|
Kemal Akkaya
Department of Computer Science
|
Office: FANER 2138
|
My research mainly focused on the problems at the network layer of wireless sensor networks. I have been working on novel and energy efficient algorithms that can handle delay-constrained routing, in-network data aggregation, routing security, base-station and sensor relocation and mobility of base-stations in severely resource constrained environments.
In addition, I am also interested in clustering, coverage and QoS problems in wireless sensor and actuator networks and wireless image/video sensor networks. I investigate new techniques which can deal with non-traditional trade-offs and technical challenges associated with special characteristics of those wireless networks.
 |
 |
 |
Mica-2 Motes from Crossbow which are used in our research
One of the problems I have worked on is the QoS routing of sensor data. The main motivation of such work is the increasing interest in sensor networks applications that require certain performance guarantees such as end-to-end delay, bandwidth etc. For instance, routing of imaging data in a battle environment requires careful handling in order to ensure that the end-to-end delay is within acceptable range and the images are received properly without any distortion. We have proposed a novel energy-efficient routing mechanism to ensure bounded delay for the data delivery in sensor networks. The proposed approach sets up energy-aware multi-hop data paths considering sensor’s transmission power and sensor energy reserve and imposes an end-to-end delay as a constraint. An end-to-end delay bounds is achieved through the use of a Weighted Fair Queuing (WFQ) based packet scheduling technique in each sensor node. WFQ also allows service sharing for real-time and non-real-time flows, giving chance to accommodate both best effort and real-time traffic at the same time. The protocol has been proven to provide timeliness without negatively affecting the energy consumption of the network.
In-network data aggregation can reduce the number of transmissions/receptions in the network and can be used in conjunction with delay constrained routing. For instance, in monitoring applications some queries that are subject to aggregation may require a bounded response time in order to ensure timely reaction to important findings. When processing such queries, in-network data aggregation should not only be performed in an energy-efficient manner but should also achieve timeliness for some designated paths from the source nodes to the sink. Therefore, we further extended the protocol in order to handle aggregation of delay constrained data in WSN.
While energy and QoS aware routing protocols, strive to find the optimum paths and employ new mechanisms in the sensor nodes, they do consider the sink node as fixed and does not move. When the sink is stationary, hops that are further from the sink have to be picked as relays to substitute the sensors close to the sink since they die quickly. If the sink has limited motion capability it will be desirable to relocate the sink close to an area of heavy traffic or near loaded nodes in order to decrease total transmission power and extend the life of the network. We have developed a novel relocation protocol for the sink in order to increase the lifetime of the network. Our protocol checks the traffic density of the nodes that are one-hop away from the sink and their proximity to the sink. Once the total transmission power for such nodes is guaranteed to be reduced more than a certain threshold and the overhead of moving the sink is tolerable, the sink starts to move to the new position.
Sink relocation can also be beneficial in applications involving real-time traffic. The quality of service achieved in these applications can start to diminish with increased volume of real-time data and most of the packets can miss their specified deadlines. In order to enhance timeliness in such situations, one of the solutions is to explore sink’s ability to move to a location where the volume of real-time data is high. We have also developed a protocol which periodically checks the deadline miss rate for real-time packets and triggers a relocation stimulus for the sink if the miss rate exceeds a certain threshold. In order to designate a new sink location, our approach finds the node that routes the largest number of real-time packets and checks whether moving to that location or close to that location affects the current routes or not.
Almost all of the energy-aware routing protocols that have been proposed for WSN relay data to a stationary sink. However, in numerous applications the mobility of the sink is desirable. One of the possible configurations of a mobile sink can be a computer installed on a moving object. The use of mobile sinks introduces a non-traditional trade-off between the need for frequent re-routing to ensure optimal network operation and the desire for minimizing the overhead of topology management. We have developed an energy aware mechanism for efficient and continual data delivery to a moving sink. Our approach tracks the distance of the sink from the last hops to the sink, and dynamically adjust the routes either via increasing the last hop nodes’ radio ranges or introducing new forwarder nodes as part of the routes. Rerouting is triggered when the current routes become unacceptably inefficient.
We further extended the protocol to accommodate both real-time and non-real-time traffic and provide a service differentiation mechanism in order to ensure timely delivery to the mobile sink. The main idea is to watch for sink’s reachability to sensor nodes and dynamically adjust the network topology in order to ensure timely data delivery. Our approach balances the energy and timeless goals of the network and prevents the potential excessive topology management overhead.
Simulation Tools
We are using our own C++ based simulator, NS-2 and Berkeley motes in order to evaluate the performance of our approaches.
 |
 |
Simulation of Wireless Sensor Networks with different tools
Data Gathering in Wireless Sensor and Actor Networks: Recently, a new type of network, namely wireless sensor and actor networks (WSANs) started to emerge and received attention with the idea of employing some powered nodes called actors within the sensor network for decision making and acting purposes. However, the research area is fairly new and new protocols at all layers of protocol stack are needed in order to fit to certain characteristics of WSANs. Currently, I am working on routing protocols for WSANs that would provide energy aware data gathering and on time acting for the actors. Since many actors can be deployed in the network, we are developing an architectural framework based on clusters each of which has a particular actor as the cluster head. Therefore, we are working on a distributed and simple clustering protocol with lower message complexity. As part of the architecture, we are also working at the level of inter-actor communication for providing an efficient way of collaboration among actor nodes.
Handling Mobility in WASNs: One of the unique characteristics of WSANs is having mobile actors. Therefore, the proposed architecture should be maintained when an actor, as a cluster-head goes out of transmission range of other sensors and/or actors. As a future work, we plan to work on algorithms for handling such cases without putting a major overhead on both sensors and actors.
Wireless Image/Video Sensor Networks: Since my wok on providing real-time routing in WSNs has many applications for transmission of images in battle environments, disaster areas, intrusion detection etc., it is important to reduce the size of transmitted images through either aggregation or clever compression techniques due to energy considerations. I plan to collaborate with people in image/video processing area in order to work on energy-efficient image/video transmission and aggregation in WSNs.
Energy-aware real-time MAC protocols for WSNs: I am planning to tackle the QoS provisioning problem at the Medium Access Layer (MAC) of WSNs. Our proposed protocols provide QoS only at the network layer. Therefore, we plan to exploit that work in order to provide certain end-to-end delay guarantees at the MAC layer of WSNs.
