Multi-channel Real-time Communications in Wireless Sensor Networks Xiaodong Wang Good afternoon, my name is XXX. I am a 2nd year Ph.D. student at the University of Tennessee. This is joint work with another student, XXX, and my advisor, Dr. Xiaorui Wang. 1
Introduction to Wireless Sensor Networks A wireless sensor network (WSN) consists of spatially distributed inexpensive miniature devices to cooperatively monitor physical or environmental conditions [Wikipedia] Cooperatively: communication with each other over wireless channels Monitor: capable of computation and sensing Inexpensive: Ideally less than 1 cent Have a wide range of potential applications industry, science, transportation, civil infrastructure, security, etc. In my presentation today, I will mainly cover the following sub topics. First,…. Then After that Finally
Comparison with Traditional Wireless Networks Integration of sensors, processors and radio Sensors are networked Very limited resource and very low cost Deeply embedded and networked Stationary locations May allow limited mobility Large Scale Deployment Data-centric communication Care about data instead of identity: mostly many to one communication Different from Internet or wireless ad hoc networks
Some Issues with WSN Energy Consumption End-to-end Communication Delay Energy resource is critical in WSN: provided by batteries. Normally don’t expect recharging or battery replacement Research goal: extend network life End-to-end Communication Delay Real-time is important for a lot of application Coverage The range of monitoring and the communication connectivity Synchronization Distributed system requires good timing knowledge. Clock drift on the local crystal causes trouble Security Wireless channel is insecure
Real-time Property Packet Reception Ratio (PRR) The probability of a packet to be successfully received by the receiver Real-time metric: , the total transmissions we need to successfully receive a packet. What is impacting real time service quality? Interference Packet cannot meet the required deadline because of collision and contention Long packet routing path High workload Application type decides traffic pattern: Event driven: End-to-end communication Data collection monitoring: May require data aggregation
Flow-Based Real-time Communication In Multi-Channel Wireless Sensor Networks Xiaodong Wang, Xiaorui Wang, Xing Fu, *Guoliang Xing, Nitish Jha Good afternoon, my name is XXX. I am a 2nd year Ph.D. student at the University of Tennessee. This is joint work with another student, XXX, and my advisor, Dr. Xiaorui Wang. EECS Department University of Tennessee, Knoxville * CSE Department Michigan State University
Empirical Study on Multi-Channel Communication With single channel, increasing sending power has significant impact to others’ transmission Almost 40% drop ratio when the communication power is low Experimental Setup Using multiple channels significantly mitigates the interference to other transmissions. Single Channel Multiple Channel Thus we conduct some empirical study on Multi-channel Communications The experimental setup is like this topology. We use three pairs of motes and synchronize their transmission. Calculate the packet drop ratio at the receiver of pair two and consider pair 1 and pair 3 as interfering pairs The first: same channel, all three pairs, Find, if using low power on the transmission of pair 2, drop ratio Then: different channels, We can see that even using low power at pair 2, the drop ratio is only about 3 percent This perfectly justify our idea to use multi-channels
Related Works Single channel real-time communication Implicit EDF Collision-free real-real time scheduling SPEED Enforcing uniform communication speed None of them takes advantage of multiple channels Multi-channel protocols and channel assignment Node-based protocols Requires channel switching Interference free channel assignment Requires synchronization Transmission power control Most do not deal with real-time requirement Several existing works are related to our paper There are some single channel real-time communication protocol Such as Implicit EDF and SPEED, but None Also several Multi-channel protocols and channel assignment has been proposed, but such as …….. But some of them require swtiching channels and others may have the synchronization requirement. Moreover, some work deal with transmission power control to decrease the interference, but …….
Outline Multi-Channel Real-Time protocol (MCRT) Network metric design Flow based channel allocation Power-efficient real-time routing (RPAR) Disjoint path search algorithm Experiment result Conclusion Based on these information, we propose our Multi-Channel Real-Time I will introduce the 1. 2. 3.
Multi-Channel Real-Time Protocol Multi-Channel Real-Time protocol (MCRT) Especially designed for the real-time application in multi-channel WSNs Designed to meet the end-to-end delay Application traffic type: many to one communication Major components: Flow-based channel allocation Power-efficient real-time routing Contributions: Formulate a constrained optimization problem Propose a heuristic to solve the problem Incorporate power efficient component Our Multi-Channel Real-Time Protocol…..is especially By proposing MCRT in the paper, we give the following contributions First, ………………..for the multi-channel real-time service problem Then, A heuristic algorithm to solve the problem is proposed Moreover, we incorporate the power efficient component in our protocol And finally we did massive simulations to illustrate the quality of our protocol
Network Metric Design Link weight calculation Worst-case one-hop delay Link weight = Number of transmissions = 1/PRR PRR : Packet Reception Ratio 3 Communication Relationships defined by PRR: Communication link: PRR > 90% Interference link: 90%>PRR>10% Cannot communicate: PRR<10% Worst-case one-hop delay End-to-end delay: Summation of the worst-case one-hop delay along a path In order to formulate our problem, we first need to design the network metric In our work, we use the 1/packet reception ratio as the link weight Also based on the value of packet reception ratio, we can have 3 communication relationships.
Flow Based Channel Allocation Channel assignment requirements: Different channels for different data flows Data flows are mutually disjoint Disjoint Path with Bounded Delay problem (DPBD) Directed graph G=(V,E) with weighted edges K source vertices s1,…, sk and a destination vertex t Goal: Find k disjoint paths, one from each source si to t All the path delay are bounded by a value W DBPD is NP-complete NP-completeness of DBPD can be proved by reducing it to MLBDP MLBDP problem: Maximum Length Bounded Disjoint Path Problem As we have the network metric finalized, we start to allocate channels for network. This is also the first main component of our protocol: We have two requirement for our channel allocation. First, we want to assign different channels for different data flows Then it gives the second requirement Different data flows are mutually disjoint
Power-Efficient Real-Time Routing Real-time forwarding Velocityrequired(s, d) =dist(s, d)/Tslack Tslack: the time remaining until the packet deadline expires Velocityprovided(n, p, c) =(dist(s, d) − dist(n, d))/delay(n, p, c) s = current node, d = destination, n = neighbor, p = power, c = n’s channel Feasible next hop: Velocityprovided > velocityrequired Power and neighborhood management Power adaptation If a set of neighbors are feasible, decrease power to transmit to the least power consumer Increasing the power to transmit to the max velocity provider, if no neighbors are feasible Neighborhood discovery If the transmission to all neighbors are requiring max power, but still cannot meet deadline, start neighborhood discovery by using Routing Request (RR) packet d Vrequired n Vprovided s The second component of our protocol is Power-Efficient Real-time Routing component, in short (RPAR). In this component, we first deal with the real-time forwarding issue: We calculate the required at the current node to the destination by the first formula, which is the distance between the current node to the destination devided by the slack time left in order to catch the deadline. Then we calculate the provided velocity from a certain neighbor choice by using the second formula. It need to calculate the euclidean between the current node and the certain neighbor and then devided by the delay we estimated for that neighbor under a certain power After we decided a neighbor choice, we will use a power adaptation scheme to manage the power consumption.
Outline Introduction Multi-channel real-time protocol Disjoint Path Search Algorithm Experiment result Conclusion
Disjoint Path Search Algorithm DBPD Problem: Directed graph with weighted edge, k sources and 1 destination Find k disjoint paths, one from each source si to t All the path delay are bounded by a value W Disjoint path search algorithm includes two phases Phase I: Initial solution set searching To search an initial solution set with some disjoint paths Dijkstra's shortest path algorithm is implemented Phase II: Augmentation algorithm Get as many disjoint paths as possible Phase I can only provide an incomplete solution set by fast searching scheme Depth first searching Matching scheme to the existing solution Phase II is iterative. Every round of phase II will add one more new disjoint path to the solution set Centralized algorithm
Phase II: Augmentation algorithm W X V S U C T B Y W X V S U C T B Y W X V S U C T B Y DFS Match New Path W X V S U C T B Y W X V S U C T B Y W X V S U C T B Y W X V S U C T B Y DFS/match One more path
Disjoint Path Search Algorithm (cont’) Analysis of the augmentation algorithm: Time complexity: O(W’2|V||E|) DFS: O(W’|E|) Matching algorithm O(W’2|V||E|) W’ is the edge number boundary V – number of nodes E – number of edges Extended DBPD problem Better fault tolerance More energy efficient neighbors to choose for real time forwarding In our work, we also extend DPBD problem. We add some shadow sources to the real source. Then in the extended problem, we need to find out disjoint paths from all the sources including the shadow source. By doing this, we can get more than one paths for a certain data flow so that the network will have more fault tolerance and more energy efficient neighbor as the next hop forwarding choice.
Outline Introduction Multi-channel real-time protocol Disjoint path search algorithm Experiment result Baseline design Simulation result Conclusion
Baseline Design SIMPLE Node-based scheme A flow-based distributed heuristic to find disjoint path Requires an initialization phase to establish path Using explorer packet Node-based scheme Every node has a default listening channel Node need to switch channel between receiving and transmitting Real-time Power Aware Routing (RPAR) Single channel real time protocol
Simulation Setup Simulation setup Major evaluation metrics NS-2 simulation, based on the characteristic of Mica2 sensor mote Probabilistic radio model from USC is implemented 130 nodes in a 150x150m2 square scenario, divided into 130 grids Major evaluation metrics Deadline miss ratio The percentage of packet that is missing the required service deadline Energy consumption per data packet Energy required for each scheme to successfully finish a work load under a certain deadline requirement Our experiment is conducted in NS-2 simulator, based on …… The probabilistic radio model from University of South California is implemented, which was not included in NS2 originally. The simulation uses 130 nodes in a 150 by 150 square meters scenario. We design two evaluation metric for our experiment result. The first one is deadline miss ratio, which is the percentage of packet that is missing the required service deadline. 2nd
Simulation Result (cont’) Performance with different deadlines MCRT outperforms other baselines on all the different deadlines Performance with different packet rate MCRT shows low miss ratio and energy consumption
Simulation Result (cont’) Performance with different number of flows MCRT is not impacted significantly by number of flows Performance for different network density MCRT is not sensitive to density
Conclusion The proposed MCRT protocol can effectively utilize the multiple channels for the many to one flow traffic pattern application MCRT greatly reduces the deadline miss ratio compared with a single channel real-time protocol and two baselines MCRT is the most energy efficient scheme among the four schemes MCRT has good scalability compared with others
Critiques No explicit end-to-end delay boundary is provided in the paper No explicit analysis for the real-time property after the channel is assigned. Worst case is a theoretical approach, but not realistic enough in real networks.
Department of EECS University of Tennessee, Knoxville Exploiting Overlapping Channels for Minimum Power Configuration in Real-Time Sensor Networks Good afternoon, my name is XXX. I am a 2nd year Ph.D. student at the University of Tennessee. This is joint work with another student, XXX, and my advisor, Dr. Xiaorui Wang. Department of EECS University of Tennessee, Knoxville 25
Introduction Multi-channel application Multiple channels supported by hardware Limited orthogonal channels, adjacent channels are overlapping Should we use overlapping channels? More channels resources to use is a benefit. The interference between overlapping channels is not negligible
Empirical Study for Overlapping Channel Overlapping channel interference Transmission pair uses Channel 16 and power level 15 Jammer pair uses Channel 15 Both pairs achieve good PRR when jammer use power level 16 -18
Problem to Solve Problem: Goal: A WSN with multiple data flows from different sources to the base station Assign channels (including overlapping channels) to the data flows Determine a transmission power level for every node Goal: To minimize overall (transmission) power consumption of the network To guarantee average end-to-end delay of each data flow to stay within a deadline
Contributions Overlapping channel interference reality Overlapping channels can be used for improving real-time performance Empirical models Received signal strength (RSS) vs. transmission power Packet reception ratio (PRR) model Power and channel configuration problem formulated Constrained optimization problem formulated Heuristic algorithm proposed Testbed established Experiment conducted on a 25-motes testbed.
Outline Introduction & contributions Related work review Empirical modeling of overlapping channel Overlapping Channel RSS Model Packet Reception Model Minimum transmission power configuration Empirical Results Conclusion
Interference Strength The interference strength is decided by the received signal strength (RSS) from the interference transmission. Higher interference signal strength -> more severe interference. Signal strength is essentially decided by the transmission power at the sender Overlapping Channel RSS Model Sender uses Channel 16 Linear RSS vs. Power model
Packet Reception Model Packet Reception Ratio (PRR) Decided by Signal to Interference and Noise Ratio (SINR) Relationship between SINR and RSS Packet reception model PRR-SINR-Channel relationship (SINRv, cv, PRRv)
? Problem Solving Flow Transmission Power RSS vs. Power Real-time constrained power minimization problem RSS SINR vs. PRR PRR
Outline Introduction & contributions Related work review Empirical modeling of overlapping channel Minimum transmission power configuration Problem Formulation Average PRR Estimation Collision Probability Calculation Algorithm Design Empirical Results Conclusion
Problem Formulation System assumption Many-to-one traffic pattern Each source generates a data flow independently All flows are disjoint Base station is assumed to be a super node, with multiple radio interfaces Power optimization problem formulation Subject to the constraints: Average packet reception ratio Soft real-time guarantees for Wireless Sensor Networks Sources generate independent random flows
Average PRR Estimation Average packet reception ratio estimation Low probability for more than two nodes to transmit at the same time Average packet reception ratio : P(u,w) : probability that node w's transmission collide with node v's reception of its own sender u PRR(u,v,w) : packet reception ratio of v's reception from its sender under w's interference Example: Case Probability PRR 1/PRR No interference at v 25% 100% 1 Interference at v 75% 50% 2
Collision Probability Calculation Probability of collision Verification of the average PRR estimation
? Problem Solving Flow Transmission Power RSS vs. Power RSS SINR vs. PRR ? Real-time constrained power minimization problem Goal: most energy efficient power configuration PRR
Algorithm Design The configuration search space is huge: jnkm m nodes forming n flows in the network. The total available number of channels on the equipment is j. Each mote can use k different power levels to transmit. Simulated Annealing (SA): probabilistic method for global optimization problems Reason for using SA: commonly used to find suboptimal solutions when the search space is huge and discrete
Algorithm Design (cont’) Algorithm based on Simulated Annealing Choose Tini, calculate the Cini and Pini Find neighbor state ΔP > 0 Calculate ΔP ΔP < 0 Accepted by Probability? No Yes Calculate Delayi Meet Constraint? No Add Penalty Yes Reduce Temperature T Program Terminate? No Yes end
Problem Solving Flow + Transmission Power RSS vs. Power RSS Simulated Annealing SINR vs. PRR Real-time constrained power minimization problem Goal: Best power configuration PRR +
Outline Introduction & contributions Related work review Empirical modeling of overlapping channel Minimum transmission power configuration Empirical Results Testbed and baselines Four different set of experiments Conclusion
Empirical Result Testbed setup and baselines 25 Tmote Invent motes are used Two Baselines: Orthogonal: orthogonal channels with simulated annealing power consumption algorithm Random: Random channel assignment with simulated annealing power consumption algorithm
Different Delay Constraints Topology I is used in this experiment Three channels, 16, 17 and 18, are used. Totally four flows are formed.
Different Delay Constraints (cont’) Overlapping scheme achieves a smaller average end-to-end transmission count and power consumptions Loose constraint gives larger search space leading to better performance
Different Packet Rates Higher packet rate leads to increased retransmission count and power consumption for each packet. higher degree of packet collision
Different Flow Numbers Topology II is used in this experiment Five channels, from 16 to 20, are used. Flow number is increased by one for each time
Different Flow Numbers (cont’) More data flows bring more interference into the network More flows are sharing the same channels for data transmissions
Different Network Size Topology III is used in this experiment Five channels, from 16 to 20, are used. Four flows are formed Node number is increased by one in each flow for each time
Different Network Size (cont’) More nodes result in larger end-to-end transmission count and power consumption More inter-channel and intra-channel interference is incurred
Conclusion Interference between overlapping channels is not negligible Overlapping channels can be utilized to reduce both end-to-end delay and power consumption with careful power configurations.
Critiques The channel assignment algorithm is a static algorithm, which has a large overhead for gathering the measurement of signal strength and packet reception ratio. The average packet reception ratio estimation only works under the proposed traffic pattern. No complexity analysis for the simulated annealing algorithm
Comparison MCRT Overlapping Goal Channel Assignment with Real-time Constraint Power Minimization with Real-time Constraint Channel Used Orthogonal Channel Only Overlapping Channels Channel Assignment Static Evaluation Simulation Real Test-bed result