Download presentation
Presentation is loading. Please wait.
Published byJack Morgan Modified over 6 years ago
1
Wireless Sensor Networks Positioning Algorithms & Energy Management
Sherry Adair Beaux Sharifi CS526 Spring 2005
2
Agenda Motivation Positioning Algorithms Energy Management References
11/23/2018 CS526 WSN Adair/Sharifi
3
Example Applications WSN’s have a variety of applications
Golden Gate has motes used to study the effect of wind, traffic, and earthquakes on the structure of the bridge (use accelerometers on a small section) 11/23/2018 CS526 WSN Adair/Sharifi
4
Example Applications (cont)
Biologists studying endangered redwood trees 2 month study using 50 motes measuring: Light, Temperature, Humidity UC Berkeley Biology Research 11/23/2018 CS526 WSN Adair/Sharifi
5
Research Focus Positioning Algorithms Energy Management
Positioning Algorithms are distributed heuristic algorithms used to determine the local or global coordinate positions of nodes in an ad-hoc wireless sensor network. Most applications implicitly require positioning information Most research topics are focused on methods for saving energy I also chose to study positioning algorithms based on sheer curiosity. 11/23/2018 CS526 WSN Adair/Sharifi
6
Positioning Agenda Background 3 Different Algorithms
Simulation Results Conclusion 11/23/2018 CS526 WSN Adair/Sharifi
7
Positioning Background
2D Trilateration 3D Trilateration a b b c Trilateration is a radio-location technique used for determining the relative position of objects using the geometry of triangles in a similar fashion as triangulation. Unlike triangulation, which uses angle measurements (together with at least one known distance) to calculate the subject's location, trilateration uses the known locations of two or more reference points, and the measured distance between the subject and each reference point. To accurately and uniquely determine the relative location of a point on a 2D Distance measurements can be made with AOA, TOA, TDOA, RSSI. d a c 11/23/2018 CS526 WSN Adair/Sharifi
8
Positioning Background (cont)
Two Major Difficulties to Positioning: Sparse Anchor Node Problem Range Error Problem Sparse Anchor Node Problem: a node requires at least 4 anchor nodes positions and relative distances between them and itself in order to derive its own positions within a 3D space. However, most nodes are not within direct radio range of all 4 anchor nodes. Range Error Problem: range measurements between nodes are imprecise. As positioning algorithms reuse previously computed ranges, errors can compound and propogate in the network, rendering the final results useless. Another problem with positioning algorithms comes from the need to have at least four points of reference with known locations in a three-dimensional space in order to determine the unique location of an unknown node. Given our assumptions above, there may only be four anchor nodes within a given network. It is very unlikely that every node within the network will be in direct communication with all four anchor nodes since the anchor nodes may be randomly located. Therefore, without some method of passing anchor positions between nodes, some nodes will not be able to determine their position within the network. This is known as the sparse anchor node problem. Unfortunately, range measurement errors can be as large as +-50% of the measurements due to factors such as whether any obstacles are present within the network, fading, and other sources of interference. Furthermore, poor triangulations results can compound as bad position information is spread and reused by additional nodes in the network. This is known as the range-error problem and can render the results of triangulation useless. 11/23/2018 CS526 WSN Adair/Sharifi
9
Positioning Algorithms
ABC – Assumption Based Coordinate [Savarese, Rabaey, Beutel, 2001] TERRAIN - Triangulation via Extended Range and Redundant Association of Intermediate Nodes [Savarese, 2002] Hop-TERRAIN [Savarese, 2002] Two-Phase [Savarese, 2002] First Phase: Hop-TERRAIN Second Phase: Refinement Both TERRAIN and Hop-TERRAIN solve the Sparse Anchor Node Problem. Two-Phase solves both Sparse Anchor Node Problem and Range Error Problem. 11/23/2018 CS526 WSN Adair/Sharifi
10
TERRAIN Algorithm Anchor Node Distance GLOBAL POSITION = ABC Algorithm
n1 = (r01, 0) n2 = (r012 + r022 – r122) , r022 – x22 ) 2 (6,6) (3,5) 2r01 1 3 3 (1,3) 4 4 (3,2) 2 (5,2) Anchor Node Distance GLOBAL POSITION = 1 ABC algorithm is an algorithm that allows a collection of nodes to locate themselves relative to each other, but not relative to any global coordinate system. TERRAIN: - each anchor node initiates ABC algorithm. - ABC algorithm propogates and allows each node to determine its relative position to each anchor node. This is the extended portion of the TERRAIN alg. - From one of these local anchor node maps, a node can compute its distance to that anchor node. - Process repeats until a node has 3/4 anchor node distances and then can perform standard trilateration to compute global position. Adv: Solves sparse anchor node problem. Disadv: Breaks down with range error problem. Compounds errors with error hop. Provides very inconsistent results = sqrt( ) = 8.5 1 3 (0,0) (5,1) 2 = 4.3 (3,0) 3 = 1.2 (18, 24) 11/23/2018 CS526 WSN Adair/Sharifi
11
Hop-TERRAIN Algorithm
Binary nature provides following benefits: No compounding of errors at each hop Provides consistent results Scales to much larger networks 3 2 2 1 3 Anchor Node Distance GLOBAL POSITION = 1 Very similar to TERRAIN except hop-counts are propogated from anchors instead of actual distances. Hop-TERRAIN: - each anchor initiates algorithm by broadcasting its position and hop-count of 0. - each neighbor forwards anchor’s position and hop-count + 1. - As a node acquires a hop count to an anchor node, it multiplies the hop-count by a hop metric to compute its distance to the anchor - Once a node has the distance to 4 anchor nodes, it can compute a standard trilateration to determine its global position. Adv: Solves sparse anchor problem. Doesn’t compound errors due to binary nature of hop-counts Provides consistent results Scales to large networks Disadv: Not very accurate though 1 = 6 = 3 * Hop Metric 2 = 4 3 = 2 (18, 24) 11/23/2018 CS526 WSN Adair/Sharifi
12
Two-Phase Refinement Algorithm
First Phase: Hop-TERRAIN Detects Edge Independence (for poor topologies) Second Phase: Refinement Iterative improvement of positions via ranges until position converges Uses Confidence Metrics (for convergence) In this phase, the sparse anchor node problem is solved by executing the Hop-TERRAIN algorithm as discussed above. This phase is only intended to give a rough approximation of positioning information and is used as the starting point for the next phase, Refinement. The objective of the Refinement phase is to improve the initial position estimates by iteratively estimating the ranges between nodes. This will, in turn, solve the range error problem by successively approximating the true ranges between nodes in the presence of range errors. Refinement: - a node broadcasts its position and receives position and range estimates from its immediate neighbors - node trilaterates its position and compares with previous position. - if position difference it greater than a threshold, its repeats process until position converges. - once converged, its broadcasts its position to its neighbors. A large assumption within the Refinement algorithm is that a node’s position estimate will actually converge to a more accurate final position. Positions typically converge due to the over-defined system of equations for trilateration (namely, Least Squares Algorithm). However, Savarese noticed a couple of reasons why nodes sometimes don’t converge: Range errors propogate quickly through the network. Nodes that have poor topology in the network In order to fix the fast error propagation problem, Savarese assigns a confidence metric to each node’s position. These confidence metrics are then used to weigh the over-defined system of linear equations in order to produce a more accurate estimation of a node’s position. This weighting helps nodes converge to a more accurate position by reducing the impact of erroneous range information (a22). In order to fix the anomaly in network topology problem, Savarese modifies the Hop-TERRAIN algorithm to make sure that none of the paths to anchor nodes share any common edges. By detecting independence of multi-hop routes to anchor nodes, anomalies such as ill-connected groups of nodes can be detected and prevented from entering the Refinement phase [a23]. Finally, the Refinement algorithm is summarized below in Figure 4. Adv: Solves sparse anchor problem with Hop-TERRAIN Solves range error problem with refinement Disadv: Range errors must be less than 40% otherwise worse than Hop-TERRAIN Uses a lot of communication (energy). Can be mitigated with infrequent running of alg. 11/23/2018 CS526 WSN Adair/Sharifi
13
Simulation Results TERRAIN vs. Hop-TERRAIN
Test of TERRAIN and Hop-TERRAIN’s sensitivity to Range Error Problem. Figure 5 demonstrates the average position error magnitude as the range error variance increases. Hop-TERRAIN shows very consistent results. Its average position error only varies from 133% to 309% (see upper table in figure). On the other hand, TERRAIN varies wildly as the range error variance increases. It demonstrates extremely high sensitivity to range errors with an average position error of 41% to 2.9E16%! Range Error Sensitivity of Hop-TERRAIN and TERRAIN (nodes = 40, anchors = 4, range = 10, grid = 30x30) 11/23/2018 CS526 WSN Adair/Sharifi
14
Simulation Results (cont) Hop-TERRAIN vs. Refinement
Figure 6 shows the average performance of the Hop-TERRAIN algorithm as a function of connectivity and anchor population. Figure 6 demonstrates that the Hop-TERRAIN algorithm has an average performance of under 100% error when there is at least a 5% anchor population and connectivity level of 7 nodes or greater. Error levels can reach as high as 250% when connectivity and anchor percentages are very low. Figure 7 shows the improvement of adding the Refinement phase after the Hop-TERRAIN algorithm. For the same 5% anchor population and connectivity of 7 nodes as we looked at before in Figure 6, Refinement improves the error level from 100% down to 33%. This is a 3-5x improvement. Maximum error levels are also improved with only a 54% maximum error level when connectivity and anchor percentages are low. Average Position Error After Refinement (5% Range Errors) Average Position Error After Hop-TERRAIN (5% Range Errors) 11/23/2018 CS526 WSN Adair/Sharifi
15
Simulation Results (cont) Hop-TERRAIN vs. Refinement
Figure 8 shows the result of comparing Hop-TERRAIN and Refinement as the level of range error increased while using a 10% anchor population and a node connectivity of 12 nodes. As we saw in Figure 5, Hop-Terrain is almost completely insensitive to changes in range accuracy. It has a consistent position error level of around 46% of range which is consistent with Figure 6. Refinement, on the other hand, offers better position estimates until the range error level reaches around 40%. At this point, it actually becomes worse than strict Hop-TERRAIN exceeding 50% position errors. Finally, the lower bound line shows the “best-case” position measurement using the true position of a node’s neighbors and the same amount of range error. It demonstrates that the Refinement algorithm has some room for improvement. Range Error Sensitivity between Hop-TERRAIN and Refinement (10% Anchors, 12 Nodes Connectivity) 11/23/2018 CS526 WSN Adair/Sharifi
16
Positioning Conclusion
Algorithm Sparse Anchor Problem Range Error Problem TERRAIN Hop-TERRAIN Two-Phase (< 40%) All three algorithms solve the sparse anchor node problem. However, only Two-Phase solves the range error problem. Except, only if range error are less than 40%. Note: that Hop-TERRAIN can be used exclusively if only ball-park accuracy is needed. 11/23/2018 CS526 WSN Adair/Sharifi
17
Future Positioning Research
Total Least Squares Algorithm Hop-Refinement First of all, since both Hop-TERRAIN and Refinement use the least squares algorithm as a primary vehicle for computing a node’s position, this algorithm becomes a primary candidate for improvement. Currently, the least squares algorithm assumes that only the range measurements contain errors. The reference positions are assumed to be without errors. This is a simplifying assumption. Savarese suggests a new algorithm called total least squares that assumes errors in both the range measurements and the reference points in order to better approximate node positions. Finally, due to the sensitivity Refinement has on range measurements and the consistency demonstrated in the Hop-TERRAIN algorithm, Savarese suggests a similar approach for Refinement called Hop-Refinement. This new algorithm would remove Hop-TERRAIN’s sensitivity to range measurement accuracy by only using hop-counts between neighboring nodes instead of range measurements. In particular, Hop-Refinement would use a local collection of multiple-hop neighbors in order to triangulate its position similar to Hop-TERRAIN. Since more than one-hop neighbors would be used, it would result in additional communication overhead which could again be mitigated by reducing the frequency of running the algorithm. 11/23/2018 CS526 WSN Adair/Sharifi
18
Energy Agenda Importance of Energy Management Sources of Wasted Energy
Methods of Reducing Energy Consumption Future Research Conclusions 11/23/2018 CS526 WSN Adair/Sharifi
19
Importance of Energy Management
Thousands of motes Not feasible to access them because of location, or quantity Reliability of application depends on motes continuing to operate Required to operate for many years 11/23/2018 CS526 WSN Adair/Sharifi
20
Source of Wasted Energy
Transmissions Collisions Overhearing Control packet overhead Idle listening Lossy links 11/23/2018 CS526 WSN Adair/Sharifi
21
Methods of Reducing Energy Consumption
Algorithms designed with power consumption in mind Special MAC protocols (S-MAC, B-MAC) Active/Sleep periods Decreasing the sensing coverage area Data Reduction Shorter, more reliable links Scavenging Power from solar, vibration using custom IC 11/23/2018 CS526 WSN Adair/Sharifi
22
Special MAC Protocols Needed to focus on energy management
Based on protocol Use active/sleep schedule Collision Avoidance Increase latency Reconfigure network based on current load (B-MAC) 11/23/2018 CS526 WSN Adair/Sharifi
23
Example of Energy Saved by Sleeping
System Components: StrongArm SA-1110 microprocessor Sensor Radio 11/23/2018 CS526 WSN Adair/Sharifi
24
Mica2 sleep savings Full operation of the sensor requires about ~15ma of current AA batteries supply ~1800 ma which would last about 120 hours or 5 days 11/23/2018 CS526 WSN Adair/Sharifi
25
Shorter, more reliable links
11/23/2018 CS526 WSN Adair/Sharifi
26
Energy Scavenging 11/23/2018 CS526 WSN Adair/Sharifi
27
Energy Scavenging (cont)
11/23/2018 CS526 WSN Adair/Sharifi
28
Energy Scavenging PicoRadio Meso-scale radio 11/23/2018
CS526 WSN Adair/Sharifi
29
Moore’s Law Capabilities increasing Costs staying the same
Power consumption staying the same Reduced power consumption for special purpose nodes 11/23/2018 CS526 WSN Adair/Sharifi
30
Future Research Renewable sources of energy
MAC protocols designed especially for WSN Custom low power ICs 11/23/2018 CS526 WSN Adair/Sharifi
31
Energy Conclusions Much energy is spent in the communication task of the mote, with almost as much energy required to listen as to send Special MAC protocols are required to address the special needs of WSN such as conserving power and adjusting to the changing network topology Active/sleep schedule is a common method used to conserve energy. Tradeoff is latency in packet delivery Possibility of extending the lifetime of motes using renewable energy sources such as solar and vibration 11/23/2018 CS526 WSN Adair/Sharifi
32
References Jason Hill, Mike Horton, Ralph Kling, Lakshman Krishnamurthy. The Platforms Enabling Wireless Sensor Networks. Communications of the ACM June 2004/ Vol47. No. 6. p C. Savarese, “Robust Positioning Algorithms for Distributed Ad-Hoc Wireless Sensor Networks”, Masters Thesis, 2002. C. Savarese, J. Rabaey, and J. Beutel, “Locationing in Distributed Ad-hoc Wireless Sensor Networks,” in IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), pages , Salt Lake City, UT, May 2001 Eugene Shih, Seong-Hwan Cho, Nathan Ickes, Rex Min, Amit Sinha, Alice Wang, and Anantha Chandraskasan. Physical Layer Driven Protocol and Algorithm Design for Energy-Efficient Wireless Sensor Networks. 11/23/2018 CS526 WSN Adair/Sharifi
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.