The impact of the timeliness of information on the performance of multihop best-select Stephan Bohacek Rick Blum Len Cimini University of Delaware Lehigh University University of Delaware bohacek@udel.edu rblum@ece.lehigh.edu cimini@ece.udel.edu Larry Greenstein Alex Haimovich Rutgers University New Jersey Institute of Technology ljg@winmain.rutgers.edu alexander.m.haimovich@njit.edu Abstract— Cooperative relaying enables nodes to actively cooperate to deliver packets to their destination. This cooper- ation allows nodes to take advantage of the diversity provided by variations in the channel gains between nodes. Best-select, a particular type of cooperation, has been shown to result in significant gains in the performance of source-to-destination communication. However, this increase in performance is achieved by exchanging channel gain measurements, which requires overhead. One way to reduce this overhead is to exchange channel gain measurements less frequently. This paper examines the trade-off between performance and the frequency of exchanging channel gains. This investigation focuses only on the channels that are impaired by multipath fading and shadow fading. I. I NTRODUCTION In traditional multihop wireless data networks, route search and packet forwarding are separated; first a route is found, and then packets are forwarded along the route. In the case that multipath routing is employed, the situation is similar, but a set of paths are found, and packets are forwarded along each route either probabilistically, or the routes are used as precomputed backups [1], [2]. In any case, in traditional routing, nodes act alone to forward the packet to its next hop. In cooperative relaying, a group of nodes act together to forward packets. While several variants of cooperative relaying are possible, one approach is to generalize the single node that forwards the packet to a set of nodes that cooperate (see [3] for an alternative approach). Such a set of nodes is called a relay-set. Thus, while traditional networking forwards packets from node to node, this form of cooperative relaying forwards packets from relay-set to relay-set. Within the relay-set paradigm, there are also many possible approaches. For example, in some cases, a number of nodes transmit the same or different parts of the packet. In such cases, the total transmission power used to transmit the data packet between two relay-sets is distributed among a number of node pairs [4], [5]. However, if the channels are known, then a simple and good performing approach is to allocate all the power to the node that has the best channel to destination [6]. Such an approach is known as best-select relaying. Best-select protocol (BSP) is an implementation and a multihop extension of best-select relaying [7]. BSP makes active use of channel measurements to select the best end- to-end path. A distinguishing feature of BSP is that it is highly dynamic; channel measurements are repeatedly made and the path that packets take varies as the channel varies. In this way, BSP is able to take advantage of diversity and achieves dramatic improvement over traditional least- hop routing that finds a single path and utilizes the path until it breaks. In [8], it was found that BSP provides large performance improvements over least-hop routing. For example, it was found that BSP has the potential to find paths that provide 5-10 times higher throughput, impose 5- 10 times less delay, utilize 10-1000 times less power, and consume 10-100 times less energy than traditional least-hop routing. As mentioned, best-select relies on the dissemination of channel measurements. Specifically, the version of best- select considered in [8] exchanges channel information between adjacent relay-sets every time a packet is delivered. On the one hand, this exchange of channel information re- sults in overhead and detracts from the overall performance. While on the other hand, it is the exchange of channel information that allows the best path to be selected and leads to the improvements in performance. This paper inves- tigates the relationship between performance and the rates at which channel information is exchanged. Several schemes to reduce overhead are considered. Specifically, while BSP exchanges channel information with every packet delivery, here the impact of exchanging channel information less regularly is investigated. Also, as explained in the next section, the direction of the exchanges (downstream or upstream) can be changed. The impact of the direction of the information exchange is investigated. Also, the impact of only "good" nodes exchanging channel information and