Optimal Configuration of Network Coding in Ad Hoc Networks
Analyze the impact of network coding (NC) configuration on the performance of ad hoc networks with the consideration of two significant factors, namely, the throughput loss and the decoding loss, which are jointly treated as the overhead of NC. In particular, physical-layer NC and random linear NC are adopted in static and mobile ad hoc networks (MANETs), respectively. Furthermore, we characterize the good put and delay/good put tradeoff in static networks, which are also analyzed in MANETs for different mobility models (i.e., the random independent and identically distributed mobility model and the random walk model) and transmission schemes.
Network coding was initially designed as a kind of Source coding. Further studies showed that the Capacity of wired networks can be improved by network coding (NC), which can fully utilize the network resources.
Due to This advantage, how to employ NC in wireless ad hoc networks has been intensively studied in recent years with the Purpose of improving the throughput and delay performance. The main difference between wired networks and wireless Networks is that there is non ignorable interference between Nodes in wireless networks.
Therefore, it is important to design the NC in wireless ad hoc networks with interference to achieve the improvement on system performance such as good put and delay/good put tradeoff.
The probability that the random linear NC was valid for a multicast connection problem on an arbitrary network with independent sources was at least (1 − d/q)η, where η was the number of links with associated random coefficients, d was the number of receivers, and q was the size of Galois field Fq.
It was obvious that a large q was required to guarantee that the system with RLNC was valid. When considering the given two factors, the traditional definition of throughput in ad hoc networks is no longer appropriate since it does not consider the bits of NC coefficients and the linearly correlated packets that do not carry any valuable data. Instead, the good put and the delay/good put tradeoff are investigated in this paper, which only take into account the successfully decoded data.
Moreover, if we treat the data size of each packet, the generation size (the number of packets that are combined by NC as a group), and the NC coefficient Galois field as the configuration of NC, it is necessary to find the scaling laws of the optimal configuration for a given network model and transmission scheme.
- Throughput loss.
- The decoding loss.
- Time delay.
Proposed system with the basic idea of NC and the scaling laws of throughput loss and decoding loss. Furthermore, some useful concepts and parameters are listed. Finally, we give the definitions of some network performance metrics.
Physical layer Network Coding designed based on the channel state information (CSI) and network topology. The PNC is appropriate for the static networks since the CSI and network topology are preknown in the static case.
There are G nodes in one cell, and node i (i = 1, 2, . . . , G) holds packet xi. All of the G packets are independent, and they belong to the same unicast session. The packets are transmitted to a node i’ in the next cell simultaneously. gii’ is a complex number that represents the CSI between i and i’ in the frequency domain.
- System minimizes data loss.
- System reduces time delay.
The networks that consist of n randomly and evenly distributed static nodes in a unit square area. These nodes are randomly grouped into S–D pairs.
The protocol model, which is a simplified version of the physical model since it ignores the long-distance interference and transmission. Moreover, it is indicated in that the physical model can be treated as the protocol model on scaling laws when the transmission is allowed if the signal-to-interference-plus-noise ratio is larger than a given threshold.
Transmission Schemes for Mobile Networks:
When the relay receives a new packet, it combines the packet it has with that it receives by randomly selected coefficients and then generates a new packet. Simultaneous transmission in one cell is not allowed since it is hard for the receiver to obtain multiple CSI from different transmitters at the same time. Hence, we employ the random linear NC for mobile models.
Analyzed the NC configuration in both static and mobile ad hoc networks to optimize the delay good put tradeoff and the good put with the consideration of the
Through put loss and decoding loss of NC. These results reveal the impact of network scale on the NC system, which has not been studied in previous works. Moreover, we also compared the performance with the corresponding networks without NC.