A social-aware routing protocol for opportunistic networks

Highlights

• Swarm intelligence-based approach designed to intermittently connected networks.

• Greedy Ant Colony Optimization and Cultural Algorithms to address the routing problem.

• The utility of nodes as message forwarders considers a set of socialaware metrics.

• Operational metrics characterize the opportunistic social connectivity between nodes.

• CGrAnt outperforms Epidemic, PROPHET, and dLife in message delivery and redundancy.

Abstract

Understanding nodes mobility is of fundamental importance for data delivery in opportunistic and intermittently connected networks referred to as Delay Tolerant Networks (DTNs). The analysis of such mobility patterns and the understanding of how mobile nodes interact play a critical role when designing new routing protocols for DTNs. The Cultural Greedy Ant (CGrAnt) protocol is a hybrid Swarm Intelligence-based approach designed to address the routing problem in such dynamic and complex environment. CGrAnt is based on: (1) Cultural Algorithms (CA) and Ant Colony Optimization (ACO) and (2) operational metrics that characterize the opportunistic social connectivity between wireless users. The most promising message forwarders are selected via a greedy transition rule based mainly on local information captured from the DTN environment. Whenever global information is available, it can also be used to support decisions. We compare the performance of CGrAnt with Epidemic, PROPHET, and dLife protocols in two different mobility scenarios under varying networking parameters. Results obtained by the ONE simulator show that CGrAnt achieves a higher message delivery and lower message redundancy than the three protocols in both scenarios. The only exception is in one of the scenarios, when messages have a time to live lower than 900 min, where CGrAnt delivers a bit less messages than dLife, although with a lower message redundancy.

Introduction

The pervasiveness of computing devices and the emergence of new applications are factors emphasizing the increasing need for adaptive networking solutions. In most cases, this adaptation requires the design of interdisciplinary approaches as those inspired by nature, social structures, games, and control systems. The approach presented in this paper combining solutions from different, but yet complementary domains, i.e., networking, artificial intelligence, and complex networks. The aim addresses the problem of efficient data delivery in opportunistic and intermittently connected networks referred to as Delay Tolerant Networks (DTNs)(Chaintreau, Hui, Scott, Gass, Crowcroft, Diot, 2007, Khabbaz, Assi, Fawaz, 2012, Tournoux, Leguay, Benbadis, Whitbeck, Conan, de Amorim, 2011). Movement of nodes in such networks is not random and is a manifestation of their routine behavior. Together with contact-based interactions among nodes, this movement generates a mobile social network where contacts occur opportunistically in social environments such as conferences sites, urban areas, or university campuses.

We note that the complex environment of opportunistic DTNs challenges the application of any routing protocol. On the other hand, given that adaptation in nature is a permanent and continuous process, we believe that Swarm Intelligence (SI) methods, including approaches based on Ant Colony Optimization (ACO) (Dorigo, Maniezzo, & Colorni, 1996) and Cultural Algorithms (CAs) (Reynolds, 1994) can be adopted even in these complex environments. Among ACO characteristics that can contribute to routing on opportunistic DTNs, we point out: (i) auto-organization that induces no need of a central element to coordinate ants action (partial paths construction in the routing context); (ii) due to its parallel and scalability characteristics, ACO is suitable for complex routing processes involving high dimensions networks; (iii) the use of pheromone that can take advantage of the repetitive behavior of nodes in opportunistic networks; (iv) the use of heuristic-based decisions that can be useful for intermittent connections; (v) the capability of trading-off between local and global searches by constructively building a set of solutions that can be partially maintained or discarded depending on the dynamic of the current environment. Besides, the environment of opportunistic DTNs presents certain features in the mobility patterns of the network nodes that can be well explored by Cultural Algorithms (e.g., knowledge stored in the belief space can guide the swarms through new or already constructed partial paths depending on the node behavior).

The main innovations and contributions of CGrAnt are listed in the following: (i) differing from other protocols that use either only global or only local information, CGrAnt contains additional flexibility, because the decision is based on all available information (both ACO operators and knowledge stored in the CA belief space); (ii) to reduce the number control messages circulating in the net, search phases start only under demand. However, considering the sparse environment, searchers agents (ants) encapsulate data into the messages; (iii) assuming that DTNs are usually intermittently connected, the proposed protocol aims to avoid missing good paths by using event-guided evaporation mechanisms instead of cyclic ones; (iv) CGrAnt maintains a set of (partial) paths instead of only the best one; (v) in CGrAnt, the total of ants is dynamically defined; (vi) to take advantage of each encounter among nodes, the transition rule of CGrAnt is greedy instead of probabilistic; (vii) due to the absence of a central element, knowledge components and communication between population and belief spaces are spread among the nodes which store only partial information regarding the net. Considering the previous characteristics, we assume that CGrAnt can be suitable for running on different mobile scenarios (ranging from highly connected to sparse connected networks) due to its adaptive nature.

Motivated by those issues, this paper aims to extend a previous work (Vendramin, Munaretto, Delgado, & Viana, 2012a) by evaluating the use of CGrAnt to identify the most promising social-aware forwarders in two different DTN scenarios. Here, opportunistic and complex information (such as frequency and duration of contacts, centrality metrics, and mobility features) is also gathered and favorable paths along which to forward each message are determined hop-by-hop, while limiting data redundancy. However, in this paper we intend to show that the forwarding approach implemented through CGrAnt is adaptive and tailored to match forwarding decisions to different mobility conditions. Hence, we applied CGrAnt to two scenarios and compared its performance with that provided by dLife (in addition to PROPHET and Epidemic).

The remainder of this paper is structured as follows. Section 2 provides an overview of the principles that drive our approach. Section 3 describes the CGrAnt routing protocol in detail, and Section 4 presents the simulation environment. Section 5 investigates how the proposed operational metrics affect the CGrAnt’s performance. Section 6 compares the performance of CGrAnt with three known DTNs forwarding protocols under varying networking parameters, and finally, Section 7 summarizes the concluding remarks and future directions.