Design and analysis of an infrared range sensor system for floor-state estimation

Similar documents
The Investigation of the Obstacle Avoidance for Mobile Robot Based on the Multi Sensor Information Fusion technology

An Improved Weighted Centroid Localization Algorithm

A Robust Adaptive Carrier Frequency Offset Estimation Algorithm for OFDM

Proceedings of the 2007 IEEE International Conference on Signal Processing and Communications (ICSPC 2007)

arxiv: v1 [cs.ro] 25 Apr 2018

DISCUSSION ON THE APPLICATION OF WIRELESS ACTIVE SENSING UNIT

A Novel Multimode Mobile Robot with Adaptable Wheel Geometry for Maneuverability Improvement

Modified Takagi-Sugeno Fuzzy Logic Based Controllers for a Static Compensator in a Multimachine Power System

FUSION OF DATA FROM INERTIAL SENSORS, RASTER MAPS AND GPS FOR ESTIMATION OF PEDESTRIAN GEOGRAPHIC LOCATION IN URBAN TERRAIN

Networked Estimation for Event-Based Sampling Systems with Packet Dropouts

Design and Simulation of PD and PID Controller for Hybrid Actuator

IEE Electronics Letters, vol 34, no 17, August 1998, pp ESTIMATING STARTING POINT OF CONDUCTION OF CMOS GATES

2005 Journal of Software. . Ad hoc ), ) A Delay Oriented Adaptive Routing Protocol for Mobile Ad hoc Networks

PSO Algorithm-Based Optimal Tuning of PSS for Damping Improvement of Power Systems

RECOMMENDATION ITU-R P

A Condition-based Preventive Maintenance Policy for Markov Deteriorating Systems

Sliding Mode Controller with RBF Neural Network for Manipulator Trajectory Tracking

100 mm. 200 mm. 300 mm. 500 mm. Wood (rough) Cardboard Black DBT DBT DBT DBT DBT - 1 K DBT K CONNECTION DIAGRAMS.

Developing a Gesture Based Remote Human-Robot Interaction System Using Kinect

Using Bayesian Game Model for Intrusion Detection in Wireless Ad Hoc Networks

The Research and Realization of A Localization Algorithm in WSN Based on Multidimensional Scaling Li Xiang, Qianzhi Hong, Liuxiao Hui

Use of ANFIS Control Approach for SSSC based Damping Controllers Applied in a Two-area Power System

Beam quality measurements with Shack-Hartmann wavefront sensor and M2-sensor: comparison of two methods

Improved Cuckoo Search Algorithm with Novel Searching Mechanism for Solving Unconstrained Function Optimization Problem

A New Type of Weighted DV-Hop Algorithm Based on Correction Factor in WSNs

Open Access An Improved Algorithm of Spatial Difference MUSIC for Direction of Arrival Estimation in a Smart Antenna

GNSS Integrity for Railway Transportation

Application of Intelligent Voltage Control System to Korean Power Systems

熊本大学学術リポジトリ. Kumamoto University Repositor

HUMANOID robots continue to be an active topic for

Uncertainty in measurements of power and energy on power networks

Low Switching Frequency Active Harmonic Elimination in Multilevel Converters with Unequal DC Voltages

This is a repository copy of Online Control of IPMSM Drives for Traction Applications Considering Machine Parameter and Inverter Nonlinearities.

Priority based Dynamic Multiple Robot Path Planning

Design of Teaching Platform Based on Information Detection System

Adaptive System Control with PID Neural Networks

Energy Efficiency Resource Allocation for Device-to-Device. Communication Underlaying Cellular Networks

Efficient Large Integers Arithmetic by Adopting Squaring and Complement Recoding Techniques

Customer witness testing guide

Research on detection system of heat faults based on multi-sensor information fusion

Multirate Anypath Routing in Wireless Mesh Networks

Research of Dispatching Method in Elevator Group Control System Based on Fuzzy Neural Network. Yufeng Dai a, Yun Du b

NOVEL BAND-REJECT FILTER DESIGN USING MULTILAYER BRAGG MIRROR AT 1550 NM

Prediction of Critical Submergence for Horizontal Intakes

Moving Target Tracking Based on CamShift Approach and Kalman Filter

On Sensor Fusion in the Presence of Packet-dropping Communication Channels

Towards a Rapid Control Prototyping Toolbox for the Stellaris LM3S8000 Microcontrollers

Time-frequency Analysis Based State Diagnosis of Transformers Windings under the Short-Circuit Shock

Kinematics of a dedicated 6DOF Robot for Tele-echography

The Dynamic Utilization of Substation Measurements to Maintain Power System Observability

Prediction-based Interacting Multiple Model Estimation Algorithm for Target Tracking with Large Sampling Periods

Comparison of Field-Oriented Control and Direct Torque Control for Permanent Magnet Synchronous Motor (PMSM)

Sliding-Mode Control of Synchronous Reluctance Motor

A Pervasive Indoor-Outdoor Positioning System

POWER QUALITY IMPROVEMENT USING SELF-TUNING PI CONTROLLED STATCOM BY PARTICLE SWARM OPTIMIZATION

Optimal Placement of PMU and RTU by Hybrid Genetic Algorithm and Simulated Annealing for Multiarea Power System State Estimation

A RF Source Localization and Tracking System

High Speed, Low Power And Area Efficient Carry-Select Adder

Fuzzy Gain Scheduling of PID Controller for a MIMO Process

TORQUE RIPPLE MINIMIZATION, SUPPRESS HARMONICS, AND NOISE OF BRUSHLESS PM SYNCHRONOUS MOTORS DERIVED BY FIELD ORIENTED CONTROL

Experiments in Outdoor Operation of RatSLAM

Multi-sensor optimal information fusion Kalman filter with mobile agents in ring sensor networks

Image analysis using modulated light sources Feng Xiao a*, Jeffrey M. DiCarlo b, Peter B. Catrysse b, Brian A. Wandell a

A Novel Impedance Measurement Technique for Power Electronic Systems

A Simple Satellite Exclusion Algorithm for Advanced RAIM

Calculation of the received voltage due to the radiation from multiple co-frequency sources

PID Controller tuning and implementation aspects for building thermal control

Dynamic Optimization. Assignment 1. Sasanka Nagavalli January 29, 2013 Robotics Institute Carnegie Mellon University

Instruction Sheet AMPMODU* MTE CONNECTORS Mar 11 Rev A

Target Tracking and Obstacle Avoidance for Mobile Robot based on Kinect

An Adaptive Energy-Conserving Strategy for Parallel Disk Systems

Inverter topology comparison for remedial solution in transistor faulty case

Open Access Research on Piecewise Linear Fitting Method Based on Least Square Method in 3D Space Points

Topic-weak-correlated Latent Dirichlet Allocation 1

Evaluate the Effective of Annular Aperture on the OTF for Fractal Optical Modulator

Wi-Fi Indoor Location Based on RSS Hyper-Planes Method

THE INTERNET-BASED TELEOPERATION: MOTION AND FORCE PREDICTIONS USING THE PARTICLE FILTER METHOD

An Online Q-learning Based Multi-Agent LFC for a MultiArea Multi-Source Power System Including Distributed

A RELIABLE SEMI-DISTRIBUTED LOAD BALANCING ARCHITECTURE OF HETEROGENEOUS WIRELESS NETWORKS

Multichannel Frequency Comparator VCH-315. User Guide

Effectiveness and Accuracy of Wireless Positioning Systems

Mooring Cost Sensitivity Study Based on Cost-Optimum Mooring Design

Parameter Free Iterative Decoding Metrics for Non-Coherent Orthogonal Modulation

To: Professor Avitabile Date: February 4, 2003 From: Mechanical Student Subject: Experiment #1 Numerical Methods Using Excel

Topology Control for C-RAN Architecture Based on Complex Network

Online Reporting. Online Reporting. A step-by-step guide. Important information for churches, schools and organisations

FPGA Implementation of Ultrasonic S-Scan Coordinate Conversion Based on Radix-4 CORDIC Algorithm

ANNUAL OF NAVIGATION 11/2006

Kalman Filter based Dead Reckoning Algorithm for Minimizing Network Traffic between Mobile Nodes in Wireless GRID

A path-specific propagation prediction method for point-to-area terrestrial services in the VHF and UHF bands

Behavior-Based Autonomous Robot Navigation on Challenging Terrain: A Dual Fuzzy Logic Approach

The Performance Improvement of BASK System for Giga-Bit MODEM Using the Fuzzy System

Fuzzy PID Based Trajectory Tracking Control of Mobile Robot and its Simulation in Simulink

PRO- CRIMPER* III Hand Crimping

Consistent cooperative localization

Performance Analysis of Multi User MIMO System with Block-Diagonalization Precoding Scheme

Ensemble Evolution of Checkers Players with Knowledge of Opening, Middle and Endgame

Combined Independent Component Analysis and Kalman Filter Based Real-Time Digital Video Stabilization

Optimal State Prediction for Feedback-Based QoS Adaptations

Insertion/Extraction Tool and Replacement Tip Kits [ ]

Transcription:

Journal of Mechancal Scence an Technology 5 (4) (011) 1043~1050 www.sprngerlnk.com/content/1738-494x DOI 10.1007/s106-011-0141-5 Desgn an analyss of an nfrare range sensor system for floor-state estmaton Mnyoung Lee an Sooyong Lee * Department of Mechancal an System Desgn Engneerng, Hongk Unversty, Seoul 11-791, Korea (Manuscrpt Receve Aprl 16, 010; Revse October 3, 010; Accepte November 19, 010) ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Abstract Most bln an vson-mpare people use a whte cane or a gue og to a ther walkng. The conventonal whte cane helps the user to perceve the envronment smply by entfyng the presence of nearby obstacles. In the present stuy, three nfrare range sensors were use to entfy the floor state. The sensor system, affxe to the user s belt, oes not requre swngng moton or any other movement by the user. The three sensors are confgure raally, an the esgn parameters are selecte base on the sensor characterstcs. The man fference from the conventonal range sensor system for the bln s that the nfrare range sensor system entfes three screte states (even surface, ascenng star, escenng star). The conton for the state transton was erve from a sensor system moel. Atonally, the effects of sensor system movement were verfe by covarance analyss. Keywors: Covarance analyss; Floor sensng; Range sensor; State estmaton ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- 1. Introucton Gue-og assstance to the bln s very effectve but not very common. Instea, bln people usually rely on whte canes, whch smply help them to perceve, by touch, whether the path n front of them s open or obstructe. Usng a cane, the user also can perceve the conton of ther walkway. Impact, bouncng, stffness an even soun are mportant ncators to the user, as combne wth hs personal experence. Use of a cane for etectng obstacles an percevng general path contons, however, s lmte to the area wthn the cane s reach. Use of sensors greatly extens the range of obstacle etecton. Optcal range sensors such as nfrare sensors an laser sensors prove accurate measurement, but mght not be able to etect glass. Laser sensors are expensve, an safety emans that they be hanle wth cauton. Optcal sensors, atonally, are not yet suffcently robust to envronmental lght conton changes. Ultrasonc sensors alternatvely, mght not be as accurate as nfrare range sensors, but they have a longer measurement range. Also, ther beam angle s much wer than that of optcal sensors, whch characterstc s sutable for canes because the measurement tells the range nformaton to the nearest obstacle n the sonc cone. On the other han, ultrasonc sensors, ue to ther we beam angle, Ths paper was recommene for publcaton n revse form by Assocate Etor Yang Sh * Corresponng author. Tel.: +8 30 1609, Fax.: +8 3 7003 E-mal aress: sooyong@hongk.ac.kr KSME & Sprnger 011 cannot fferentate object shapes. Ths makes the task of fferentatng between an obstacle an, for example, ascenng stars, ffcult. Determnng whether a surface s even or not, whch stncton can be very crtcal nformaton to bln peestrans, also s problematc. Ref. [1] evelope a han-hel envronment scovery tool that ncorporates a laser-base range sensor. As the user swngs the system aroun, he receves local range nformaton. The tme profle of the range s analyze to etect envronmental features that are crtcal to moblty, such as curbs, steps an rop-offs. An extene Kalman flter s use to track the range ata an etect envronmental features of nterest. Ref. [] presente an onlne surface-trackng algorthm base on a Jump-Markov moel for automatc etecton of geometrc sngulartes. The algorthm escrbes the evoluton of range measurements n a pecewse planar worl moel. Ref. [3] aresse relable feature extracton from contnuously scannng outoor-operatng range sensors. An algorthm etects features on-lne as soon as the pertnent range has been sense. The etecton process moel aapts ts valaton regon accorng to the spatal graent of the surface beng sense, an the ata s mplemente n extene-kalmanflter-recursve form. Ref. [4] ntrouce the new concept of an electronc cane base on a combnaton of several sensors. Dfferent types of electronc travel as have been evelope over the past thrty years. Commercally avalable as nclue the Senero Group's MnGue [5], the Nuron Laser Cane [6] to a the etecton of rop-offs an overhea obstacles,

1044 M. Lee an S. Lee / Journal of Mechancal Scence an Technology 5 (4) (011) 1043~1050 1000 900 SENSOR OUTPUT MODEL 800 700 A/D Value 600 500 400 300 00 Fg. 4. Use of nfrare range sensor system. 100 50 100 150 00 50 300 350 400 450 500 RANGE [cm] Fg. 1. Infrare range sensor characterstcs. concurrent measurements. By analyzng the three range values, states of the floor can be successfully fferentate.. Sensor system moel Fg.. Infrare range sensor system (Desgn). Fg. 3. Infrare range sensor system (Photo). an the Sonc Pathfner [7], a heaban-mounte ultrasonc sonar evce that proves nformaton on obstacles n the mmeate surrounngs, n the form of notes on a major muscal scale. Ref. [8] ntrouce the three nfrare range sensor system for etectng ascenng an escenng stars. The ecsons of the system are mae base on current sensor reangs; however, sturbance ue to the user s movement was not consere. In the present stuy, a three nfrare range sensor system was use to sense the surfaces of forthcomng ascenng an/or escenng stars. The three sensors are nstalle raally, takng avantage of nfrare s narrow beam, to prove Ths work employe GPY0A710K (Sharp) sensors. Ths type can measure ranges from 80cm to 500cm, though the resoluton eterorates as the range lengthens, as shown n Fg. 1. Fgs. an 3 show the three-sensor raal arrangement. The user receves the floor-state estmaton results from the wast-belt-affxe sensor system, as llustrate n Fg. 4. Ths system nforms the user of the floor state va ether a synthesze voce or a haptc evce [9]. The beam wth of the nfrare range sensors s narrow enough that the sensors o not nterfere wth each others. Fg. 5(a) shows the sensor system confguraton. The lower nfrare sensor (S1) s nstalle as orentate at the angle of wth respect to the gravty recton. The center (S) an upper (S3) sensors are nstalle as orente at the angles 1 an 3, separate raally wth each postone at stance c from the center. The sensor system s fxe at the heght of h. As t s affxe to the user s belt, the value of h vares epenng on the leg length of the user. For the purposes of our experment, h was set at 100 [cm]. Each sensor proves range nformaton, 1, an 3, from the tp of the sensor to the obstacle (usually the floor surface n our case, because the sensor s artculate ownwar). The sensor reangs vary accorng to the floor state, as shown n Fgs. 5(b)-(e). Raw sensor values for three walkng stuatons are presente n the followng three fgures. Fgure 6 shows the sensor output when the user walks on an even surface an then stops. The three sensor outputs vary whle the user walks, but they reman synchronze, an the fference between two sensors remans almost constant. Note that the output from the top sensor (S3 n the fgure) vares more than those from the other two, because the angle of ncence to the floor s much larger than for the sensors at the bottom an the center, an prove less accurate measurement. Fg. 7 shows the sensor output when the user walks on an even surface, then steps up an ascenng star, an fnally comes to an even surface agan. The sensor reangs, here, clearly show the fference from the even-floor case.

M. Lee an S. Lee / Journal of Mechancal Scence an Technology 5 (4) (011) 1043~1050 1045 (a) Knematc moel of sensor system (b) Ascenng star 1 (c) Ascenng star () Descenng star 1 Fg. 5. Sensor system an envronment. (e) Descenng star Fg. 6. Range sensor raw ata - even surface. Fg. 7. Range sensor raw ata ascenng stars.

1046 M. Lee an S. Lee / Journal of Mechancal Scence an Technology 5 (4) (011) 1043~1050 h 1 3 3 1 Fg. 10. Smplfe sensor system moel. 1 : 1 st sensor reang : n sensor reang : 3 r sensor reang. 3 Fg. 8. Range sensor raw ata escenng stars. We represent the change of the sensor system heght h ue to the -th sensor reang change. From the moel, we erve h = cos 1 1 (1) h = cos( + ) 1. () Fg. 9. State transton. Fg. 8 shows the sensor output when the user approaches a escenng star from an even surface, then steps own the escenng star an contnues to walk on an even surface agan. It s clear that the sensor reangs are stnct for fferent floor states. The rules for entfcaton of the floor state are evelope n the followng secton. 3. Analyss The sensor system moel was evelope for etermnaton of floor states from sensor nformaton. The floor-state transton s efne as n Fg. 9. Note that the state transton between the ascenng state an the escenng state was not consere. Specfcally, the floor-state change s etermne on the bass of the sensor reang change. Therefore, we moel the system accorng to the sensor value change, not the current value. The avantage of usng the state transton s the utlzaton of prevous state nformaton. For the purposes of the analyss, the sensor moel represente n Fg. 10 below s a smplfe verson of Fg. 5(a). The parameters use n the moel are: h : stance from the floor to the center of the sensor system : angle of the frst sensor ray from the gravty recton 1 : angle between the 1 st sensor ray an the n sensor ray 3 : angle between the n sensor ray an the 3 r sensor ray If the floor s even, the change of the heght ue to the three sensor reangs shoul be the same: h = h = h 1 3 (3) cos = cos( + ) = 1 1 (4) cos( + + ). 3 1 3 We also represent the change of the angle th sensor reang change. Then, = arccos arccos 1 + 1 1 1 = arccos arccos + + 3 3 3 ue to the - = arccos arccos 3. (7) Smlarly, f the floor s even, the change of the angle ue to the three sensor reangs shoul be the same: = = (8) 1 3 arccos arccos = + 1 1 1 arccos arccos = + arccos arccos. + 3 3 3 (5) (6) (9)

M. Lee an S. Lee / Journal of Mechancal Scence an Technology 5 (4) (011) 1043~1050 1047 (a) Ascenng Star Fg. 11. Values of (b) Descenng Star h n Eqs. (1)-(3). Star heght: 159mm. Fg. 14. Experment wth 3-D Measurement System. 1050 (a) Ascenng Star 1000 Z [cm] 950 (b) Descenng Star Fg. 1. Values of h n Eqs. (1)- (3). Star heght: 106 [mm]. 900 0 00 400 600 800 1000 100 1400 X [cm] Fg. 15. Measure Poston of Sensor System. 00 195 (a) Ascenng Star 190 185 [eg] 180 175 170 Fg. 13. Values of (b) Descenng Star h n Eqs. (1)-(3). Star heght: 1 [mm]. Eqs. (4) an (9) are the contons necessary for the even state to reman n the same, even state. Fg. 11(a) shows the values of h n Eqs. (1), () wth an ascenng star, an Fg. 11(b), wth a escenng star (heght=159 [mm]). The state of the floor clearly s fferentate by the values of h. The same experments were conucte wth stars of fferent heght, an the results are shown n Fgs. 1 (heght=106 [mm]) an 13 (heght=1 [mm]), respectvely. 165 160 0 00 400 600 800 1000 100 1400 X [cm] Fg. 16. Measure Ptch Angle of Sensor System. 4. System performance varaton The sensor system affxe to the user s belt mght move vertcally or rotate whle the user walks; that s, the values of h an can change. In orer to evaluate the sensor system movement, a 3-D moton capture system was use, as shown

1048 M. Lee an S. Lee / Journal of Mechancal Scence an Technology 5 (4) (011) 1043~1050 n Fg. 14. The poston an the angular splacement (ptch angle) of the sensor system were measure, an the results are shown n Fgs. 15 an 16. An experment that entale followng a 1400 [cm] straght lne showe that the stanar evaton of the heght of the sensor system s 4.69 [mm], an the stanar evaton of the ptch angle, 1.3 [egrees]. We expecte that the movement of the sensor system woul affect the sensor measurements; performng a covarance analyss, we estmate the effects of the changes of the heght an the ptch angle. The change of the -th sensor reang ue to the heght change h an the ptch angle change s represente by h(cos cos( + )) + hcos = 1 cos( + ) cos (10) N = D (11) where an = h(cos( + ) cos( + + )) N 1 1 + h cos( + ) 1 = cos( + + ) cos( + ) D 1 1 3 N = 3 3D (1) where an = h(cos( + + ) 3N 1 3 cos( + + + )) + h cos( + + ) 1 3 1 3 + + + + + 3D = cos( ) cos( ). 1 3 1 3 The covarance matrx C, whch represents the strbuton of the sensor reangs, s estmate from the covarance matrx Ch,, whch represents the strbuton of the heght an the ptch angle C = F C F (13) T h, h, h, where Ch, s efne as Ch, = σ 0 h 0 σ an σ, σ are stanar evatons of heght an ptch h angle, respectvely. Smlarly, C s efne as C σ 0 0 1 = 0 σ 0 0 0 σ 3 where σ, σ σ are stanar evatons of the 1 st, n, 1 3 an 3 r sensors, respectvely. The Jacoban matrx Fh, s efne as h 1 1 1 = = h, F h h 3 3 3 h of whch each element s 1 1 h+ h = = h h cos( + ) cos( + ) h+ h 1 = = h h cos( + a + ) cos( + a + ) 1 1 h+ h 3 = h h cos( + a + a + ) 1 3 1 = cos( + a + a + ) 1 3 1 h+ h = cos( + ) ( h+ h)sn( + ) = + cos ( ) h+ h = cos( + + ) ( h+ h)sn( + + ) 1 = cos ( + + ) 1 1 3 cos ( + + + ) 1 3 1 h+ h 3 = cos( + + + ) 1 3 ( h+ h)sn( + + + ) = (14)

M. Lee an S. Lee / Journal of Mechancal Scence an Technology 5 (4) (011) 1043~1050 1049 Fg. 17. Estmate an measure sensor characterstcs. Fg. 19. Sensor output changes ue to the angle of ncence. The sensor outputs, wth several fferent angle of ncence values, are shown n Fg. 19. We can see that a large angle of ncence (larger than 60 egrees) egraes sensor accuracy. Fg. 18. Change of sensor outputs ue to ambent lghts. From the measure values of σ an σ, the value of C h was estmate. We also measure the strbuton of the outputs for the nomnal confguraton of h =1 [m], =30 [egrees], =10 [egrees], 1 3 =10 [egrees]. The stanar evatons of the sensor reangs were σ =0.0076 [mm], 1 σ =0.0139 [mm], σ =0.0310 [mm]. Fg. 17 shows the 3 estmate values of the stanar evatons from Eq. (13) an the measure values. On ths bass, we conclue that the effects of sensor system movement are neglgble. Another factor that can affect system performance s sensor accuracy. The nfrare range sensor GPY0A710K, base on the poston senstve evce [10], uses 870nm nfrare lght. As such, ambent lght can affect the sensor characterstcs. To test the sensor, four fferent experments were conucte uner the followng contons, respectvely: (a) fluorescent lamp lght an sunlght, (b) fluorescent lamp lght only, (c) sunlght only, an () no lght. Fg. 18 shows the sensor outputs n each case. It s apparent that the output change sgnfcantly uner conton (a), fluorescent lamp lght an sunlght, but ths stuaton rarely arses. We also neee to check the effect of the angle of ncence, because the sensor uses a poston senstve evce. If the user approaches stars roughly perpencular to the walkng path, the angle of ncence s close to zero. However, ths mght not be zero, epenng on the approachng angle of the user. 5. Conclusons In orer to facltate the walkng of bln an vsonmpare people, a sensor system that scans floor surfaces an etects the presence of stars was evelope. The system ncorporates three nfrare range sensors fxe raally, an knematc moelng was evelope for t. Three floor states, even floor, ascenng star an escenng star, can be accurately entfe from the fferences of the sensor values. For that entfcaton, the system, nstea of relyng on current sensor values, references ther changes as reflecte n state transtons. The sturbance ue to sensor system movement were consere an accounte for usng covarance analyss. Acknowlegment Ths work was supporte by a Natonal Research Founaton of Korea grant fune by the Korean Government [No. 010-0007794]. The authors are grateful to Husung Km an Sangcheol Park for ther help wth the harware experments. References [1] D. Yuan an R. Manuch, A Tool for Range Sensng an Envronment Dscovery for the Bln, 004 Conference on Computer Vson an Pattern Recognton Workshop (CVPRW'04), 3 (004) 39. [] D. Yuan an R. Manuch, Dynamc Envronment Exploraton Usng a Vrtual Whte Cane, Proceengs of the 005 IEEE Computer Socety Conference on Computer Vson an Pattern Recognton (CVPR 05), 1 (005) 43-49. [3] M. D. Aams, On-Lne Graent Base Surface Dscontnuty Detecton for Outoor Scannng Range Sensors, Proceengs of 001 IEEE/RSJ Internatonal Conference on Intellgent Robots an Systems (001) 176-1731. [4] A.-C. Scherlen, J. C. Dumas, B. Guej an A. Vgnot, Rec-

1050 M. Lee an S. Lee / Journal of Mechancal Scence an Technology 5 (4) (011) 1043~1050 ognze Cane, The new concept of a cane whch recognzes the most common objects an safety clues, The 9th Annual Internatonal Conference of the IEEE Engneerng n Mecne an Bology Socety (007) 6356-6359. [5] The Mngue ultrasonc moblty a, http://www.gpresearch.com.au/ultra.htm. [6] J. M. Benjamn, N. A. Al an A. F. Schepss, A laser cane for the bln, Proceengs of San Dego Bomecal Symposum, 1 (1973) 53-57. [7] D. Bsstt an A. D. Heyes, An Applcaton of Bofeeback n the Rehabltaton of the Bln, Apple Ergonomcs, 11 (1980) 31-33. [8] H. Km, M. Lee, S. Park an S. Lee, Exploraton of Floor Surface for the Bln, The 6th Internatonal Conference on Ubqutous Robots an Ambent Intellgence (URAI 009), (009) 575-579. [9] L. Km, S. Park, S. Lee an S. Ha, An electronc traveler a for the bln usng multple range sensors, IEICE Electroncs Express (6) 11, (009) 794-799. [10] Devce Specfcatons for GPY0A710K, Reference manual, Spec. No. ED-06G06, August 4, 006, SHARP Corporaton. Mnyoung Lee receve hs B.S. egree n Mechancal an System Desgn Engneerng from Hongk Unversty, Seoul, Korea, n 010. He s workng towars hs M.S. egree there. Hs research focuses on sensor fuson. Sooyong Lee receve hs B.S. an M.S. egrees n Mechancal Engneerng from Seoul Natonal Unversty, Seoul, Korea, n 1989 an 1991, respectvely, an hs Ph.D. egree from MIT, Cambrge, MA, n 1996. He worke as a Senor Research Scentst at KIST an then as an Assstant Professor n the Department of Mechancal Engneerng at Texas A&M Unversty. He jone Hongk Unversty, Seoul, Korea, n 003, an s currently an Assocate Professor n the Mechancal an System Desgn Engneerng Department. Hs current research nclues localzaton an sensor fuson.