In this paper, we study how to adapt the routing according to dynamic network conditions in wireless ad hoc networks. We present a method that dynamically chooses routing agent between ad hoc on-demand distance vector routing protocol (AODV) with TCP (endto-end transport) and delay-tolerant networking (DTN) routing and bundle protocol (hop-by-hop transport). We use simulations to confirm that DTN routing and the bundle protocol leads to significantly shorter end-to-end delays and higher message delivery ratios than AODV and end-to-end TCP when the wireless node density is low. However, with high node density, DTN routing, especially epidemic routing, suffers from multiple bundle copies and simultaneous transmissions that lead into collisions and retransmissions at the wireless link (MAC) layer. Thus, we propose a simple adaptive scheme that uses only local information to transmit the messages from source to destination using either AODV or DTN routing, depending on current node density, message size, and path length to destination.

No abstract

Abstract. This paper investigates congestion control in opportunistic networks that use delay-tolerant networking (DTN) as a basis for communication. We propose a mechanism that advertises buffer occupancy information to adjacent nodes and avoids forwarding through nodes with high buffer occupancy. The nodes then achieve global congestion avoidance simply based on locally available information. The proposed mechanism works independent of the routing protocol and is thus applicable to wide array of scenarios. Extensive simulations, with different node mobility models and radio modeling, indicate that the proposed mechanism improves message delivery ratio and decreases end-to-end delay.

In this paper, we present two simple-to-implement measurement-based admission control methods for real-time services in IEEE 802.16e networks. The first method simply utilizes the averaged number of free slots as input in admission decisions while the second method is more advanced and it tunes the admission thresholds according to current traffic load. Our simulations show that the proposed methods lead into more efficient use of scarce radio resources -without sacrificing the Quality of Service (QoS) requirements -than using parameter-based, i.e., static, connection admission control.