Research On Large-scale And Continuous Neighbor Discovery Techniques In MSN

The development of cheap sensor chips and mobile sensor networks has made docking appli-cation attract more and more attention and feasible for widespread deployment.The main goal of docking applications is to record the time when moving objects reach and leave particular places so as to learn the information such as the moving traces and behavior pattern of the moving object-s.Typical docking applications include tourists tracking in scenic spots or cultural heritage sites,hikers tracking and rescuing and wildlife habitat monitoring.In docking applications,there are usually two kinds of sensor nodes:static sensor nodes deployed at some particular places(also known as rendezvous points)and mobile sensor nodes attached on moving objects(also known as objects of interest).Through interaction between these two kinds of sensor nodes,the rendezvous time can be timely sensed and the related information exchanged.Thus,docking applications re-ly on the continuous neighbor discovery between mobile and static nodes.The accuracy of time recording and availability of applications depend on the latency of each discovery while the energy consumption in the discovery process determines the life time of sensor nodes and the robustness of applications.Therefore,the main challenge of neighbor discovery in docking application is how to reduce both energy consumption and discovery latency which are in conflict with each other.As for such challenge,based on the discovery feature between mobile nodes and static nodes in docking applications,this dissertation carries out the following research work:1.This dissertation firstly proposes mobility-assisted slot index synchronization to reduce the continuous discovery latency in docking applications.The neighbor discovery in docking applications is a large-scale and continuous process.This dissertation divides the neighbor discovery in such continuous process into two kinds:the initial neighbor discovery and sub-sequent neighbor discovery.Initial neighbor discovery happens when mobile nodes firstly join the network and reach the rendezvous points while the following neighbor discovery when mobile nodes move to other rendezvous points can be treated as the subsequent neigh-bor discovery.This dissertation finds that if all the static nodes can keep slot index synchro-nized,the subsequent neighbor discovery latency between mobile nodes and static nodes can be greatly reduced.However,in docking applications,static nodes can not directly commu-nicate with each other because of the deployed distance and complex environments.Thus,the slot index synchronization between static nodes faces great challenges.For this problem,this dissertation firstly proposes a distributed slot index synchronization technique:MASS(Mobility-Assisted Slot index Synchronization),where the mobile nodes are used as proxies to relay information among static sensor nodes,and gradually achieving the slot index syn-chronization between static nodes through the distributed reference selection and slot index synchronization algorithm based on priority.For the clock drift between sensor nodes,a dynamic clock skew estimation and an adaptive clock drift compensation method are also proposed.Meanwhile,for the accumulative error introduced by the synchronization which can lead to the worst-case discovery latency,this dissertation optimizes the state-of-the-art neighbor discovery protocols and avoid the worst case effectively.The experiment results show that MASS can reduce the average discovery latency by up to 2 orders of magnitude without increasing energy consumption of sensor nodes.2.This dissertation studies slot size optimization techniques to fully reduce the initial discovery latency.For initial neighbor discovery,slot size is a key factor which greatly affects the discovery latency.Given the fixed energy consumption,discovery latency will decrease proportionally while reducing slot size in theory.However,with the current working pattern in active slots adopted by most neighbor discovery protocols,the initial discovery latency will explode when the slot size is reduced to some extent.For this problem,based on the commonly used working pattern in active slots,this dissertation firstly studies the changing model between reducing slot size and discovery latency through measurement study and finds that the high probability of beaconing collision is the reason that causes the discovery latency to explode.Then,we propose slot size optimization techniques including reducing beacon density and introducing random jittering to reduce the the probability of discovery failure.The experiments results show that within small slot sizes,the proposed techniques can reduce the discovery latency up to 1 order of magnitude.3.This dissertation optimizes the active slot working pattern in traditional neighbor discovery protocols and firstly proposes heterogeneous working pattern in active slots to further reduce initial discovery latency.Meanwhile,based on the state-of-the-art neighbor discovery proto-col Searchlight,we argue that the working pattern in active slots can be changed to further reduce discovery latency when the slot size is reduced to some extent.This dissertation firstly proposes hetero geneous active slots working pattern,changing the homogeneous active slots into beaconing slots and listening slots.And then for the discrete listening slots which can also lead to discovery failure,a new continuous listening-based neighbor discovery protocol called Spotlight is proposed."The experiments show that Spotlight can achieve a 50%reduc-tion in discovery latency over existing state-of-the-art neighbor discovery protocols without increasing power consumption in existing sensor networks.At last,we take the tourists tracking in Mogao Grottoes which is a typical docking application as an example and validate the performance of the proposed neighbor discovery techniques through simulations based on the deployed tracking system and accumulated tracking records.Simulation results show that the proposed techniques can reduce the discovery latency of existing tracking system by 20 times and extend the life time of mobile nodes from 11 hours to 44 days.