Novel Battery-Less Wireless Sensors for Traffic Flow Measurement

Transcription

1 Novel Battery-Less Wireless Sensors for Traffic Flow Measurement Final Report Prepare by: Krishna ijayaraghavan Rajesh Rajamani Department of Mechanical Engineering University of Minnesota CTS 8-

2 Technical Report Documentation Page. Report No.. 3. Recipients Accession No. CTS 8-. Title an Subtitle 5. Report Date Novel Battery-Less Wireless Sensors for Traffic Flow Measurement November Author(s) 8. Performing Organization Report No. Krishna ijayaraghavan an Rajesh Rajamani 9. Performing Organization Name an Aress. Project/Task/Work Unit No. Department of Mechanical Engineering University of Minnesota Mechanical Engineering Church Street S.E. Minneapolis, Minnesota 5555 CTS project # 77. Contract (C) or Grant (G) No.. Sponsoring Organization Name an Aress 3. Type of Report an Perio Covere Intelligent Transportation Systems Institute University of Minnesota Transportation an Safety Builing 5 Washington Ave. SE Minneapolis, Minnesota Supplementary Notes Abstract (Limit: wors) Final Report. Sponsoring Agency Coe This project presents a novel battery-less wireless sensor that can be embee in the roa an use to measure traffic flow rate, spee an approximate vehicle weight. Compare to existing inuctive loop base traffic sensors, the new sensor is expecte to provie increase reliability, easy installation an low maintenance costs. The sensor uses power only for wireless transmission an has ZERO ile power loss. Hence the sensor is expecte to be extremely energy efficient. Energy to power this sensor is harveste entirely from the short uration vibrations that results when an automobile passes over the sensor. A significant portion of the project focuses on eveloping low power control algorithms that can harvest energy efficiently from the short uration vibrations that result when a vehicle passes over the sensor. To this effect this report evelops an compares three control algorithms Fixe threshol switching, Maximum oltage switching an Switche Inuctor for imizing this harveste energy. The novel Switche inuctor algorithm with a ual switch control configuration is shown to be the most effective at imizing harveste energy. All three of the evelope control algorithms can be implemente using simple low power analog circuit components. The evelope sensor is evaluate using a number of experimental tests. Experimental results show that the sensor is able to harvest aequate energy for its operation from the passing of every axle over the sensor. The sensor can reliably an accurately measure traffic flow rate. 7. Document Analysis/Descriptors 8. Availability Statement Traffic sensor, wireless traffic sensor, energy harvesting, battery-less wireless sensor, traffic flow measurement No restrictions. Document available from: National Technical Information Services, Springfiel, irginia 9. Security Class (this report). Security Class (this page). No. of Pages. Price Unclassifie Unclassifie

3 Novel Battery-Less Wireless Sensors for Traffic Flow Measurement Final Report Prepare by Krishna ijayaraghavan Rajesh Rajamani Department of Mechanical Engineering University of Minnesota November 8 Publishe by Intelligent Transportation Systems Institute Center for Transportation Stuies University of Minnesota Transportation an Safety Builing 5 Washington Ave SE Minneapolis, Minnesota 5555 The contents of this report reflect the views of the authors, who are responsible for the facts an the accuracy of the information presente herein. This ocument is isseminate uner the sponsorship of the Department of Transportation University Transportation Centers Program, in the interest of information exchange. The U.S. Government assumes no liability for the contents or use thereof. This report oes not necessarily reflect the official views or policy of the Intelligent Transportation Systems Institute or the University of Minnesota. The authors, the Intelligent Transportation Systems Institute, the University of Minnesota an the U.S. Government o not enorse proucts or manufacturers. Trae or manufacturers names appear herein solely because they are consiere essential to this report.

4 ACKNOWLEDGMENTS The author(s) wish to acknowlege those who mae this research possible. The stuy was fune by the Intelligent Transportation Systems (ITS) Institute, a program of the University of Minnesota s Center for Transportation Stuies (CTS). Financial support was provie by the Unite States Department of Transportation Research an Innovative Technologies Aministration (RITA).

5 TABLE OF CONTENTS I. Review of Current Traffic Sensors... II. New Battery-Less Wireless Traffic Sensor... A. Overview... B. Principle... C. Harware... D. Controller Cont... E. System Moel for Control... III. Control System Design an Analysis... 7 A. Fixe Threshol Switching... 8 B. Max oltage Switching... C. Switche Inuctor... 3 D. Effect of MOSFET on Switche Inuctor... 8 E. Comparison of Control Algorithms... I. Experimental Results.... Traffic Sensor Networks... A. Introuction... B. Encoer Decoer... C. Experimental Results... 7 I. Completion of Project Tasks... 3 A. Full Sensor s Basic Sensor... 3 B. Use of SAW Device for Battery-Less Wireless Operation... 3 II. Conclusions References... 3

6 LIST OF FIGURES Figure : Photograph of Sensor Base on the First Design... 3 Figure : Photograph of Re-Designe Sensor... 3 Figure 3: Sensor Dimensions... Figure : Energy Harvesting Circuit... Figure 5: Force Input Use for Simulation... 7 Figure : Mechanical Response of the System... 8 Figure 7: State Transition Diagram for Fixe Threshol Switching... 8 Figure 8: oltage Output for Fixe Threshol Switching Algorithm... 9 Figure 9: Loa Current for K Loa with Fixe Threshol Switching Algorithm... 9 Figure : Instantaneous Power Consume by K Loa with Fixe Threshol Switching Algorithm... 9 Figure : State Transition Diagram for Max oltage Switching... Figure : oltage Output for Max oltage Switching Algorithm... Figure 3: Loa Current for K Loa with Max oltage Switching Algorithm... Figure : Instantaneous Power Consume by K Loa with Max oltage Switching Algorithm... 3 Figure 5: Energy Harvesting with Inuctor... Figure : State Transition Diagram for SW L ( Switche Inuctor )... Figure 7: State Transition Diagram for SW P ( Switche Inuctor )... Figure 8: oltage Output for Switche Inuctor Algorithm... 7 Figure 9: Loa Current for K Loa with Switche Inuctor Algorithm... 7 Figure : Instantaneous Power Consume by K Loa with Switche Inuctor... 7 Figure : Mosfet in Max Switch... 9 Figure : Simulation Results... 9 Figure 3: Results from Two Sets of Experiments... Figure : Theoretical oltage Obtaine from Switche Inuctor vs the Theoretical oltage Obtaine from Max Switching Algorithm... Figure 5: Fixe Threshol Algorithms with a Threshol of Figure : Max Switching Algorithm... 3 Figure 7: Switche Inuctor Algorithm... 3 Figure 8: Max Switching... Figure 9: Switche Inuctor... Figure 3: Capacitor oltage Depenent on ehicle Weight (Left: Car Driven Close to the mph; Right: Motorcycle Driven Close to the 5 Figure 3: Circuit Schematic... 7 Figure 3: Storage Capacitor oltage ( st Configuration of Encoer)... 8 Figure 33: Decoer Output ( st Configuration of Encoer)... 9 Figure 3: Storage Capacitor oltage ( n Configuration of Encoer)... 9 Figure 35: Decoer Output ( n Configuration of Encoer)... 3

7 EXECUTIE SUMMARY This project presents a novel battery-less wireless sensor that can be embee in the roa an use to measure traffic flow rate, spee an approximate vehicle weight. Compare to existing inuctive loop base traffic sensors, the new sensor is expecte to provie increase reliability, easy installation an low maintenance costs. The sensor uses power only for wireless transmission an has ZERO ile power loss. Hence the sensor is expecte to be extremely energy efficient. Energy to power this sensor is harveste entirely from the short uration vibrations that results when an automobile passes over the sensor. A significant portion of the project focuses on eveloping low power control algorithms that can harvest energy efficiently from the short uration vibrations that result when a vehicle passes over the sensor. To this effect this paper evelops an compares three control algorithms Fixe threshol switching, Maximum oltage switching an Switche Inuctor for imizing this harveste energy. The novel Switche inuctor algorithm with a ual switch control configuration is shown to be the most effective at imizing harveste energy. All three of the evelope control algorithms can be implemente using simple low power analog circuit components. The evelope sensor is evaluate using a number of experimental tests. Experimental results show that the sensor is able to harvest aequate energy for its operation from the passing of every axle over the sensor. The sensor can reliably an accurately measure traffic flow rate. The with of the wireless transmission pulse from the sensor is roughly proportional to the weight of the vehicle passing over the sensor. Thus the vehicle weight can be approximately measure. Each sensor is provie with a uniquely ientifie encoer that enables the receiver to ientify the specific sensor from which it receives wireless transmission. This will enable networking of multiple sensors (at intersections an on highways) with a single transceiver.

8 I. REIEW OF CURRENT TRAFFIC SENSORS Transportation agencies all aroun the country monitor traffic flow rates on most major highways using inuctive loop etectors (ILDs). The Minnesota Department of Transportation (MnDOT) for example, monitors the flow rates at over points in the Minneapolis/St. Paul metro area using such ILDs. An ILD consists of a big loop of metallic coil burie in the lane. This loop is connecte to a station which powers the loop an processes the information obtaine from the loop to etermine if a vehicle passes over the sensor. The flow rate information from such sensors is use to control ramp meters, ientify congestion points, etect incients an for a number of other applications. Inuctive loop etectors exhibit high accuracy in etecting vehicles ([7]). Thus espite various new non-intrusive technologies for etecting vehicles such as image processing base etectors ([5],[8],[3],[5]) an systems base on auio processing ([], []), inuctive loop etectors remain the most wiely use technology. Despite their popularity, ILDs are far from perfect an there has been consierable work to improve etection using better moels, better filtering technology an by using better ientification techniques such as Fuzzy Logic an Artificial Neural Networks (ANNs) ([], [], [5]). Despite many improvements, the installation of the ILD involves cutting a large section of the roaway in each lane an therefore causes consierable traffic isruption. Owing to its operating principle, the ILD nees to be continuously powere resulting in consierable ile power loss. For example, an ILD nees to be continuously powere uring the night, even if there is very little traffic flowing on a particular highway. Research continues to be one on evelopment of other new traffic sensors. The Hi-Star portable traffic sensor from Quixote Transportation Technologies is a portable traffic analyzer that can be place on the roa surface in the traffic lane an connecte to a computer for ata retrieval. It is useful for applications where a temporary traffic sensor is require, for example for conucting traffic an turn analysis at an intersection or for surveying traffic on a brige or parking garage. The traffic sensor evelope in this project is unique an ifferent from all the sensor technologies escribe above. Its uniqueness comes from the fact that it is the first ever sensor that is battery-less, wireless an is powere entirely by harvesting energy from vibrations for its operation.

9 II. NEW BATTERY-LESS WIRELESS TRAFFIC SENSOR A. Overview This project has evelope a novel battery-less wireless traffic sensor, which is extremely energy efficient. The sensor is completely autonomous an can be embee in the lane without the nee for control/ata cables. In the absence of any automobile, the sensor is completely turne off, consuming no power. Thus, the sensor has ZERO ile power loss. When an automobile passes over the sensor, the sensor is turne on an a RF pulse is transmitte wirelessly to the station. The sensor requires no external power source as it is powere by harvesting all its energy from vibrations that result when a vehicle passes over it. Further this sensor has smaller imensions an can be installe with much lower traffic isruptions. This is especially true because the sensor oes not nee a power source an power lines o not nee to be run to the sensor. This new sensor, like the ILD, oes not use complex image processing or auio processing techniques an woul hence provie the same level of high reliability. Owing to the battery-less an wireless nature of the sensor low maintenance can also be expecte. Further the sensor can measure the number of axles an the approximate weight of the passing vehicle in aition to the traffic flow rate. It is also possible to configure several sensors to transmit to a single station by transmitting a unique coe using a programme encoer. B. Principle The propose sensor is base on the principle of vibration harvesting energy (HE) to enable wireless transmission of signals. Soana et. al () ([]) provies a goo review of many of these HE techniques. Some of the earlier work has also focuse on eveloping control algorithms to optimize the amount of energy harveste ([], []). However, the HE techniques in literature focus preominantly on harvesting energy from a continuous source of vibration. When a vehicle passes over the sensor, the resulting vibrations mechanical are of short uration. Hence, although the concept of HE is not new, it has never before been use to power a traffic sensor. Further the optimal algorithms that have been propose earlier cannot be implemente in a stan-alone sensor as they require an external control input (an possibly an external power source). Hence new algorithms have been evelope an implemente in this project. C. Harware The propose sensor consists of a ouble beam structure with a main beam an two aitional support beams. The first sensor esigne by the team was essentially a two-imensional structure with the two lower support beams being perpenicular to the main beam. A photograph of this sensor is shown in figure.

10 Figure : Photograph of Sensor Base on the First Design This sensor was later replace by a much more compact esign in which the two lower beams are along the same axis as the upper main beam. This leas to an essentially one-imensional structure. A photograph of the re-esigne one-imensional sensor is shown in figure. Figure : Photograph of Re-Designe Sensor 3

11 The main beam of the sensor is (or.8 meters) long an the two support beams are long (or 5 mm) at the ens. A schematic of the sensor with imensions is shown in figure 3. A total of eight Piezo elements (four s for each of the support beams) are bone at the locations shown in figure 3 an connecte electrically in parallel. Finite element simulations using ANSYS reveale that the average of the over the area of all the s epene only on the total loa acting on the main beam. The ouble layer beam configuration was chosen since the average voltage evelope by the woul be inepenent of the lateral location of the loa an the sensor can thus etermine the weight of the passing vehicle. It shoul be note further that the spee of the passing vehicle can be measure by measuring the time ifference in the loaing between two consecutive sensors place a short longituinal istance apart. Since each axle applies a istinct loa on the sensor, the number of axles on the vehicle can be counte. D. Controller Cont In this paper, we evelop a controller to optimize energy harveste from short uration inputs from near impact loaing. This technique coul be extene to harvest energy from other sources such as shock absorbers an laning gear in airplanes uring touch own. The emphasis on this paper has been on eveloping control strategies for the Energy Harvesting Systems (EHS) that are completely powere from the energy that is harveste. The esign has been restricte to a controller that can be implemente using simple onboar analog electronics. The efficacy of these control strategies have been verifie using simulations an experiments. E. System Moel for Control The EHS consists of electric substrate bone to a beam structure as shown in figure 3. When a vehicle passes over the sensor, the experiences a from the loaing. The on the results in a voltage being evelope in the. This voltage causes a reaction force on the beam structure. This force woul couple the mechanical ynamics of the beam structure with the electrical ynamics. However the Piezo element has a cross-section of 5mm.9mm which is much smaller cross section of the parent material. Hence the force generate by the can be neglecte in calculating the overall, effectively ecoupling the mechanical ynamics from the electrical ynamics. The overall ynamics of the system can be moele as a cascae ynamics system. Figure 3: Sensor Dimensions The mechanical system that rives the electrical system consists of the vibrating beam structure. For a simple beam structure in vibration, the various moes of vibration can be calculate using equation (). More complicate structures require a FEM moel solution.

12 ν ( x, t) ν ( x, t) ν ( x, t) ρ A + b + EI = () t t x The ominant frequencies thus obtaine can be use to construct a low orer system moel. Thus for the purposes eveloping the control system, the mechanical system can be consiere to be a spring mass amper system with the equation () an the is calculate from equation (3) u& + ζω u& + ω u n n = F m & () u = l ε (3) where u is the isplacement of the mechanical system l is the length scale associate with the mechanical system. At low frequencies, the Piezo electric material which is the critical part of the EHS, is moele as a voltage source in series with a capacitance using equation () & (5) ([], [8]). A more sophisticate moel can be foun in Weinbert et. al.([9]). () = i pt C = δ ε (5) where is the voltage measure across the Piezo, is the voltage open circuit voltage generate ue to the C is the capacitance of the Piezo i is the current through the Piezo p is the Piezo constant efine as the evelope per unit applie electrical fiel δ is the thickness of Piezo ε is the in the Piezo, from eq. (3) The electrical system shown in figure also consists of a brige rectifier (enote by Dioe Brige ) connecte to the Piezo element. The ioe brige is connecte to a storage capacitor C, which in turn is connecte to transmitter via a loa switch (SW s L ). The transmitter is moele as a loa resistor R in the circuit. The control circuit is not inclue in the ynamics owing to its L extremely small current consumption. In orer to calculate the overall electrical ynamics, each ioe making up the brige is moele by a piecewise linear moel ([]). The current p i can then be written as equation () an the ynamics of the capacitor voltage can be given by equation (7). 5

13 R i p = C & s + RL = i p C & = i s p sign if ( ) > otherwise if SW if SW L L + is close is open () (7) where is the forwar voltage rop across each ioe (about.7 to. base on the type of ioe use) Figure : Energy Harvesting Circuit

14 III. CONTROL SYSTEM DESIGN AND ANALYSIS Of the total energy generate in the Piezo, only the fraction transferre to the storage capacitor C is available to rive the loa. This energy can be calculate in terms of the imum voltage s across the storage capacitor C s as. In the following section, the available voltage is etermine for each of three ifferent control algorithms. For the purpose of simulation, we consier the mechanical system with a mass of Kgs, a amping ratio of.7 an a spring constant of N/m. These parameters were chosen so as to provie a natural frequency of Hz in the simplifie mechanical moel. The loa is shown in figure 5, an the response of the mechanical system is seen in figure. The system was esigne so that the Piezo woul prouce = for the static weight of the automobile. The loa switch, SW L, shown in figure is implemente using a MOSFET transistor. The loa resistor is of the orer of KΩ. If SW L is always close, this loa resistor is always connecte to the Piezo. In fact by using equations (-7) in a Simulink moel, it can be shown that the peak magnitue of woul be.7. This voltage cannot rive the transmitter, which requires a supply of.5 at the very least. 3 5 Force (N) time (sec) Figure 5: Force Input Use for Simulation 7

15 3.5 3 Displacment (mm) Figure : Mechanical Response of the System A. Fixe Threshol Switching This is an existing algorithm for energy harvesting an the simplest []. In this algorithm, the loa is connecte to the storage capacitor ( C ), by setting the control input to logic high () when s the voltage across this capacitor ( ) crosses a preetermine on-threshol high. The control is turne off (), if falls below an off-threshol. The control signal to SW low L, can be given by the control law state transition iagram shown in figure 7. > high <= control high = = < control low >= low Figure 7: State Transition Diagram for Fixe Threshol Switching Once the loa switch SW L is close, the voltage across the storage capacitor oes not buil any further. The imum of value is equal to the on-threshol high. The fixe threshol switching is the simplest algorithm, an woul serve as a baseline for evaluating the performance of the other control algorithms. Simulation results obtaine using this baseline control law is shown below. The voltage in the storage capacitor is seen in figure 8, while figure 9 an figure, respectively, show the instantaneous loa current an power at the loa. 8

16 3.5 3 oltage () time (sec) Figure 8: oltage Output for Fixe Threshol Switching Algorithm Current (ma) time (sec) Figure 9: Loa Current for K Loa with Fixe Threshol Switching Algorithm Power consume at Loa (mw) time (sec) Figure : Instantaneous Power Consume by K Loa with Fixe Threshol Switching Algorithm 9

17 B. Max oltage Switching In this new algorithm, the loa is connecte to the storage capacitor ( C ), when the voltage s across this capacitor ( ) reaches a imum value. The control SW L is turne off, if falls below the off-threshol low. The control law can be given by the state transition iagram shown in figure. The occurrence of imum can be etermine using analog electronics. For instance, the -etector can be realize using a high pass RC filter given by equations (8). A imum is eclare when the output of this filter falls below a threshol. The value of this threshol is small an etermines how close to zero the erivative must become for the voltage to be recognize as imum. R R + = (8) filter is imum control = = control or < low < low >= low Figure : State Transition Diagram for Max oltage Switching If the isplacement of the beam has only one extremum value u (with a corresponing = ), then can be calculate using C C + C = C = s + C ( ) ( ) if otherwise if otherwise > > (9a) (9b) Thus for a sufficiently large voltage, the ifference between between C an s C in the inverse ratio of their capacitance. an is istribute

18 It must be note that for the Fixe Threshol Switching algorithm to work reliably, nees high to be chosen to etect every single vehicle. Hence high must be chosen a volt or two lower than the lowest obtaine from (9) corresponing to the smallest. Thus, although it is not irectly evient, obtaine from Max oltage Switching algorithm is necessarily larger than that obtaine from Fixe Threshol Switching algorithm. Equation (9) can be erive as follows. It is clear that = + (refer figure ). Thus if C C >, the brige circuit rectifies the current an charges the storage capacitor. When, the ioes block the flow of current thus preventing C storage capacitor from ischarging. If i p oes not change signs an C > effective voltage riving the current thought the resistive element in the circuit is given by effective () t = sign( () t C () t ) ( () t () t () t,) C () In moeling overall ynamics, the first orer nonlinear electrical ynamic equations (-7) are ominate by the much slower ynamics of the mechanical system. The system exhibits a two time scale property an the faster electrical ynamics nees to be moele by its quasi-steay state value ([9], []) which correspons to ( t) =. Equivalently effective () t C () t = () t () Since signs Thus C an s, the C are two capacitor connecte in series, as long as & () t oes not change () t C s = () C () t C C + C = C s C + = s C C ( ) if ( ) otherwise if otherwise > > (3) The above equation irectly yiels equation (9).

19 For the current setup, with =, C = 85nF, = μf, C s is calculate to be.8. 5 oltage () time (sec) Figure : oltage Output for Max oltage Switching Algorithm 5 Current (ma) time (sec) Figure 3: Loa Current for K Loa with Max oltage Switching Algorithm

20 3 Power consume at Loa (mw) time (sec) Figure : Instantaneous Power Consume by K Loa with Max oltage Switching Algorithm Since the storage capacitor is allowe to charge to a higher voltage, this algorithm elivers a larger peak power in comparison to the Fixe Threshol switching. It is seen that the imum power of the Max oltage Switching (figure ) is nearly twice the imum power of the Fixe Threshol switching (figure ). By virtue of its esign, high for fixe threshol algorithm, nee to be necessarily chosen to be less than the value of obtaine from equation (9) for the lightest vehicle. Hence this algorithm is always more efficient at harvesting vibration energy than the simple fixe threshol algorithm escribe in the previous section. C. Switche Inuctor This section proposes a thir algorithm that woul further enhance. This algorithm uses a circuit shown in figure 5. The new circuits uses an inuctor ( L ) an switch SW P in aition to the components shown in figure. The voltage rop across L is given by. SW L P is turne on when reaches a imum an SW L is turne on when reaches a imum. The switches SW P an SW L are turne off when the respective voltages + an L rops below an off-threshol low As iscusse in the previous section, the occurrence of imum can be etermine using analog electronics. The control law for SW P is given by state transition iagram shown in figure an the control law for SW L is given by state transition iagram state transition iagram shown in figure 7. 3

21 C s Figure 5: Energy Harvesting with Inuctor is imum control = = control L + or L + < low >= low L + < low Figure : State Transition Diagram for SW L ( Switche Inuctor ) is imum control = = control or < low < low >= low Figure 7: State Transition Diagram for SW P ( Switche Inuctor )

22 If the isplacement of the beam has only one extremum value, then can be given by where mc + C = ( ) ( π ( ζ ) m = + exp ζ, m C ζ = R L C p C = C + C p s if otherwise > () By comparing equations () with (9), we notice that has increase by a factor ( m ), which equals the peak overshoot of the LCR circuit. The Switche Inuctor algorithm yiels a higher ue to the presence of the inuctor. In the absence of the ioe brige rectifier, the secon ynamics of the LCR circuit will exhibit an oscillatory behavior for an extremely small time perio t an the ynamics woul eventually converge to its steay state value. The brige rectifier in the circuit woul however clamp to the overshoot voltage, resulting in higher available voltage. Equation () can be erive as follows. When SW P is close, the circuit is similar to the circuit in section III B, an the effective voltage riving the resistive an inuctive components of the circuit is given by effective () t = sign( () t C () t ) ( () t () t () t,) C (5) In the absence of SW P, the overall ynamics is ominate by the mechanical system an woul be given by equation (9). There woul be no gain in. If SW P is close at some t = t when =, it woul be result in a step input to the electrical circuit. If ( t ) >, the ioe brige will begin to conuct when SW P is close. i p L t + R i = ( ) () p effective t 5

23 When i p is uniirectional, the electrical ynamics can be written in terms of Q = i p t. Substituting Q = i p t an Q C = i pt = C C (7) == i C t p = Q (8) an noting that sign( ) = sign( Q) written as LQ&& + R Q& + = ( C + ) sign( ), the piece-wise linear ynamics of the LCR system can be Q (9) Now i p = Q& an i p is uniirectional up to the first imum of Q. Since equation (9) is vali when i p is uniirectional, it can be use to etermine this first imum Q. For this secon orer system, Q C m = () Closing SW P when is imum, woul result in step voltage input the electrical LCR circuit (equal to ). Thus mc + C = ( ) if otherwise > () Simulation results using the switche inuctor controller are shown in the following figures. When compare to the Max voltage switch controller, it is seen from figure an figure 8 that the available voltage has increase by a factor of over.5, an from figure an figure, it is seen that peak power is increase by a factor of. Using equation (3) to (3), a value of = 9.5 is obtaine for the first pulse. The estimate value for is seen to be in close agreement with the simulation result from figure 8.

24 oltage () time (sec) Figure 8: oltage Output for Switche Inuctor Algorithm Current (ma) time (sec) Figure 9: Loa Current for K Loa with Switche Inuctor Algorithm Power consume at Loa (mw) time (sec) Figure : Instantaneous Power Consume by K Loa with Switche Inuctor 7

25 D. Effect of MOSFET on Switche Inuctor In estimating the value of, section III.C assumes ieal switches an neglects the voltage rop across SW P. The practical realization of the switch involves an analog electronics circuit with a mosfet. This circuit woul uses the voltage store in the storage capacitor to controls a mosfet as shown in figure. In the off state, the mosfet offers nearly infinite resistance an accurately moels an ieal open switch. When the mosfet is turne on using a control signal, it offers a finite voltage rop which is a function of DS. The voltage rop in turn affect the DS electrical ynamics an the over all electrical ynamics is no longer linear. Since there are no stanar solutions for this nonlinear problem, a numerical solution framework has been presente in this sub-section. This solution can be use in conjunction with section III.C to estimate the performance of the Switche Inuctor Algorithm with greater accuracy. For this numerical moel, the iniviual ioes in the brige rectifier as well as the mosfet are moele using the stanar nonlinear equation ([7]). I = + I ioer if GS < T k[ ( GS T ) DS DS ], = if GS > T an k( GS T ) if GS > T an ioe ioe ln + I S I D ioe DS DS < > ( ) GS ( ) GS T T () (3) where the subscripts refers to the source-gate voltage GS refers to the source-rain voltage DS is the mosfet threshol voltage T is a mosfet constant k Clearly = + () GS DS Hence = [ + I K ] ( ) + ( ) if > / ( I K ) ( ) DS D T T T T otherwise D / (5) 8

26 The parameters for the moel that were obtaine from the ata sheets of the electronic components are given below. 5 I S = = R D =. Ω T =. K = 9.85 L = mh R =. 8 Ω -3 A R is the resistance of the inuctor which is ae in series. figure an figure 3 compare the experimental response of the system to the theoretical of the moel of the system for equivalent parameters of = μf an = 5 C p L C Piezo G Piezo Electrical moel of Piezo Crystal Dioe Brige Piezo Crystal ip sw P G, S an D respectively enote the gate source an rain of the mosfet C s S D Figure : Mosfet in Max Switch Simluate storage capacitor voltage () Time(ms) Figure : Simulation Results 9

27 Using this moel, the theoretical voltage for the switche inuctor is plotte versus the theoretical response of the switch for the sensor in figure for an initial storage capacitor voltage of.5. Storage capacitor voltage () 8 experiment # experiment # Time(ms) Figure 3: Results from Two Sets of Experiments oltage gain for the switche Inuctor () oltage gain with switch () Figure : Theoretical oltage Obtaine from Switche Inuctor vs. the Theoretical oltage Obtaine from Max Switching Algorithm The theoretical moel oes not consier the equivalent series resistance of the electrolytic storage capacitor. Hence the theoretical preiction shown in figure woul give an upper boun for the voltage from the switche inuctor algorithm.

28 E. Comparison of Control Algorithms The table below shows a comparison on the imum voltage across the storage capacitor for the three algorithms. Algorithm Fixe Threshol Switching high C Max oltage Switching ( ) + C mc + C Switche Inuctor ( ) As note earlier, for the Fixe Threshol Switching algorithm to work reliably, high must necessarily be chosen a volt or two lower than the lowest obtaine from equation (9) corresponing to the smallest that can be expecte. Also, from equation, it is clear that m >. Thus, for Fixe Threshol is smaller than for Max oltage Switching which in turn is smaller than for the Switche Inuctor algorithm.

29 I. EXPERIMENTAL RESULTS A passenger sean car was riven over the sensor at Kmph. This resulte in separate loaing from the two axle, first by the front tire an then by the rear tire. For the purpose of experiments, ramps were constructe to enable the test vehicles to be riven smoothly over the sensor. Since the imum range of our ata-acquisition system was limite to was altere to limit the imum voltage. All the experiments were performe using the same setup with only the electronic circuit being moifie to realize the three control algorithms. ±, the sensor configuration Figure 5 show the results from the testing of the Fixe threshol algorithm. The on-threshol for the algorithm was chosen at.75 so that the sensor woul etect light vehicle such as motorcycles. The mosfet use in the switching circuit, con the off-threshol to.75. As a result, SW L turns on when the capacitor voltage reaches.75 an turns off when falls below.75. The electronic circuits were moifie so that SW L turns on when reaches a imum. In orer to etect a global imum, an to collect energy from both axles, SW L was moifie to turn on when oes not increase for a perio of ms. The off-threshol was once again chosen to be.75. Figure show the results from Max Switching algorithm. It is seen from figure, that the first axle prouces.5 an the secon axle results in an aitional.5. The theoretical moel evelope in section III.D, estimates when the Switche Inuctor Algorithm is use in the place of the Max Switching algorithm. The theoretical estimates that the first axle woul prouce a voltage of.5 an the secon axle woul prouce. It preicts an upper boun of.5 with the Switche Inuctor Algorithm. Figure 7 show the results from the Switche Inuctor Algorithm experiments an the capacitor voltage is foun to be Storage capacitor voltage () Time(s) Figure 5: Fixe Threshol Algorithms with a Threshol of.75

30 9 8 Storage capacitor voltage () Time(s) Figure : Max Switching Algorithm 9 8 Storage capacitor voltage () Time(s) Figure 7: Switche Inuctor Algorithm In orer to compare the energy amount of energy harveste, the loa switch was isable an the open circuit voltages generate at the storage capacitor were recore. Figure 8 an figure 9 respectively show the voltage generate by the Max oltage Switching an Switche Inuctor. It is apparent that if SW P is controlle as prescribe, the Switche Inuctor offers significant improvement. 3

31 9 8 Storage capacitor voltage () Time(s) Figure 8: Max Switching 9 8 Storage capacitor voltage () Time(s) Figure 9: Switche Inuctor Figure 8 shows the influence of vehicle weight on the voltage across the storage capacitor. The figures on the left correspon to a passenger sean while the figures on the right correspon to a motorcycle passing over the sensor. It can be clearly seen that the storage voltage for the motorcycle is smaller than the storage voltage for the car. Further results also showe that the storage voltage for a mini-van was higher than the storage voltage for the passenger sean an reache a value as high as 7 volts. This emonstrates that the sensor is capable of roughly measuring vehicle weight. Since the transmission pulse with is proportional to the storage capacitor voltage, a rough measure of the vehicle weight can be obtaine by measuring the timewith of the receive wireless transmission pulse.

32 Storage capacitor voltage () Storage capacitor voltage () Time(s) 8 Time(s) Storage capacitor voltage () Storage capacitor voltage () Time(s) Time(s) Figure 3: Capacitor oltage Depenent on ehicle Weight (Left: Car Driven Close to the mph; Right: Motorcycle Driven Close to the 5

33 . TRAFFIC SENSOR NETWORKS A. Introuction The feasibility of a single battery-less wireless traffic sensor has been experimentally emonstrate. The current sensor will transmit an RF pulse per axle to a eicate receiving station upon the arrival of an automobile. In the presence of multiple sensors foun in sensor networks, such as at an intersection or on a highway with many lanes, the sensor shoul sign the transmission so that the receiver will be able to ientify the source of transmission. This section proposes a solution for such a sensor signature. B. Encoer Decoer Outline The propose solution involves assigning a signature to each sensor in the form of a unique ientification number. When a vehicle passes over the sensor, it woul harvest energy from the mechanical vibrations. An encoer is use to encoe this ientification number to a sequence of s an s. A transmitter moulates an RF carrier wave to transmit this sequence of s an s to a receiver. Upon receiving this signal, the receiver emoulates the RF signal an recreates the series of s an s at its output. A ecoer is use to convert this sequence back to the ientification numbers number that was transmitte. The ecoe signature is use to ientify the sensor over which the automobile has passe. The schematic of the circuit to implement this scheme is shown in figure 9.

34 sw L Piezo Crystal Dioe Brige C S 8 5 IN SHDN MAX GND SENSE OUT SET 5 IN D D ENCODER 3, 7 & 8 Data_Out GND 7 IN DATA 5 ANT LYNX TMX- 8LC LADJ/GND GND GND 3 GND 8 GND Figure 3: Circuit Schematic Power Calculations For the receiving station to reliably ecoe the transmitte signature, a minimum transmission of ms is require an the supply voltage nees to be maintaine at atleast.5. Hence the system shall be esigne to transmit for 5 ms. From our experiments, the total current require by the new circuit was calculate to be ma. For a μf storage capacitor that is being use, this correspons to a voltage change of 5. Since the minimum acceptable voltage for the storage capacitor is.5, enough energy must be harveste from the vibrations to charge the storage capacitor to.5-8. This is expecte to be possible with the switche inuctor control system. C. Experimental Results The circuit shown in figure 9 was implemente an the traffic sensor was excite by a short uration loa. For the first experiment, the encoer was configure to transmit a for bit (D ) an a for bit (D ). Figure 3 shows the storage capacitor voltage for the first experiment. From figure 3, showing the outputs of bit (D ) an bit (D ) of the ecoer, it is seen that the sensor ientification number can be successfully ecoe. For the secon experiment, the encoer was configure to transmit a for bit (D ) an a for bit (D ). Figure 3 shows the storage capacitor voltage for the first experiment. From figure 33, showing the outputs of bit 7

35 (D ) an bit (D ) of the ecoer, it is seen that the sensor ientification number can be successfully ecoe. 7 Storage capacitor voltage () Time(s) Figure 3: Storage Capacitor oltage ( st Configuration of Encoer) 8

36 3.5 D 3 D.5 Decoer Outputs () Time (s) Figure 33: Decoer Output ( st Configuration of Encoer) 7 Storage capacitor voltage () Time(s) Figure 3: Storage Capacitor oltage ( n Configuration of Encoer) 9

37 3.5 D 3 D.5 Decoer Outputs () Time (s) Figure 35: Decoer Output ( n Configuration of Encoer) 3

38 I. COMPLETION OF PROJECT TASKS The tasks in the project that summarize the major activities which were accomplishe are as follows: ) To evelop robust roaworthy electronic harware that can be use for evaluation of all the sensors evelope in this project. This will be one by fixing problems with the current ata acquisition harware, eveloping a new PCB that can accommoate ifferent control systems for energy harvesting an finally by carrying out vehicle tests to verify the working of the new measurement harware. ) To evelop an simulate key control algorithms to imize energy harvesting from electric system. The best algorithm will then be selecte an use for the remaining relate tasks in this project. 3) To evelop electric sensor esign to enable weigh-in-motion in aition to vehicle counting, axle counting an spee measurements (Full sensor). ) To evaluate the use of surface acoustic wave (SAW) evices to enable battery-less wireless operation of the roaway embee sensors. 5) To evelop electric sensor esign with minimal size an optimal ease of installation to enable vehicle counting, axle counting an spee measurements (Basic sensor). ) To conuct experimental vehicle tests to evaluate performance, reliability an accuracy of full sensor. 7) To conuct experimental vehicle tests to evaluate performance, reliability an accuracy of basic sensor. 8) To evelop an evaluate a sensor network in orer to enable short term traffic sensor applications such as turn analysis at rural intersections. A single micro-processor capable of simultaneously receiving an analyzing signals from multiple sensors will be evelope. 9) To write a comprehensive project report escribing the technology evelope, results obtaine an conclusions reache from the work complete in the project. All of the above tasks have been complete. The previous chapters provie etails of the esign, technical analysis, simulation results an experimental test ata. Two important issues nee to be escribe with respect to the list of tasks above: A. Full Sensor s Basic Sensor When the project proposal was written, the PI envisione two ifferent types of sensors: ) A full sensor that woul be capable of measuring traffic flow rate, number of axles on vehicle an approximate vehicle weight. ) A basic sensor that woul be capable of measuring only traffic flow rate an woul be more compact an size-optimal than the full sensor. Analysis uring the project showe that the sensor must be feet long in orer to ensure that one set of wheels of every passing vehicle woul travel over the sensor. It turne out that the wheels of the passing vehicle woul have to travel irectly over the sensor in orer to be able to generate enough energy to power the electronics for wireless transmission. Hence the optimal sensor size 3

39 that coul be obtaine woul be a sensor that is feet long with minimal height an minimal with. This is the final size of the sensor as evelope an teste in this project. While this sensor is a basic sensor in terms of its size optimality, it is also capable of measuring not only traffic flow rate but also approximate vehicle weight an number of axles on vehicle. Hence the istinction between a full sensor an a basic sensor coul not really be mae in the project. Figures an in this report coul be consiere to represent the original full sensor an basic sensor respectively. However, the basic sensor is also able to measure the approximate vehicle weight an therefore provies the same functionality as the full sensor. B. Use of SAW Device for Battery-Less Wireless Operation A surface acoustic wave (SAW) evice was consiere as an alternate metho of obtaining battery-less wireless operation. SAW evices were fabricate an the interrogator electronics for interfacing with the SAW evice were evelope. Battery-less wireless operation with the SAW evice was teste. However, it was eventually abanone ue to the fact that we coul only obtain wireless telemetry istances of the orer of a few feet using this technology. On the other han, the metho of harvesting energy from the sensor vibrations using electric elements an power electronics an using this energy to power a wireless transmitter was able to provie telemetry istances of over a hunre feet. Hence the clear superiority of this approach le us to abanon the surface acoustic wave approach. 3

40 II. CONCLUSIONS This project evelope a battery-less wireless traffic sensor for measurement of traffic flow rate, number of vehicle axles an approximate vehicle weight. To the best of our knowlege, this is a unique invention an is the first ever presentation of a traffic flow sensor that obtains energy for its operation entirely by harvesting vibration energy from the passing of a vehicle over the sensor. The evelope sensor is feet long, about an inch in with an inches in height. It can be place in a rectangular slot mae in the roaway. This particular project focuse on esign of the sensor, evelopment of control algorithms to imize the energy harveste from the vibrations, simulation stuies an experimental tests to evaluate the performance of the sensor. Experimental results showe that the sensor was able to harvest aequate energy for its operation from the passing of every axle over the sensor. The sensor coul reliably an accurately measure traffic flow rate. The with of the wireless transmission pulse from the sensor was roughly proportional to the weight of the vehicle passing over the sensor. Thus the vehicle weight coul also be approximately measure. Compare to existing inuctive loop etectors, the evelope new sensors have the following avantages: ) Installation: To install a loop etector an calibrate it, it is sometimes necessary to shut own traffic on the roa for as much as ays. The new sensors can be installe by rilling a slot across the lane in the roa surface of inch with an inches epth. Most importantly, no wiring is neee from the traffic lane to the roasie ata acquisition unit. It is expecte that the installation will only take a few minutes. ) Energy consumption: The sensors on the roaway require no external power supply while inuctive loop etectors have to be continuously powere all the time, even uring the night when traffic flow might be really low. 3) Cost: A -channel loop etector package (for example, from Eberle Design, Inc or Reno A&E) on average has a harware cost of aroun $7. The installe cost of each loop is typically aroun $5. The propose sensors on the other han are expecte to cost only $5 - $ each. ) Aitional ariables: The new sensors can measure number of axles an vehicle length, in aition to traffic flow rate. Thus they can be use for vehicle classification. With some further evelopment, the sensors can likely also be use for weigh-in-motion. Future work will focus on further enhancing the telemetry istance of the sensor so that it can irectly transmit wireless reaings to a central metro location, improving the weigh-in-motion ability of the sensor an fiel testing of the sensor by placing it in a pavement location where it can be repeately teste with known traffic flow rates an known vehicle loas. 33

41 REFERENCES [] C. Chen, J. Kwon, J. Rice, A. Skabaronis an P. araiya, Detecting errors an imputing missing ata for single-loop surveillance systems, Transportation Research Recor, pp.-7, 3. [] S. Chen, Z. Sun an B. Brige, Traffic Monitoring Using Digital Soun Fiel Mapping, IEEE Transactions on ehicular Technology, vol. 5, no., pp ,. [3] S. Chen, Z. P. Sun an B. Brige, Automatic traffic monitoring by intelligent soun etection, IEEE Conference on Intelligent Transportation Systems, pp.7-7, 997. [] B. Coifman, Improve elocity Estimation Using Single Loop Detectors, Transportation Research Part A: Policy an Practice, vol. 35, no., pp.83-88,. [5] C. W. De Silva, Mechatronics: An Integrate Approach, CRC Press,. [] N. Ferrier, S. Rowe an A. Blake, Real-time traffic monitoring, Proceeings of the Secon IEEE Workshop on Applications of Computer ision, pp.8-88, 99. [7] W. E. L. Grimson, C. Stauffer, R. Romano an L. Lee, Using aaptive tracking to classify an monitor activities in a site, Proceeings of the IEEE Computer Society Conference on ision an Pattern Recognition, pp.-9, 998. [8] P. Horowitz an W. Hill, The Art of Electronics, Cambrige University Press, 989. [9] S. Kamijo, Y. Matsushita, K. Ikeuchi an M. Sakauchi, Traffic monitoring an accient etection at intersections, IEEE Transactions on Intelligent Transportation Systems, vol., no., pp.8-8,. [] H. K. Khalil, Nonlinear Systems, Upper Sale River, New Jersey, Prentice Hall,. [] S. W. Kim, Y. Eun, H. Kim, J. Ko, W. J. Jung, Y. Choi, Y. G. Cho an D. Cho, Performance comparison of loop/ an ultrasonic sensor-base traffic etection systems for collecting iniviual vehicle information, Proceeings of the th Worl Congress on ITS (CD-ROM), 998. [] S. W. Kim, K. Kim, J. Lee an D. D. Cho, Application Of Fuzzy Logic to ehicle Classification Algorithm In Loop/Piezo-Sensor Fusion Systems, Asian Journal of Control, vol. 3, no., pp.-8,. [3] P.. Kokotovic, J. O'Reilly an H. K. Khalil, Singular Perturbation Methos in Control: Analysis an Design, Society for Inustrial an Applie Mathematics, Philaelphia, 98. [] C. Li, K. Ikeuchi an M. Sakauchi, Acquisition of Traffic Information Using ieo Camera with D Spatio-Temporal Image Transformation Technique, IEEE Transactions on Intelligent Transportation, pp.3-38, 999. [5] E. Lefeuvre, A. Bael, C. Richar, D. Guyomar, Piezoelectric Energy Harvesting Device Optimization by Synchronous Electric Charge Extraction, Journal of Intelligent Materials, Systems an Structures, vol., no., pp.85-87, 5. [] L.J. Mountain, J. B. Garner., Application of photography to traffic surveys, The Journal of the Institute of Highway Engineers, pp.-9, 98. [7] P. T. Martin, Y. Feng an X. Wang, Detector Technology Evaluation, Mountain-Plains Consortium, 3. [8] P. G. Michalopoulos an C. A. Anerson, Costs an benefits of vision-base, wie-area etection in freeway applications, Transportation Research Recor, pp.-7, 995. [9] O. D. Nwokah, Mechanical Systems Design Hanbook Moeling Measurement an Control, CRC Press,. 3