Multipoint relaying

The concept of multipoint relaying is to reduce the number of duplicate retransmissions while forwarding a broadcast packet. This technique restricts the set of nodes retransmitting a packet from all nodes, to a subset of all nodes. The size of this subset depends on the topology of the network.

Figure: Flooding a packet in a wireless multihop network. The arrows show the way information is passed, not all transmissions.
\includegraphics[width=2.5in]{gfx/flood_one_way.eps}
Figure: Flooding a packet in a wireless multihop network from the center node using MPRs(black). The arrows show the way information is passed, not all transmissions.
\includegraphics[width=2.5in]{gfx/mpr_flood_one_way.eps}

This is achieved by selecting neighbors as Multipoint relays(MPRs). Every node calculates its own set of MPRs as a subset of its symmetric neighbor nodes chosen so that all 2 hop neighbors can be reached through a MPR. This means that for every node $n$ in the network that can be reached from the local node by at minimum two symmetric hops, there must exist a MPR $m$ so that $n$ has a symmetric link to $m$ and $m$ is a symmetric neighbor of the local node. In the scenario illustrated in figure 3.6, node A selects the black nodes as MPRs. This way all two hop nodes can be reached through a MPR. Node B will not retransmit traffic from A that is to be flooded.

Figure: Node A has selected the black nodes ar its MPRs.
\includegraphics[width=2.5in]{gfx/mpr_selection.eps}

OLSR lets nodes announce their own willingness to act as MPRs for neighbors. 8 levels of willingness are defined from the lowest WILL_NEVER(0), which indicates that this node must never be chosen as a MPR, to the highest WILL_ALWAYS(7), which indicates that this node should always be chosen as a MPR. The willingness is spread through HELLO messages and this information must be considered when calculating MPRs.

Finding the optimal MPR set has been proved to be a NP-complete problem in [62]. RFC 3626 proposes a rather simple heuristic for MPR calculation. The MPR scheme has been further studied and analyzed in amongst others, [52], [15] and [51]. In this thesis the MPR technique is not further analyzed.

Andreas 2004-07-29