Implementation

In this section, we are dealing with the implementation issues of the scheme described in this chapter. The main concern is the design of a proper mech-anism that ensures the selection of the next hop as described by our routing strategies.

The first assumption stated in this chapter is the ability of the transmit-ting nodes to discover their routes or their neighborhood. This translates in practice into a routing protocol that relies on geographical locations of nodes in order to compute routes towards destination. For example, this a-priori information regarding neighbor and destination coordinates is used in geographical routing protocols as described in [63, 65]. A simple and effi-cient way to obtain these geographical locations is to assume that all nodes are equipped with Global Positioning System (GPS) receivers and capabili-ties. This GPS-based solution is realistic nowadays and easy to implement.

One can argue that other location methods can be used for ad hoc wire-less networks such as signal strength, angle of arrival and can be suitable for determining local information (i.e., information in the vicinity of the node). However all these methods would be expensive in terms of compu-tations and bandwidth (exchange information overhead). Thus, GPS based location determination method can be an important parameter in reducing information overhead, thus simplifying the distribution of information and limiting infrastructure reliance. Moreover, in case not all the nodes have a GPS receiver, one can design a procedure so that the subset of nodes

4.4. Throughput expressions 103

DATA

Slot Time WT = Wait Time

Rx1 Burst

Rx2 Burst

... ...

Figure 4.6: Slot format for the case where the transmitter selects the next hop.

with GPS capability succeed in supporting the maximum possible number of nodes without GPS capability. As a result local determination is fully enabled.

Once the neighbor and destination coordinates are available at the trans-mitter, the latter computes the next hop (or the next relay) towards the destination. The computation of the next hop could be either determined by the transmitter (as described in Fig.4.6) or by the receiver at the next hop (as described in Fig.4.7). For the first solution, the transmitter know-ing the direction of the final destination and the coordinates of nodes in its vicinity, computes the forward progress (for all strategies described in this chapter) and determines the best node to be the next hop. However this computation is based on the knowledge of the nodes locations, their oper-ating mode (MAC mode, either transmitter or receiver) and an information on the node density. Indeed the transmitter selects the best hop among the receivers. If the station elected as the next hop happens to be in transmit mode, a collision occurs. To cope with this problem, we can include in the spatial throughput derivations this event as a collision with certain proba-bility. Another approach described in Fig.4.6 relies on the signal bursts sent by the receiver nodes. A portion of the slot time is dedicated to these signal bursts. Just before data transmission in the next slot, the nodes that will operate in the receive mode in the next slot will send some burst signals or

DATA

CT[i]

ACK Slot Time

CT = Receiver Contention Time

CT[n]

CT[k]

Figure 4.7: Slot format for the case where the receiver elects itself as the next hop.

pilots in a broadcast manner. Thus all the transmit nodes in the vicinity will update their receive nodes table in order to keep track of all the poten-tial relay nodes in the next slot. These signal bursts are sent just after the ACK/NACK signaling while all the potential transmitters are listening. In order to avoid collisions between different nodes burst, one can imagine that each receive node has a position within the time allocated for these signals in the slot format. This is similar to a PPM modulation, making all these signals orthogonal at the transmitting nodes, and any other physical layer solution (e.g. OFDM) is also suitable. This scheme has other practical ad-vantages. It ensures a simple and efficient way to estimate the node density at the transmitting nodes, making the computation of the forward progress easy. These received burst signals (or pilots) allow also the computation of a spatial empirical average of the SINR’s at the transmitting nodes. This is crucial for the routing strategy (RS3).

The second solution, described in Fig.4.7, is a receiver oriented approach.

The receiver that realizes the optimal forward progress (depending on the routing strategy used) elects itself as the next hop or the relay for the current

4.4. Throughput expressions 105 transmission. The issue raised by this solution is how a potential relay notifies the other receivers about the forward progress realized for the current transmission. This is important since it prevents collisions between all the ACK signaling from all the receivers (potential relays). The idea one can use to solve this problem is inspired from backoff algorithms and contention windows used in 802.11 protocols. As depicted in Fig.4.7, receiver[i] has to wait for a duration of time CT[i] before sending the ACK signaling.

This duration is closely linked (for example inversely proportional) to the forward progress realized by the corresponding receiver. Thus the receiver corresponding to the maximal forward progress will be first in accessing the channel and transmitting the ACK signaling. This prevents all the nodes listening to this ACK signaling from sending ACK signals to the corresponding transmitter node and stops the election process of the next hop node.

A more detailed study of such schemes is needed and the impact of these signaling schemes on the overhead and the throughput must be taken into account in a more rigorous manner. This is beyond the scope of this work.