Investigating the use of Topology Adaptation for Robust Multi-Path Transport: A Preliminary Study Marco Arguedas, Carlos Perez, Marco Carvalho Institute for Human and Machine Cognition 40 S. Alcaniz St., Pensacola, FL 32502 Email: {marguedas,cperez,mcarvalho}@ihmc.us Kelli Hoback, Wayne Kraus Rockwell Collins 400 Collins Rd NE, Cedar Rapids, IA Email: {kahobak,wakraus}@rockwellcollins.com Abstract—In this paper we introduce a topology control algorithm for increasing transport robustness and efficiency by creating two simultaneous communication paths between the source and destination nodes of a data flow. We make use of a cross-layer substrate [1], [2] that allows us to detect the flows of data in the network while also allowing for a finer control of the routing. Because information of flow traffic in the network is available, the algorithm attempts to achieve intra-flow interference reduction by exploiting node mobility. I. I NTRODUCTION The issue of routing in mobile ad-hoc networks (MANETs) has been of interest to the research community in recent years. Several routing protocols that try to maximize net- work connectivity have been proposed[3], [4], [5]. However, some MANETs might have other concerns such as load balancing, fault-tolerance and/or improvement in bandwidth on top of node connectivity, and multi-path routing is one of the preferred mechanisms for addressing these issues. In this paper we present dpath, a distributed topology control algorithm designed to create, on demand, two simultaneous data communication paths between the source and destination nodes of a data flow. The algorithm creates the interference free data paths by moving nodes for the duration of the data flow. Assuming a common and fixed transmission frequency for all nodes, the algorithm adjusts the position of the nodes to create two data paths that are both link and edge disjoint, thus interference free under certain simplifying assumptions 1 . Given enough resources, the only two points of contention between the data paths are the source and destination nodes. Once the disjointed path has been created, the sender node may choose to utilize both paths to improve robustness by duplicating the transmission of each data packet over the paths. Alternatively, the sender node may chose to improve throughput instead by interleaving the transmission over each link. Figure 1 shows an instance of a network with disjointed paths between nodes ‘0’ and ‘11’. 1 This is strictly true for a unit disc topology. In most practical cases, however, the transmission interference range is significantly higher than the connectivity range (defined by the RSSI threshold) which would technically require the algorithm to be defined over two topologies, one representing the network connectivity and one for interference (a contention graph). Nevertheless, as currently proposed, the algorithm relies on the simplifying assumption that network propagation can be approximated to a unit-disc, but we expect that it will well tolerate small violations of the assumptions. Fig. 1. Dual disjoint path in place, allowing sender to choose between end- to-end robustness and throughput. II. RELATED WORK Popular choices of multi-path routing protocols include SMR (Split Multipath Routing) [6] and AODVM (Ad hoc On-Demand Distance Vector Multipath) [7]. Both of these protocols are on-demand (reactive) routing protocols. These modified versions of DSR[3] and AODV[4], respectively, have been adapted to handle multiple node-disjoint paths. In these algorithms, the selection of the disjointed path is realized at the destination node. Such decision is based on global information gathered from and provided by route-request messages. These algorithms, however, are bound by a common set of assumptions. First, the algorithms consider that the topology of the nodes is fixed, most likely because neither the mobility of the nodes or the transmission power of the devices can be controlled. Second, the mechanisms of route creation and selection do not take into consideration historical information of data traffic. Improvement of bandwith is an issue that has been at- tempted in [8], where the authors make use of a cross-layer approach. However, the authors try to solve the multipath routing problem using genetic algorithms, a solution that might be too complex to be performed in real-time.