AHNs imply a completely new approach to networking design and bring the architectures of nodes and transmission protocols very close to the specific field of application of each considered case study :

• scenario parameters such as the nodes’ density
• their different resources
• their relative positions in the considered environment
• the extension of the operating area
• the type of established communications (e.g., one-to-one, one-to-many, etc.), and
• the specific types of traffic (e.g., real-time, best-effort, etc.)

Just to list a limited number of items, pose totally different requirements and may appreciably influence the choice of design. From this perspective, it appears even clearer that applications are very different from each other: the particular choice in the first context might not be as good in another.

Keeping in mind this variety of scenarios and flexibility of use, a number of characteristics of such networks are usually considered in the design process. Obviously, issues related to the radio channel are important. Several solutions foresee the use of the Industrial, Scientific, and Medical (ISM) 9 frequency band.

On the one hand, this is advantageous because of free license usage and a range of frequencies common to many countries. On the other hand, the ISM band requires coexistence with many other standards (such as Bluetooth 10 or WaveLAN 11 ), already allocated to the same radio bands, or even with other generic electronic appliances (such as cordless phones or microwave ovens), which are potential sources of interference.

Moreover, transmissions will have to comply with precise and strict transmission rules (e.g., low maximum output power level and specific spread spectrum modulation schemes). In some cases, to effectively exploit the available frequency range, the entire bandwidth is sliced into small parts by creating a number of nonoverlapping subchannels whose management necessarily also involves the MAC sublayer.

Also, the adopted transmission power levels assume great importance in AHNs, especially in the case of high-density and battery-powered nodes. Minimal transmission power, together with the implementation of smart strategies for energy saving, may considerably increase node life duration and channel utilization.

However, using less transmission power may involve longer multi-hop paths with a major impact on routing protocols and overall topology management. Additionally, the use of variable power levels may give rise to unidirectional links; that is, a pair of nodes cannot communicate with each other because of the mismatching of covered radio range.

The handling of such cases is cumbersome, and the MAC sublayer usually acts like a filter so that upper-layer protocols operate in bidirectional-link conditions.

Even when all nodes transmit at maximum power, particular topology constraints may require multi-hop transmissions to guarantee full network coverage. Considering also that no centralized control system is provided, new challenges for wireless AH MAC protocols, with respect to classical wired cases, are then posed.

Moreover, effective usage of the medium requires that the MAC implement certain quality of service (QoS) levels whenever they are imposed by applications. The creation of subsets for nodes — such as those identified by network organization algorithms — can be exploited to implement alternative MAC coordination schemes by enabling some types of centralized transmission policies.

As stated before, node mobility is just another key feature of AHNs. It obviously impacts on their topology, often in a very unpredictable way, in space and in time: the setting up or the dropping of a link, an occasional event in wired as well as in single-hop wireless networks, becomes very frequent, and routing protocols have to react to these events and manage them promptly, implementing fast tracking of changing topologies and quick recalculations of the available paths.

The optimized route calculation should be based on a plurality of parameters of the network; for instance, an algorithm could discard links showing a received power level under a specified threshold. A too-low level is considered the sign of a very weak link, prone to too many errors or even to termination.

In a different manner, an algorithm could evaluate the relative mobility among the nodes and require that only those hosts showing a low relative mobility will act as relayers, aiming at potentially increasing the stability of routes.

Alternatively, an algorithm can exploit the presence of a predetermined virtual infrastructure, assigning to it crucial routing functions such as route discovery ; finally, an algorithm could determine paths that for certain reasons can guarantee, more than other paths, certain degrees of QoS.

Generally speaking, any routing protocol has to guarantee the robustness of a path over time. This may be achieved by predicting and preallocating alternative paths to be used whenever the first choice fails ( backup paths ) or quickly recalculating paths on the basis of local information ( local recovery procedures ).

In any case, the number and the repetition rate of control data exchanges has to be minimized to preserve bandwidth for data and to minimize energy consumption. When considering transport layer issues, the high degree of distributed intelligence naturally present in an AHN should be taken into account.

Many features, in fact, are usually provided by the system already, and an accurate analysis should concentrate on the provisioning of such complementary functions as are strictly necessary to manage the application’s traffic flow.

It would be possible, in this way, to avoid the replication of any useless functionality. A “tiny” transport protocol, for instance, could implement only the flow control feature and leave to the underlying network mechanisms the execution of congestion control and the notification of transmission errors.

Many MAC protocols, for instance, already provide acknowledgment frames. Security can pose several problems. Usually the Admission Control procedures are implemented by a central unit and are not provided in AHNs.

In principle, this is a weakness that becomes even more noticeable when looking at all nodes actively participating in overall operations. That is why in AHNs there is the need to take a comprehensive approach, protecting a system already at very low networking levels.

Also, considering both scenario and application characteristics, some precautions are required even at the MAC sublayer to keep a malicious node from receiving or transmitting sensitive data. But this is only one facet of the problem because a malicious node could even cause impaired operations, for instance, by communicating false control data to mix up routing schemes, with a major, even global, impact.

For such reasons, communications, and consequently participation in networking activity, should be restricted to authorized nodes only. A continuous monitoring of each element is required in this regard with suitable trusting criteria to allow rapid isolation of all items exhibiting illicit behavior.

A final consideration should be scalability. An AHN may spread and grow over a territory with different levels of density: it follows a clear requirement for distributed algorithms. For reasons of simplicity, such algorithms should use local control data processing and involve the smallest part of neighboring nodes.