In DSR, each retransmission RREQ adds the address of the retransmission node to the RREQ packet. Additionally, we assume that the sender's location is embedded in the GRREQ packet. Therefore, the routing cost of the proposed GAODV is very robust to the change in node density.
InTi, this event can be considered a passive acknowledgment of GREQ(Ti, S, D) because GREQ(Ti+1,S, D) is the retransmission of GREQ(Ti, S, D) induced by GRREQ(Ti− 1, S, D). The instantaneous property of the link can be explained by Eq (3), but it is not straightforward to analyze the end-to-end link using Eq (3) due to the randomness of the GRREQ travel path.
Uncertainty of the destination’s location
If we approximate AODV as GAODV with r= 0, which has 0.3910 at distance R, the required density of AODV nodes for connectivity of 0.9 and 0.99 can be around 5 and 9, respectively. Note that Figure 5 becomes exact connectivity if the number jumps open path two. As shown in Figure 6, if the location of Dis in SRR(T) is known, GRREQ propagation stops at T.
In this case, if D moves to a point within O(D,γ)-O(T,R), the GAODV cannot find the route fromStoD. However, if we use D1 instead of D, for example, a node of SRR(T) broadcasts GREQ(T, S, D1) because D1 is out of O(T,R), and this additional retransmission can provide an extra coverage . Since the real destination naΔt can be located at any point within O(D,γ), we propose to artificially set another destination, referred to as the imaginary destination, and make the GAODV routing the imaginary destination, not D. The goal is twofold: 1) to make the real destination listen to the GREQ being redirected within O(D,γ) while being routed to the imaginary destination, and 2) to shorten the routing path to the imaginary destination.
It is immediately apparent from Figure 6 that a large part of the area within O(D,γ) can be covered if the GRREQ routing propagates along the diameter line of O(D,γ). In addition, to shorten the route path, we propose to choose as an imaginary destination the intersection point on the other side of O(D) and C(S, D), for example D2 in Figure 6. Note that D1 and D3 in Figure 6 are not good options for the imaginary destination, because a large part of the area within O(D,γ) cannot be covered in both cases.
GAODV is implemented by modifying the RREQ-related AODV functions in the QualNet version.
Message format of GRREQ
Even if we choose small enough for high connectivity, path discovery may still fail due to incomplete connectivity. The size of the SRRD field must be large enough to accurately fit r, which means increasing the size of the GRREQ packet. The RREQ AODV reserved field is large enough to accommodate SRRD and SF.
The size of LoS and LoD is dependent on the required precision of location. Most of the time, 32-bit quantization is sufficient because GAODV does not require exact placement. For example, if we assign 6/6/4-bit integers to the last digit of degrees, minutes, and seconds, respectively, this quantization of the GPS coordinates can cover approximately 1,000 km×1,000 km, where the quantization error between two points is less than 100 m.
The marine VHF communication range is usually several tens of kilometers, and the quantization error normalized by the communication range is less than 1%. Therefore, 32-bit quantization for LoS and LoD is sufficient for marine VHF communication networks. The RREQ packet size of AODV is 192-bit in RFC 3561, and thus, the overhead of GRREQ due to the newly added fields is about 1/3 that of AODV.
Origination of the GRREQ at the source
If the GRREQ is retransmitted despite a successful retransmission by the node in the SRR(S), e.g. if the passive acknowledgment is not recognized due to a temporary poor state of the wireless channel, the RCC discards this ReTx at one -hop of the source neighbor. Therefore, the impact of ReTx is limited to one-step neighbors of the source. ReTx reduces the path discovery error (RDF) due to temporary degradation of the physical communication channel, such as fading.
After waiting for WaitRREP, if RREP is not received and SF is TRUE, GRREQ is restarted up to MaxNrSF-times per. If route discovery fails until SRRD is greater than MaxSRRD, GAODV gives up the selective retransmission mechanism by setting SF to FALSE and returns to conventional AODV. This fallback mechanism is needed to deal with unicast-like operation failure.
For example, the destination may be outside the coverage of the imaginary destination method.
Parsing the received GRREQ
Example for the GRREQ propagation in a unicast manner
Ifris equal to zero (SF is FALSE), L1 and L2 transmit GRREQ(L1, S, D) and GRREQ(L2, S, D) respectively, but L3, L4 and L5 do not retransmit GRREQ(S, S, D) despite disabled SF, in which we can see the avoidance of GRREQ backpropagation. Since the DISCARD Flag ofL1andL2 is TRUE from the second event, L1andL2 immediately discard this duplicate GREQ. Group M rejects GRREQ(E1, S, D) because group M is outside SRR(E1), which is similar to the second event.
Therefore, the PAC at the source is satisfied by GRREQ(E1, S, D), and the FlagPACK, which is the passive acknowledgment at the sourceS, is set to TRUE. Similar to the third event, each player inserts GRREQ(Fi, S, D) into the queue and stores the IPD with respect to GRREQ(E1, S, D). Since group M's FlagDISCARD is TRUE as of the 7th event, group M immediately discards this double GRREQ.
E1receives GRREQ(F1, S, D), which is considered a duplicate of GRREQ(S, S, D). E1 can parse GRREQ(F1, S, D) because the FlagDISCARD of E1 is FALSE. The FlagPACK, which is the passive acknowledgment at the intermediate node E1, is set to TRUE. To demonstrate the performance of the proposed GAODV, we compare the GAODV with the original AODV through QualNet computer simulations.
Through this comparison, GAODV can also be indirectly compared with the other existing methods by examining their abilities to reduce the overhead of AODV.
Connectivity
The horizontal axis is the distance between the source and the destination normalized by R, and the range of this normalized distance is from 1.5 to 5. In this case, the required node density of GAODV by 0.4 dher= 0.6 is greater than that of AODV by 2.7 and 5 times, respectively. It is clear that the required node density of GAODV decreases as less decreases, but the decrease leads to an increase in the number of hops as shown in Fig 11.
It can be observed in Fig 11 that the hop count of the GAODV is not dependent on the node density, which shows that the average hop distance is determined by r. The hop count of the AODV with ρ= 7 is greater than that of the GAODV because the amount of zigzag is large due to the low node density. Note that the hop count of the AODV converges to a step function as the node density goes to infinity.
In this table, ρreq,0.9 and ρreq,0.99 are the required node density to achieve the connectivity of 0.9 and 0.99, respectively. The GRREQ propagation metr= 0.4 andr= 0.6 is completed in two hops at a distance of 1.3RandEq (3) represents the analytical connectivity at this distance, which explains why we choose the distance of 1.3R. It can also be seen that the change of ρreq,0.9 and ρreq,0.99 obtained by the simulation is less than 3%, despite almost a doubling due to the decrease from 0.6 to 0.4.
Imaginary destination
CR is defined by the fraction of the area O(D,γ) covered by GRREQ retransmissions. To avoid the GRREQ propagation break due to imperfect connectivity, ρ is set to 100. For each γ, CR increases as d(D, DI) increases from 0 to γ, but CR decreases as d(D, DI ) greater than γ .
The imaginary destination of the GAODV ID is set as the optimal location discussed in section 2.5. The CR of the GAODV-ID decreases as d(S, D) increases, but the CR degradation of the GAODV-ID is relatively small. If we recall Figure 6, which illustrates the above scenario, we can see why the CR of the GAODV-WID decreases significantly at d(S, D) of 1.6R.
The CR fluctuation of GAODV-WID decreases with increasing d(S, D) due to the averaging effect of the randomly selected retransmission node. The CR gap between GAODV-ID and GAODV-WID increases with increasing γ. Coverage ratio for change in distance from origin to destination: radius of mobility R/2 and R.
Coverage ratio for changing the distance from the source to the destination: Mobility radius of 3R/2 and 2R.
Throughput, latency, and routing overhead
In these figures, the solid and dotted lines are the GAODV and AODV plots, respectively. It can be seen in Fig 17 that the overall routing cost of GAODV is smaller than that of AODV by an order of two or more. If NCBR= 5 andvmax= 5m/s, i.e., a low traffic scenario, the PDR of GAODV and AODV is greater than 90% and the PDR difference between GAODV and AODV is less than 6%.
If the NCBR is increased to 25, the GAODV PDR is still greater than 90%, but the AODV PDR is reduced to 60%. If vmax increases, the number of path errors also increases due to the increased rate of topology change. The PDR degradation of AODV due to increased mobility is greater than that of GAODV for allNCBR in Figure 18 .
Nevertheless, the delay of the AODV increases to five times that of the GAODV as the traffic becomes heavier. We can conclude that the GAODV is less sensitive to changes in mobility and traffic than the AODV. Figures 22–24 show the PDR, delay, and routing overhead for the packet interval change, respectively.
On the other hand, the PDR and routing surface of GAODV are slightly improved by reducing the packet interval in Figs22 and 24 because the number of path errors is almost unchanged regardless of the packet interval reduction. We have proposed GAODV that discovers a route in a unicast fashion using the RREQ sender and destination locations. GAODV is also applicable to fully mobile scenarios with the help of the proposed imaginary destination method.
