Compared to single-hop networks such as WiFi, multihop infrastructure wireless mesh networks (WMNs) can potentially embrace the broadcast benefits of a wireless medium in a more flexible manner. Rather than being point-to-point, links in the WMNs may originate from a single node and reach more than one other node. Nodes located farther than a one-hop distance and overhearing such transmissions may opportunistically help relay packets for previous hops.
This phenomenon is called opportunistic overhearing/ listening. With multiple radios, a node can also improve its capacity by transmitting over multiple radios simultaneously using orthogonal channels. Capitalizing on these potential advantages requires effective routing and efficient mapping of channels to radios (channel assignment (CA)). While efficient channel assignment can greatly reduce interference from nearby transmitters, effective routing can potentially relieve congestion on paths to the infrastructure.
Routing requires that only packets pertaining to a particular connection be routed on a predetermined route. Random network coding (RNC) breaks this constraint by allowing nodes to randomly mix packets overheard so far before forwarding. A relay node thus only needs to know how many packets, and not which packets, it should send. We mathematically formulate the joint problem of random network coding, channel assignment, and broadcast link scheduling, taking into account opportunistic overhearing, the interference constraints, the coding constraints, the number of orthogonal channels, the number of radios per node, and fairness among unicast connections.
We develop a suboptimal, auction-based solution for overall network throughput optimization. Performance evaluation results show that our algorithm can effectively exploit multiple radios and channels and can cope with fairness issues arising from auctions. Our algorithm also shows promising gains over traditional routing solutions in which various channel assignment strategies are used. Channel Assignment for Throughput Optimization in Multichannel Multiradio Wireless Mesh Networks Using
The notion of network coding was first introduced and shown to be a promising strategy for improving network throughputs for multicast. Instead of just replication and forwarding, network coding allows intermediate nodes to algebraically combine packets before forwarding them to next hop neighbors established that such combining can be linear in order to achieve the maximum multicast capacity of a given network. The subsequently showed that random coefficients, rather than the deterministic ones, can also be used to achieve the same capacity. By doing such random combining, generally speaking, it does not matter what is received or lost at a destination, but it only matters that enough is received.
We can think of routing as being a special case of network coding, where for each transmission there is only one packet to combine and coefficients for such combining are all ones. Although most promising results have been presented for multicast in the wired domain, the broadcast nature of a wireless medium turns out to be very useful for extracting the benefits of network coding for unicast as well. In general, a single wireless transmission is often received by more than one node. Nodes located farther than a one-hop distance may overhear transmissions and help relay packets for previous hops.
The explicit modeling of opportunistic listening may increase the computational complexity of the formulated optimization problem; we ask if the throughput gains are worth the computational efforts when compared with a traditional routing approach. The answer is positive. Also, it turns out that our resulting optimization framework is general enough to embrace multipath routing as a special case. We only deal with the distributed version of intraflow random network coding (RNC). The choice of interflow coding2 is desirable here because it tends to incur fixed and less overhead than interflow coding. For interflow network coding, packets are coded and decoded hop by hop.
Each wireless mesh router with access point functionality serves as an ingress or egress for the aggregate traffic associated with the mobile/wireless clients in its coverage area. Such traffic is routed to and from the wired infrastructure via multiple wireless hops formed by the wireless mesh routers some of which also function as gateways to the wired infrastructure. Each wireless router may be equipped with multiple wireless interfaces (radios) each of which operates on an orthogonal channel.
Each node in our system can algebraically combine incoming packets according to the random linear network coding (RNC) scheme before forwarding the resulting combined packets to other nodes via its broadcast link. We assume that wireless nodes are in promiscuous mode and that all wireless transmissions are in broadcast mode. Those wireless nodes that hear such transmissions may engage in packet forwarding. It is also assumed that our system operates synchronously in a time-slotted model