Section: New Results
New services and protocols
The user of a mobile network very quickly experience problems with quality of service: links fade, connectivity disrupts, delays accumulate.
In a wireless network, the set of neighbors which with one node can communicate depends on transmission range, and numerous factors, and in addition the transmission range is often lower than the interference range (the range within which a node prevents correct transmissions of other nodes). Thus bandwidth reservation, a crucial step of quality of service, is an important and difficult problem.
The services and protocols that need careful adaptation are
The connectivity continuity is the most important problem. Trivial in the wired world where a link failure is a rare event, it becomes problematic in the mobile world where link failure caused by mobility are frequent and normal. The first experiments of mobile ad hoc networks with regular internet protocols miserably failed simply because either the protocol was to slow to recover link failure, or when tuned appropriately was generating such a huge overhead that the network collapsed under its own weight. A new generation of routing protocols has arised that allow a suitable control of connectivity in mobile networks. Among them the Optimized Link State Routing combines the optmization of overhead for mobile networks and the full internet legacy. It naturally provides path redundancy which accelerate link failure recovery.
The most important lesson that must be retained is that most of these optimization become NP complete, which is a significant complication compared to their counterpart in the classical wired world. The reason for the NP-completeness is two-sided: on one side the co-interferences make impossible an optimization link by link, on the other side, the large dispersion of performance measurement makes simple heuristic ineffective. As an example, routing with respect to shortest delay average does not guarantee smallest probability of high delay.
Since the bandwidth is scarce, any multimedia application such as video streaming is resource demanding. For example a TV broadcast that uses a mesh network will rapidly exhaust the bandwidth if all connections are point to point. In this case multicast protocols that allows to gather all these point to point connections in a single flow is a need.
There are two classes of multicast protocols: the tree based protocols and the network coding protocols. In the first class the protocols take advantage of the relatively small size of the recipient node set. One can show equivalent results of Gupta and Kumar scaling properties but in the multicast plan when the ratio of recipient versus network size is a fundamental parameter. When this ratio tends to one the performance naturally worsen.
When the recipient set is the whole network, one can apply the network coding scheme with random packet combination. In network coding the packets are no longer isolated: relay nodes makes linear combination of packets and transmitted mixed packets. In theory the performance of network coding is better than isolated packet multicast. In practice network coding is simpler to operate does not need topology management such as spanning trees or Connectedc Dominating Set. The reason for this is highly non intuitive, as if packet superposition was acting like state superposition in quantum mechanic, leading to non expected results.
Quality of service has become the central requirement that users expect from a network. High throughput, service continuity are critical issue for multimedia application over the wireless internet where the bandwidth is more scarce than in the wired world. A significant issue in the ad-hoc domain is that of the integrity of the network itself. Routing protocols allow, according to their specifications, any node to participate in the network - the assumption being that all nodes are behaving well and welcome. If that assumption fails - then the network may be subject to malicious nodes, and the integrity of the network fails. An important security service over mobile networks is to ensure that the integrity of the network is preserved even when attacks are launched against the integrity of the network.
Optimized Link State Routing (OLSR)
The routing protocol OLSR is universally known in the mobile wireless community (more than 475,000 hits on Google). It has numerous implementations and is used in many wireless networks. It is a proactive protocol with full internet legacy which is based on partial topology information exchange, that non the less provide optimal path with additive metrics (such as BGP/OSPF). It is an experimental RFC within IETF and soon will become a full standard under the name OLSRv2.
Bandwidth reservation in mobile ad hoc networks
We have shown that the search of a good path for a new connection that does not destroy the quality of service of existing connections is an NP-hard problem. The result is independent on how the bandwidth nodes interfer as long they interfer at least on one hop. In this area, one contribution was the definition and testing of an efficient reservation algorithm bandwidth reservation, respecting wireless network constraints. A second contribution is more accurate computation of remaining link bandwidth by considering bandwidth on other links multiplied by the average packet retransmission on this link (inverse of packet successful transmission rate).
We have also proposed a solution called QoS-OLSR that enhances OLSR with Quality of Service support. This solution, taking radio interferences into account, ensures that QoS flows, if accepted by the admission control, will receive a bandwidth close to this requested. This solution has been implemented on the MANET/OLSR demonstrator of CELAR (MoD).
Routing based on packet delay distribution in multihop ad hoc network
Since propagation delays between routers are negligible, most delays occur in queueing and medium access control processing. Contrary to previous common belief there is no need of network synchronization. The objective is to proactively determine the delay in absence of packet data traffic. The estimate of delay distribution is done via analytical method. In order to keep control on quality of service flows we use source routing forwarding options.
Multicasting in mobile ad hoc networks
The goal of multicast protocols is to allow the network to deliver the multicast information to interested users. The multicast protocol builds and maintains a structure that will provide routes to all nodes in the multicast group; hence, they will receive the information multicast in their group. Multicast protocols can be classified according to the following criteria:
Multicast structures maintained by the multicast protocol: trees or meshes. We distinguish:
Shared tree. In the shared tree based family only one tree is built for each multicast group. Sources are not required to be a part of the multicast structure; they need an entry point to send their data to (the root of the tree for example, or the nearest tree member).
Source tree based. In the source based family, a tree is built for each tuple <source, multicast group>. For each multicast group we have several trees. Notice that IGMPv3  enables multicast source selection, which is straightforward with this kind of multicast tree.
Mesh based protocols maintain a structure containing all the participants to the multicast group; all the multicast sources and the multicast receivers. The target is to have several paths from one sender to each destination. Data is relayed and delivered through different paths to the receivers. Hence, it increases the robustness against link breakages. This robustness against the topology changes in mesh based protocols, are however more demanding in terms of bandwidth consumption compared to the tree based protocols which are more efficient in terms of resource usage.
Flat/Overlay structure. In the flat category, all nodes are assumed to handle multicast data and can participate in the multicast structure building and maintenance (tree, mesh). In the overlay category, multicast nodes of a same group build and maintain a virtual structure on top of physical structure that links all the participants using unicast tunnels. In this case, not all nodes within the network are supposed to know about the multicast protocol routing, they only have to forward the encapsulated multicast data that flows inside the unicast tunnels.
Performance evaluation of multicast protocols
The HIPERCOM team-project has designed three multicast protocols:
SMOLSR, an optimized broadcast protocol using the multipoint relays defined in OLSR;
MOLSR, a multicast protocol maintaining a source tree structure and using the topology information provided by OLSR;
MOST, a multicast protocol maintaining a shared tree structure and using overlays. It also uses the topology information provided by OLSR.
We have performed extensive simulations on the INRIA cluster with NS2 to quantitatively study the behavior of each protocol in different scenarios and configurations. The quality of the multicast is evaluated by the delivery ratio. The overhead induced by the multicast protocol is given by the number of retransmissions per multicast packet. We have increased:
the number of multicast groups,
the number of sources,
the number of clients in a group,
the source rate,
the number of network nodes,
With these results, we can deduce the applicability domain of each multicast protocol studied: SMOLSR, MOLSR and MOST.
Theoretical upper bound
We have derived a theoretical upper bound of the multicast capacity in wireless network. This result is an extension of Gupta and Kumar result about unicast capacity in wireless network. It is shown that the multicast delivery allows an increase of capacity of the order of the square root of the size of the multicast group compared to the attainable capacity if only parallel unicast connections were used. We have also shown that the protocol MOST actually attains this upper bound.
Geo-broadcast in wireless sensor or ad hoc networks
The technique of Multipoint Relays (MPRs) has proved its efficiency to optimize network flooding in mobile ad hoc networks (MANETs) and wireless sensor networks. Indeed, the number of redundant retransmissions of a message broadcast in a wireless network is reduced by ensuring that only a subset of nodes selected as multipoint relays are allowed to forward the received packets.
We have designed efficient solutions for broadcasting in a geographical area of mobile ad hoc and sensor networks. A solution is constituted by two modules:
the first one deals with MPR selection,
the second one defines the strategy for forwarding the received information.
We have then studied an example of application in a sensor network structured in different subareas. Each subarea is controled by a robot in charge of replacing failed sensors. We have evaluated the benefit brought by our solution in this specific context.
In traditional communication systems, nodes exchange data in packets, through relaying by intermediate nodes without modification of their content (routing). Seminal work from Ahlswede, Cai, Li and Yeung in has introduced the idea of network coding, whereby intermediate nodes are mixing information from different flows (different bits or different packets), for instance performing "exclusive or" between packets, before retransmitting them.
The Hipercom team is studying network coding specifically for MANET networks. It is mostly used as an efficient multicast/broadcast method (limiting the number of transmissions), and also as a reliable flooding mecanism. Hence, we are studying network coding in the context exclusively of energy-efficiency (and not capacity maximization).
We have proved theoretical results that extend previously obtained results. In another direction of work, we have designed a practical protocol to perform broadcast with network coding, DRAGONCAST in  ,  and  , which builds on these theoretical results. The main advantages of DRAGONCAST are its energy-efficiency and its simplicity. We analyzed it by simulations and simple models; it successfully illustres how (and how well) energy-efficient broadcast with a simple method could be performed with network coding.
Thomas Clausen is co-chair of the IETF working group Autoconfiguration in MANET .
A preconditioning for all routing protocols, OLSR included, is that each node is identifiable through an unique identifier We have developed, and published, a simple auto-configuration mechanism for OLSR networks, aiming a solving the simple but common problem of one or more nodes emerging in an existing network. Our solution is simple, allowing nodes to acquire an address in two steps: first, acquiring a locally unique address from a neighbor node. Then, with that locally unique address and using the neighbor from which the address was acquired as proxy, obtaining a globally unique address.
Security in OLSR
This issue is a hot issue in ad hoc networks since these networks are inherently open networks. We have reached the following results:
we have designed two security mechanisms to counter most of the attacks when we assume that there is no compromized nodes in the network; the first one has been implemented on the MANET/OLSR demonstrator of CELAR (MoD).
in presence of compromized nodes we have proposed mechanisms to detect compromised nodes or links and to remove such nodes or links in a numerous configurations of attacks.
OLSR with metrics
In practical networks, one property of many networks is that wireless transmissions may be done with the same equipment but with different parameters, such as modulations (with various payloads), transmission power, etc...This is true, for instance, for 802.11 networks, where different modulations are standardized.
However in the common OLSR routing protocol, this is not addressed, since the view is a binary view of links, which are considered either symmetrical (and then equivalent) or not usuable.
The question is how to take into account this ability to transmit in several manners, so that routing (with OLSR) is performed efficiently. We have proposed in  and in  , an extension of the OLSR routing protocol using metrics, that are well adapted to wireless networks with the characteristics of 802.11 networks.
Cross layer, sensor networks, energy efficiency
The diversity of the applications supported by wireless sensor networks explain the success of this type of network. These applications concern as various domains as environmental monitoring, wildlife protection, emergency rescue, home monitoring, target tracking, exploration mission in hostile environments... Sensor nodes are characterized by a small size, a low cost, an advanced communication technology, but also a limited amount of energy. This energy can be very expensive, difficult or even impossible to renew. That is why, energy efficient strategies are required in such networks in order to maximize network lifetime. Solutions to maximize network lifetime can be classified into four categories:
Topology control: These strategies adjust the transmission power of wireless nodes to spare energy;
Reduction of the volume of information transferred: These strategies aggregate data with or without clustering, optimize network flooding, tune the periodicity of information refreshment;
Nodes activity scheduling: as the sleeping state is the radio state consuming the least energy, these strategies make nodes sleep in order to spare energy, while ensuring network and application functions. Large benefits are expected.
Energy efficient routing: Such strategies notice that a multihop transmission is energy consuming and reducing the energy spent in the transmission of a packet from its source to its destination would increase network lifetime. Moreover, avoiding nodes with a low residual energy would also contribute to prolong network lifetime. Avoiding nodes that already have a high traffic load would reduce medium access contention, collisions if the medium access type is CSMA-CA and then spare energy lost in useless transmissions.
Energy efficient routing
Energy efficiency is a key issue in wireless ad hoc and sensor networks. Energy efficient routing is a way to improve energy efficiency and prolong network lifetime. We have shown how to extend the standardized OLSR routing protocol, in order to make it energy efficient. We have first defined an energy model for multihop transmissions. The energy cost of a one-hop transmission is evaluated, taking into account the energy lost in transmitting, receiving, overhearing and interferences. We have then evaluated the energy cost of multihop transmissions. Because of radio interferences, the selection of a unicast path, between a source and a destination, ensuring that each node has sufficient residual energy is NP-hard (see Mans 2006).
The OLSR extension we propose, called EOLSR, selects the path minimizing the energy consumed in the end-to-end transmission of a flow packet and avoids nodes with low residual energy. To take into account residual node energy, the native selection of multipoint relays of OLSR is changed. It considers the weighted residual energy of the multipoint relay candidate and its 1-hop neighbors. The cost associated with a multipoint relay candidate represents the maximum transmission duration that can be sustained by this node. Each two-hop neighbor must be covered by the candidate of maximum cost. These new multipoint relays are called EMPRs. They are used to build energy efficient routes, whereas the native MPRs are used to optimize network flooding.
No additional message is required in EOLSR. In order to select the EMPRs, the Hello messages include the residual energy of the sending node and of its one-hop neighbors. In order to compute the energy cost of a flow, we need to know the number of nodes up to two-hop of the node considered, assuming that interferences are limited to two hops. Hence, the TC (Topology Control) messages include the number of nodes belonging to the interference area of the TC originator.
As it has been shown that two-path routing is energy efficient, we compare EOLSR with a two-path source routing strategy: DL a two-path source routing with different links and DN a two path source routing with different nodes. As expected, native OLSR provides the smallest network lifetime. This shows that the selection of the shortest path is not sufficient to save energy. Concerning the two multipath source routing strategies, DN provides better results than DL. This is not surprising insofar as energy is dissipated per nodes and not per wireless link. Hence, DL that allows common nodes in the two paths can exhaust the energy of these common nodes more quickly. The main conclusion of these simulation runs is that EOLSR significantly outperforms DN and DL whatever the number of nodes. Moreover, the gain is increasing with the network size. EOLSR prolonges the network lifetime of 50% compared with OLSR for a network of 200 nodes. Notice that in the same conditions, DN prolonges the network lifetime of only 10%. Indeed, the two paths chosen by the source of the flow are used for all flow packets independently of the residual energy of these nodes. So the intermediates nodes exhaust their energy more quickly. This extensive performance evaluation allows us to conclude that EOLSR maximizes both network lifetime and the amount of data delivered.
The EOLSR protocol will be implemented in the OCARI project aiming at developing a wireless sensor communication module, based on IEEE 802.15.4 PHY layer and supporting EDDL and HART application layer and targeting applications in power generation industry and in warship construction and maintenance.
Nodes activity scheduling
In wireless ad hoc and sensor networks, an analysis of the node energy consumption distribution shows that the largest part is due to the time spent in the idle state. This result is at the origin of SERENA, an algorithm to SchEdule RoutEr Nodes Activity. SERENA allows router nodes to sleep, while ensuring end-to-end communication in the wireless network. It is a localized and decentralized algorithm assigning time slots to nodes depending on their color. Any node stays awake only during its slots and the slots assigned to its neighbors, it sleeps the remaining time. SERENA is based on distributed three-hop coloring. The node's color is then mapped in time slot. Such a solution supports late node arrivals. We have chosen node coloring because, unlike link coloring, it enables broadcast transmissions. Moreover, the existence of radio interferences requires at least two hop coloring. If a receiver is allowed to immediately acknowledge the received transmission, three-hop coloring is needed.
A performance evaluation allows us to compare SERENA coloring algorithm with existing ones such as Distributed Largest First, denoted DLF, both in terms of number of colors and complexity. SERENA and DLF use a similar number of colors, whereas the complexity of SERENA expressed in numbers of rounds is significantly lower. For a network of 200 nodes with a density of 10, the number of rounds in DLF is 273, whereas it is only 145 with SERENA. Moerover, it turns out that the number of colors used by SERENA depends (i) strongly on the network density and (ii) weakly on the number of nodes.
Simulation results show that SERENA maximizes both network lifetime as well as the amount of data delivered to the application. Moreover SERENA improves efficiency in the the node energy consumption. The first benefit of SERENA is that less energy is lost in the idle state. Indeed, if a node has nothing to transmit and its one-hop neighbors are not transmitting, the node is sleeping. The second benefit is that SERENA contributes to significantly reduce the interference phenomenon that becomes negligible. Hence, SERENA considerably improves the energy efficiency of wireless ad hoc and sensor networks. Moreover, SERENA increases the utilization of network resources such as bandwidth by means of spatial reuse.
The SERENA protocol will be implemented in the OCARI project. A strong cooperation with the MAC layer enables an efficient time slot allocation and an early detection of color conflicts caused by mobility. This cooperation improves the performances of SERENA in a network where bandwidth and energy are limited.
Many contractual collaborations:
MoD (QoS, security, interconnection between the OLSR and OSPF routing domains),
Hitachi (Vehicular applications, OLSRv2),
OCARI project (QoS, cross layer, energy efficiency),
SARAH project (QoS, localization),
Com2react (vehicular applications, multicast),
STIC INRIA-Tunisian Universities: the team of Prof. Leila Saidane at ENSI (Performance improvement in a sensor network),
Luceor (OLSR with metrics).