node j . V i and V j be the speeds of nodes i and j respecti v ely . i and j ( 0 < = i , j < 2 ) be the mo ving directions of nodes i and j respecti v ely . So, the link e xpiration time (length of the longest time interv al during which the tw o nodes are within the transmission range of each other) D t , of the link between the tw o nodes, as defined in [28], is gi v en as sho wn in (8) : D t = ( ab + cd ) + q ( a 2 + c 2 ) r 2 ( ad bc ) 2 a 2 + c 2 (8) Where : a = i cos i j cos j b = X i X j c = i sin i j sin j d = Y i Y j If the Link Expiration T ime v alue is smaller than the Link Duration T ime in the modified RREQ, the recei ving node replaces the Link Duration T ime v alue by the ne w one. In case the recei v er is not the destination of the RREQ, the node broadcasts it to other nodes. 4. EXPERIMENT AL MODELING, AN AL YSIS AND RESUL TS 4.1. Experimental modeling 4.1.1. Simulation model In this paper , the simulation of our proposed protocol (ME-MCMI A OD V), A OD V and A OD V [8] is done by using netw ork simulator (NS-2) softw are due to its a v ailability . NS-2 is a discrete e v ent Simulator written in C++ and O TCL, C++ for data per e v ent pack ets and O TCL are used for periodic and triggered e v ent. NS-2 includes a netw ork animator called Nam Animator , which pro vides visual vi e w of simulation. A WK scripts are used to analyze output of TCL and get the netw ork performance. 4.1.2. Model parameters In these simulations, we used a wireless netw ork, which is a 1km x 1km simulation en vironment. W e emplo yed MA C 802.11 protocol, with node transmission range of 250m.The constant bit rate (CBR) traf fic under the user datagram protocol (UDP) is used to accurately compare dif ferent routing protocols with a send- ing rate of 4 pack ets per second, 512 bytes of pack et size and simulation time of 600s. The random w aypoint mobility (R WP) [29] model is used as a mobility model with randomly selected speed between 1 m/s and 20 m/s. The performances of protocols are e v aluated by v arying both the netw ork size (number of nodes) and the pause time. W e consider 10 random simulation runs and the performance of the considered f actor is the a v erage of these outputs. The parameter settings are listed in T able 1. Indonesian J Elec Eng & Comp Sci, V ol. 22, No. 1, April 2021 : 315 325 Evaluation Warning : The document was created with Spire.PDF for Python.
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752 r 321 T able 1. The parameter settings of our simulation P arameter V alues Netw ork simulator NS-2.35 Simulation area 1 km x 1 km Number of mobile nodes 10, 20, 30, 40, 50, 100, 150 Simulation time (s) 600 Mobility model R andom w ay point P ause time (s) 10, 20, 30, 40, 50, 100, 150 P ack et generation rate 4 pack ets/s P ack et size (bytes) 700, 800, 900, 1000, 1100, 1200 T ransmission range (m) 250 Number of interf aces 2, 3, 4 4.2. Results and discussion 4.3. Choosing the number of interfaces In our model, de vices can communicate by means of multiple interf aces. So, we ha v e to choice the optimal number of interf aces to implement in ph ysical layer . The T able 2 sho ws the results of our simulation study in multi-channel multi-interf ace en vironment. From the table we can notice that be yond tw o interf aces the performance of our protocol decreases. This is due to se v eral f actors such as RF interference, antenna cor - relation. In addition, each antenna in the MIMO system needs a radio-frequenc y (RF) unit, so the battery gets drain f aster due to processing of comple x and computationally intensi v e signal processing algorithms.That’ s wh y we opted to use only tw o interf aces per node in the rest of our contrib ution. T able 2. Performance analysis vs number of interf aces Number of interf ace PDR % End to end delay T otal Dropped P ack ets Lifetime 2 96.9411 338.87 364 105.949859 3 94.19 345.66 387 109.243 4 93.45 376.634 392 111.324 4.3.1. P ack et deli v ery ratio P ack et deli v ery ratio (PDR) is calculated by di viding the number of pack ets recei v ed by the CBR sink at the final destination by the number of pack ets originated by the “appl ication layer” CBR sources. The PDR needs to be high for ef fecti v e performance of routing. Figure 4 sho ws deli v ery rat io of the data pack ets of ME-MCMI A OD V , A OD V and A OD V [3] in terms of v ariation of number of nodes, pause time and pack et size. W e observ ed that the protocols ha v e higher PDR when the nodes mo v e at lo w speeds. When the speed increases, routing protocols suf fer a decrease in PDR. This is normal because higher speeds of nodes mak e routes unstable which leads to an increase of loss data. W e notice also that our proposed protocol sho ws higher v alues re g ardless of v ariation of parameters (node density , pause time and pack et size) as compared to A OD V and A OD V [8], because proposed protocol select the nodes which ha v e suf ficient ener gy and high link duration time. This mechanism reduces the chances of route f ailure especially in case of high mobility (lo w pause time), which result in impro ving the pack et de- li v ery ef ficienc y . 70 75 80 85 90 95 100 10 20 30 50 100 150 Pack et   D el i very  R ati (%)  40 N um ber  of   nodes   A O D V AODV[ 8 ] O ur  a ppr oa ch (a) 65 70 75 80 85 90 95 100 10 20 30 40 50 100 150 P ack et  D el i very  R ati (%)   Pause  ti m (s )   A O D V AODV[ 8 ] O ur a ppr oach (b) 70 75 80 85 90 95 700 800 900 1000 1100 1200 Packet  Delivery  Ratio  (%)  Pack et   s i ze  (Byte )   A O D V A O D V [ 8] O ur  a ppr oa ch (c) Figure 4. P ack et Deli v ery Ratio by changing the: (a) number of nodes, (b) pause time and (c) pack et size Mobility-pr ediction and ener gy optimization for multi-c hannel multi-interface ... (Hassan F aouzi) Evaluation Warning : The document was created with Spire.PDF for Python.
322 r ISSN: 2502-4752 4.3.2. T otal dr opped pack ets The T otal Dropped P ack ets is the number of pack ets that is not recei v ed by the destination. The pack- ets may be lost due to man y f actors such as transmission errors and congestion. This loss may tak e place at both, netw ork and MA C layer . The result of these f actors is related with the host mobility , number of connections, traf fic load and pack et size. From Figure 5, the v ariation in number of mobile nodes, pause time and pack et size depicts that the A OD V and A OD V [8] ha v e more dropped pack ets than ME-MCMI A OD V . The number of nodes in the netw ork, pause time and pack et size will af fect the requirement of route disco v ery between dif ferent pairs in the netw ork. So, it can be seen from this figure that the number of dropped pack ets increases when the pause time decreases and pack et size increases, because higher mobility leads to more brok en links and higher pack et size mak es the chances of loss v ery significant due to collisions and interf ace o v erflo ws. In our approach, the paths of the high residual ener gy and li nk duration time are selected, so the route will not brok e quickly , thus it reduces the number of dropped pack ets. 500 1000 1500 2000 2500 3000 10 20 30 50 100 150 T otal   D ropp ed  P ack ets   40 N um ber  of   nodes  A O D V AODV[ 8 ] O ur  a ppr oa ch (a) 200 400 600 800 100 0 120 0 140 0 160 0 10 20 30 40 50 100 150 Total   D ropped  Pack et s   P ause  ti m (s )   A O D V AODV[ 8 ] O ur  app r oach (b) 350 400 450 500 550 600 650 700 700 800 900 1000 1100 1200 T otal   D ropp ed  P ack ets   Pack et   s i ze  (Byte )   A O D V A O D V [ 8] O ur a ppr oac h (c) Figure 5. T otal Dropped P ack ets by changing the: (a) number of nodes, (b) pause time and (c) pack et size 4.3.3. End to end delay The end to end delay is the ratio of tim e dif ference between numbers of pack et send and recei v ed o v er the total time require to reach the destination. It is a significant parameter for e v aluating a protocol, the more delay is reduced, the performance of netw ork gi v es better output. Figure 6 sho ws end to end delay for number of nodes from 10 to 150, pause time from 10s to 150s and pack et size from 700 to 1200 bytes. From this figure, we notice that this parameter is decreased as the pack et size, density of the netw ork and mobility of nodes increased. Firstly because the probability of success in accessing the medium decreased when a greater number of nodes contend for access to the channel in stable netw ork, b ut also because a lar ger pack et needs more ti me to reach destination than smaller pack ets due to more pack et drops and pack et retransmi ssions are needed. W e observ e also that the a v erage end to end delay of ME-MCMI A OD V is smaller than both A OD V and A OD V [8] in all simulation scenarios (number of nodes, pause time and pack et size). The reason is that our protocol reduce the traf fic load by selecting the stable paths and this reduces queuing and propag ation delays. 35 40 45 50 55 60 65 70 75 10 20 30 50 100 150 E ndtoE nd   de l ay  (m s)   40 N um ber   of   node s   A O D V AODV[ 8 ] O ur  a ppr oa ch (a) 0 50 100 150 200 250 300 350 400 450 500 10 20 30 40 50 100 150 E ndtoE nd   de l ay  (m s)   P ause  ti m (s )   A O D V AODV[ 8 ] our  a pp r oa c h (b) 300 350 400 450 500 550 600 700 800 900 1000 1100 1200 En to  end  del ay (m s)   Pack et   s i ze  ( Byte )   A O D V A O D V [ 8] O ur a ppr oac h (c) Figure 6. End to end delay by changing the: (a) number of nodes, (b) pause time and (c) pack et size 4.3.4. Lifetime From Figure 7, the results sho w that the netw ork lifetime increases as the node density or pause time increases. Because if the number of nodes in the netw ork is too small, feasible routes between sources and destinations may not e xist in this netw ork, so lar ger o v erhead messages need to k eep and disco v er routes, which lead to high consumption of ener gy . F or the third parameter , we observ e that as pack ets size increases, Indonesian J Elec Eng & Comp Sci, V ol. 22, No. 1, April 2021 : 315 325 Evaluation Warning : The document was created with Spire.PDF for Python.
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752 r 323 more ener gy will be required to transmit data pack ets from one end to the other , hence reducing the lifetime of the netw ork nodes and the o v erall lifetime of the netw ork. The results re v eal t hat our modified algorithm outperforms both A OD V and A OD V [8] by achie ving long duration of time for the first node witch e xhausts its ener gy on the netw ork. The impro v ement in netw ork lifetime is due to the f act that ME-MCMI A OD V pre v ents small residual ener gy nodes to be a relay node or selects a path that has long duration time than shorter path between source and destination. 50 100 150 200 250 10 20 30 50 100 150 L ifetime  (s) 40 N um ber  of   nodes   A O D V AODV[ 8 ] O ur a ppr oach (a) 90 110 130 150 170 190 210 230 250 270 290 10 20 30 40 50 100 150 L ifetime  (s) Paus t i m ( s )   A O D V AODV[ 8 ] O ur  a ppr oa ch (b) 55 65 75 85 95 105 700 800 900 1000 1100 1200 L i f eti m e (s)   P ack et  si ze  (Byte)   A O D V A O D V [ 8] O ur a ppr oac h (c) Figure 7. Lifetime by changing the: (a) number of nodes, (b) pause time and (c) pack et size 4.3.5. Reason behind the r esult As mentioned before, the results demonstrate that the ME-MCMI A OD V generates better performance results as compared with a basic A OD V and A OD V [8]. The impro v ed performance of ME-MCMI A OD V compared to the tw o others protocols can be attrib uted to se v eral design f actors. One of the major f actors is the incorporation of the enhanced RREQ mechanism, which lo wers the rate of problems in our approach. In ME-MCMI A OD V , the route created between an y pair of nodes consists only of nodes whose ener gy le v el is higher than the threshold, so our protocol ensures a more stable link, without unnecessary link breakages, and as a result more successful pack et deli v ery to destination nodes. Consequently , ME-MCMI A OD V sends out less number of control pack ets that can reduce the o v erhead and increase lifetime of the netw ork. 5. CONCLUSION The stability of the route and lifetime of the netw ork are considered as challenging tasks in MANET . This paper proposed a multi-channel multi-interf ace on-demand routing algorithm (ME-MCMI A OD V) with a mobility prediction that tak es into account the location errors and residual ener gy of nodes. V ia simulations, our proposed algorithm sho ws significant performance impro v ements in terms o f pack et deli v ery ratio, total dropped pack ets, end-to-end delay and netw ork lifetime compared with other protocols in the field, especially in a netw ork with more connections, high mobility of nodes and lar ge pack et size. Because ME-MCMI A OD V protocol select the nodes which ha v e suf ficient ener gy and high link duration time so the route will not brok e quickly . T aking for granted the benefit of the solution proposed in this paper , in the future w ork we will try to e xpand the solution by proposing a model that will tak e into account other parameters in the process of establishing routes between source and destination nodes such as channel queue length, signal to noise ratio and v arying parameters of simulations (adding some results v ersus the maximum speed in mobility model, generation and rate. REFERENCES [1] A. O. al Janaby , A. Al-Omary , S. Y . Ameen, and H. M. Al-Rizzo, “T racking high-speed users using snr -cqi mapping in lte-a netw orks, 2018 International Conference on Inno v ation and Intelligence for Informatics, Computing, and T echnologies (3ICT) , 2018, pp. 1–7, doi: 10.1109/3ICT .2018.8855771. [2] A. O. Al Janaby , A. Al-Omary , S. Y . Ameen, and H. Al-R izzo, “T racking and controlling high-speed v ehicles via cqi in lte-a systems, International Journal of Computing and Digital Systems , v ol. 9, pp. 1–10, Jul. 2020. [3] C. E. Perkins, E. M. Ro yer , Ad hoc ondemand dis tance v ector (aodv) routing, Proceedings WM- CSA ’99. Second IEEE W orkshop on Mobile Computing Systems and Applica tions , v ol. 3561, 2003, doi: 10.1109/MCSA.1999.749281. [4] D. B. Johnson and D. A. Maltz, “Dynamic source routing in ad hoc wireless netw orks, Mobile com- puting , pp. 153–181, 1996. Mobility-pr ediction and ener gy optimization for multi-c hannel multi-interface ... (Hassan F aouzi) Evaluation Warning : The document was created with Spire.PDF for Python.
324 r ISSN: 2502-4752 [5] C. E. Perkins and P . Bhagw at, “Highly dynamic destinati on-sequenced distance-v ector routing (dsdv) for mo- bile computers, A CM SIGCOM M computer communication re vie w , v ol. 24, no. 4, pp. 234–244, 1994, doi: 10.1145/190314.190336. [6] Z. Haas, “The zone routing protocol (zrp) for ad hoc netw orks, IETF Internet draft, draft-ietf-manet- zone-zrp-01. txt , 1998. [7] V . P ark, “T emporally-ordered routing algorithm (tora) v ersion 1 functional specification, Internet Draft, draft-ietf- manet-tora-spec-04. txt , 2001. [8] T . K. V u and S. Kw on, “Mobility-assisted on-demand routing algorithm for manets in the presence of location errors, The Scientific W orld Journal , v ol. 2014, 2014, doi: 10.1155/2014/790103. [9] H. F aouzi, H. Mouncif, and M. Lamsaadi, Aodv ener gy routing mechanism for multi-channel multi-interf ace ad hoc netw orks (emcmi-aodv) using a dynamic programming algorithm, International Journal of Mobile Computing and Multimedia Communications (IJMCMC) , v ol. 7, no. 4, pp. 1–16, 2016, doi: 10.4018/IJMCMC.2016100101. [10] R. A. Calv o and J. P . Campo, Adding multiple i nterf ace support in ns-2, Uni v ersity of Cantabria , 2007. [11] I. L. Cherif, L. Zitoune, and V . V ` eque, “Throughput and ener gy consumption e v aluation in directional antennas mesh netw orks, 2016 IEEE 12th International Conference on W ireless a nd Mobile Computing , 2016, pp. 1–8, doi: 10.1109/W iMOB.2016.7763211. [12] N. K et and S. Hippar gi, “Modified aodv ener gy a w are routing for optim ized performance in mobile ad-hoc netw orks, 2016 International Conference on W ireless Communications, Signal Processing and Netw orking (W iSPNET) , 2016, pp. 1030–1034, doi: 10.1109/W iSPNET .2016.7566293. [13] H. Ashwini, V . R. KP , and I. Ginima v , “Cm-aodv: an ef ficient usage of netw ork bandwidth in aodv protocol, 2018 In- ternational Conference on Design Inno v ations for 3Cs Compute Communicate Control (ICDI3C) , 2018, pp. 111–114, doi: 10.1109/ICDI3C.2018.00032. [14] A. Ab u-Ein and J. Nader , An enhanced aodv routing protocol for manets, International Journal of Computer Science Issues (IJCSI) , v ol. 11, no. 1, pp. 54, 2014. [15] Z. Zhaoxiao, P . T ingrui, and Z. W enli, “Modified ener gy-a w are aodv routing for ad hoc netw orks, 2009 WRI Global Congress on Intelligent Systems , v ol. 3, pp. 338–342, 2009. [16] A. Sharma, A. Bansal, and V . Rishiw al, “Sbadr: stable and bandwidth a w are dynamic routing protocol for mobile ad hoc netw ork, International Journal of Perv asi v e Computing and Communications , 2020. [17] A. P al, P . Dutta, A. Chakrabarti, J. P . Singh, and S. Sadhu, “Biogeographic-based temporal prediction of link stability in mobile ad hoc netw orks, W ireless Personal Communications , v ol. 104, no. 1, pp. 217–233, 2019. [18] A. M. Bamhdi, “Ef ficient dynamic-po wer aodv routing protocol based on node den- sity , Computer Standards and Interf aces , v ol. 70, pp. 103406, 2020. [Online]. A v ailable: http://www .sciencedirect.com/science/article/pii/S0920548919304453. [19] K. A. Darabkh, M. G. Alf a w ares, and S. Althunibat, “Mdrma: Multi-data rate mobility-a w are aodv-based protocol for flying ad-hoc netw orks, V ehicular Communications , v ol. 18, pp. 100163, 2019, doi: 10.1016/j.v ehcom.2019.100163. [20] H. F aouzi, M. Er -rouidi, H. Moudni, H. Mouncif, and M. Lamsaadi, “Impro ving netw ork lifetime of ad-hoc netw ork using ener gy A OD V (e-aodv) routing protocol in real radio en vironments, International Conference on Netw ork ed Systems , 2017, pp. 27–39. [21] A. K. Dogra, et al ., “Q-A OD V : A flood control ad-hoc on demand distance v ector routing protocol, 2018 first inter - national conference on secure c yber computing and communication (ICSCCC) , 2018, pp. 294–299, doi: 10.1109/IC- SCCC.2018.8703220. [22] P . Rani and G. Bisw as, A OD V enhancement based on the minimization of route-request pack ets, International Conference on Computer Science and Information T echnology , 2012, pp. 442–454. [23] S. R. Mal we, N. T aneja, and G. Bisw as, “Enhancement of DSR and A OD V protocols using link a v ailability predic- tion, W ireless Personal Communications , v ol. 97, no. 3, pp. 4451–4466, 2017, doi: 10.1109/ICSCCC.2018.8703220. [24] S. Ahn, Aodv e xtensions for multipath routing, ie tf internet draft, Uni v ersity of Seoul , No v . 2017. [25] H. Jhajj, R. Datla, and N. W ang, “Design and implementation of an ef ficient multipath A OD V routing algorithm for manets, 2019 IEEE 9th Annual Computing and Communication W orkshop and Conference (CCWC) , 2019, pp. 0527–0531, doi: 10.1109/CCWC.2019.8666607. [26] Y . Mai, F . M. Rodriguez, and N. W ang, “Cc-ado v: An ef fecti v e multiple paths congestion control A OD V , 2018 IEEE 8th Annual Computing and Communication W orkshop and Conference (CCWC) , 2018, pp. 1000–1004, doi: 10.1109/CCWC.2018.8301758. [27] P . K yasa nur and N. H. V aidya, “Routing and interf ace assignment in multi-channel multi-interf ace wire- less netw orks, IEEE W ireless Communications and Netw orking Conference , 2005, pp. 2051–2056, doi: 10.1109/WCNC.2005.1424834. [28] W . Su, S.-J. Lee, and M. Gerla, “Mobility prediction and routing in ad hoc wireless netw orks, International journal of netw ork management , v ol. 11, no. 1, pp. 3–30, Feb . 2001, doi: 10.1002/nem.386, doi: 10.1002/nem.386. [29] E. Hyyti ¨ a, H. K oskinen, P . Lassila, A. Penttinen, J. Roszik, and J. V irtamo, “Random w aypoint model in wireless netw orks, Netw orks and algorithms: Comple xity in ph ysics and computer science , v ol. 590, Jan. 2005. Indonesian J Elec Eng & Comp Sci, V ol. 22, No. 1, April 2021 : 315 325 Evaluation Warning : The document was created with Spire.PDF for Python.