|
|
 OPNET Technologies
| University: Rensselaer Polytechnic Institute "Asynchronous Slot Assignment and Neighbor Discovery Protocol for Wireless Networks (ASAND)" presented at OPNETWORK 2007. 1. Contention-Free MAC Protocols for Wireless Sensor Networks We study contention-free MAC protocols for WSNs. In our basic approach, time is divided into frames. Frames are further divided into timeslots. The objective is to assign a timeslot to each node such that no two nodes in a 2-neighborhood are assigned the same timeslot. Unlike previous time-division protocols, there is no central coordination point in the network, nor a perfect time synchronization among sensor nodes. The first protocol we proposed is based on a simple idea: each node randomly and uniformly chooses an index k between 1 and n, corresponding to the time slot tk, where n is called the global frame size. The local time that it chooses such a tk value is marked as init-time. When its current time is equal to init-time + tk, the node transmits a hello-message which contains its ID, the tk value, and its local transmission-time as the time stamp. If the transmission of hello-message is collision-free then the sender and all of its immediate neighbors mark their local time slots with and refrain using their local timeslot corresponding to tk (and we call that the sender becomes ready) In case of collision, the inequality tk ³ t1 + 2TT; k=1,...,n and TT=total transmission time, ensures that collisions occur only if more than one node choose the same ti value. In this case, colliding nodes should revise their timeslots by another random selection. We design and develop algorithms based on this simple idea and its more complex variants. We have designed a sensor node model in OPNET Modeler with a contention-free (CF) MAC using the above idea, and a new routing layer on top of it. The routing layer is a realization of our current work Green-Wave (GW) Routing, in which the basic idea is to route data from sensors to the closest sink node, though minimum-delay paths, considering the time slots assigned to nodes at the CF MAC layer. The preliminary version of the corresponding OPNET process, node, and network models can be downloaded as a bundle here, including a sample scenario with 20 nodes (18 sensors and 2 sinks). The CF MAC model is an exact implementation of the pSimpleMAC protocol we proposed in [1]. The GW Routing model is a variation of distributed Bellman-Ford algorithm, which was proposed in [2]. At this point, the execution of the algorithms can be observed by the OPNET Console during simulation. The debug information prints on the screen which time slots nodes obtain at the MAC layer and the routes selected by the nodes at the routing layer. The OPNET models are given in the following figures. (Click on the process models for full-scale versions.)
Sensor Node Model
CF MAC Process Model
GW Routing Process Model [1] Costas Busch, Malik Magdon-Ismail, Fikret Sivrikaya and Bulent Yener, "Contention-Free MAC protocols for Wireless Sensor Networks", submitted to the Journal of Distributed Computing. [2] Fikret Sivrikaya, Bulent Yener, "GreenWave Routing for Wireless Sensor Networks", Technical Report, Rensselaer Polytechnic Institute, 2005. | ||||||||||||||
| 2. Joint Problem of Power Optimal Connectivity and Coverage in Wireless Sensor Networks This work considers a multi-hop sensor network and addresses the problem of minimizing power consumption in each sensor node locally while ensuring two global (i.e., network wide) properties: (i) communication connectivity, and (ii) sensing coverage. We propose CARES: Connectivity-Assuring Randomized Energy-Saving Protocol, in which a sensor node saves energy by suspending its sensing and communication activities according to a Markovian stochastic process. We have developed Opnet process models representing this Markov model, and we will test the connectivity and energy saving characteristics of the porposed model on random sensor networks. Our primitive idea is to define two transition matrices that a node follows depending on whether there is an event or not. The Markovian models (as an Opnet process model) and the corresponding transition probabilities are given below. We first formalize and solve an optimization problem to optimize the transition probability parameters (a, b, g, d, p) so that the energy spent at a node is minimized (given the average energy usage for each state). Then we will feed this parameters into Opnet so that the nodes turn their transceivers/sensors on and off dictated by the Markov model with these parameters.
3. Power-aware / Interference-based Routing Algorithms for Multi-hop Wireless Networks. We study the energy efficient routing algorithms, coupled with power selection / power assignment problem in multi-hop wireless networks. We also integrate the effect of interference in wireless medium, and try to devise algorithms that minimize the interference throughout the network and thus increase signal quality of wireless communication. The decrease in interference also facilitates energy efficiency, by increasing reliability, i.e. success ratio of wireless links. We use OPNET Modeler with wireless module for simulations of designed algorithms, and compare their outcome with our analytic results. OPNET provides us the required flexibility, and saves us considerable time by the availability of a comprehensive library of models. 4. Random Walk-based Geometric Routing for Multi-hop Wireless Networks. In this project, we investigate the use of random walks for geometric multipath routing in multi-hop wireless networks. Having the network represented as a graph, a multi-hop packet transmission between a pair of nodes in the network can be regarded as a random walk between corresponding two vertices in the graph. We investigate ways of efficiently constructing the transition probabilities of random walks so that the packets reach their destination using small number of hops with guaranteed delivery and the packets of a single session might be routed over diverse paths, facilitating the spread of load or energy consumption throughout the network. We work on the design of algorithms that use local information only, hence are fully distributed and scalable. The OPNET Modeler provides us a convenient framework to test the performance of our algorithms by extensive simulations. | |||||||||||||||
