In Opportunistic Networks most of Internet''s basic assumptions do not hold true. Due to sparse density of nodes and frequent changes in network topology, an endto- end contemporaneous path may not exist. However, sporadic links emerging from coarse-grained mobility of nodes can be construed over a period of time, as presence of a complete path between a pair of nodes. Nodes hold a packet in permanent storage until an appropriate communication opportunity arises, which can help in further forwarding of the packet. In order to avoid packet loss, multiple copies of a single message are generally sent within the network, independently making their way to eventual destination. This design decision poses extra burden over network resources, and unnecessary utilization may result in degrading performance in resource-stringent environments. Hence, there is need to reduce this extra overhead, by determining e ective next-hop utility of nodes, and to better utilize network capacity with real time comprehension of dynamic network characteristic. Heterogeneity of nodes, in terms of capabilities or mobility patterns poses several challenges in de ning a utility function thatts all. Moreover, multi-hop routing protocols generally assume altruistic behavior of nodes. However, this assumption is not always true, as by agreeing to forward messages a node is contributing its resources such as memory, processing power, energy etc. Non-cooperative behavior may reduce e ective node density and can be devastating in opportunistic environments, where intermediary hops are required to share custody of messages. We target these issues in this thesis. In order to addressrst problem, we present a \Multi-Attribute Routing Scheme" (MARS) based on \Simple Multi-Attribute Rating Technique" (SMART) that collects samples of important information about a node''s di erent characteristics. This stochastic picture of a node behavior is then e ectively employed in calculatvii ing its next-hoptness. We also devise a method based on learning rules of neural networks to dynamically determine relative importance of each dimension. Hence, estimations based on an optimized combination of multiple parameters help in taking wiser decisions in relay nodes selection with inherent advantage of e cient utilization of network capacity. In second part of thesis, we analyze the aspect of nodes cooperation in challenged networks. We propose a novel framework to stimulate cooperation among nodes, which is deployed as an overlay to assist Destination-Dependent (DD) utility-based schemes. We envision that such an assistance mechanism to stimulate cooperation among nodes have the potential to help with practical deployments of DD utility schemes in real scenarios a icted with sel sh nodes.