Biomed Eng Lett (2013) 3:47-50 DOI 10.1007/s13534-013-0088-9
ORIGINAL ARTICLE
Proposal of a Simultaneous Ultrasound Emission for Efficient Obstacle Searching in Autonomous Wheelchairs Chang-Geol Kim and Byung-Seop Song
Received: 15 February 2013 / Revised: 27 March 2013 / Accepted: 28 March 2013 © The Korean Society of Medical & Biological Engineering and Springer 2013
Abstract Purpose Highly disable patients, who use the powered wheelchair, require an autonomous wheelchair for movement. Fundamental problem of the autonomous wheelchairs is safety. In order to dodge the obstacles, efficient searching method of the obstacles is necessary. Methods The proposed method utilizes multiple ultrasound emitters that generate signals of identical frequency and intensity. Corresponding sensors detect the reflected ultrasound signals, and the position of the obstacle is calculated based on the time of flight (TOF) of the ultrasound wave. It was carried out the obstacle searching experiments with proposed method. Results The results showed that the proposed method’s errors are within 1~2% which is much lesser than the general scan method. Therefore, the proposed method is suitable for autonomous wheelchairs as it facilitates detection of the nearest obstacle, and yields more accurate estimation of the position. Conclusions Therefore, the proposed method based on a simultaneous emission strategy is suitable for autonomous wheelchairs as it facilitates detection of the nearest obstacle, and yields more accurate estimation of the position. Keywords Rehabilitation, Autonomous wheelchair, Ultrasound, Simultaneous emission
INTRODUCTION Assistive technology (AT), which aids the disabled and elderly Chang-Geol Kim, Byung-Seop Song ( ) Department of Rehabilitation Science & Technology, Daegu University, 201 Daegudae-ro Jillyang Gyeongsan, Gyeongbuk, 712-714, Korea Tel : +82-53-850-4343 / Fax : +82-53-850-4349 E-mail :
[email protected]
people, is garnering worldwide attention, and the latest information technology (IT) and robotics technologies are being integrated with AT, resulting in novel devices [1-3]. One such life-changing device for the severely disabled people who cannot transport themselves is the autonomous wheelchair, the commercialization of which is still being impended. The overall purpose of an autonomous wheelchair is to transport the user to a destination safely and precisely. To operate the device without external aids, autonomous wheelchairs employ voice recognition, automatic control, radar, navigation, and robotics technologies. These technologies were used in several blind guide systems [3-9]. An essential capability that every autonomous wheelchair must have for user safety is obstacle detection and automatic avoidance. The obstacle detection and automatic avoidance technology, which is also used in autonomous mobile robots, consist of emission the ultrasound waves, locating the obstacles based on the detected reflection, and adjusting the path to avoid them. An autonomous wheelchair uses multiple ultrasound sensors. Each sensor sequentially emits and detects ultrasound waves. In the traditional method, an obstacle in the direction of each sensor is detected based on the time of flight (TOF) of the wave, and a map of surrounding obstacles is generated [10, 11]. However, this method has a drawback in calculating accurate distances when the signal is emitted from a moving wheelchair because the error in obstacle detection grows as the source moves, and each emitter sends signals at different positions and times. There were many efforts to resolve this problem and several researches could reduce the error [12]. But these research results can’t be the ultimate solution as the sources of error, i.e. the delay time, still exists. This paper proposes a simultaneous ultrasound emission technique that utilizes “crosstalk” as additional information to counter this problem. Crosstalk is an interference signal at
48
Biomed Eng Lett (2013) 3:47-50
a sensor in a multi-sensor configuration, which occurs due to simultaneous signal waves being emitted by many emitters. The proposed method emits ultrasound signals of identical frequency and intensity from all emitters at the same time, to reduce the error due to time delay that a sequential emission technique inherently generates. Each sensor detects the very first reflected signal from an obstacle, and in this way, the distance to the closest obstacle can be calculated with significantly improved accuracy levels.
the wheelchair moves at a constant velocity, the U2 sensor emitting place is changed from A position to B. The difference of distance (∆L2) from A to B can be calculated by multiplying the wheelchair velocity (v) and delay time (td). Therefore, the distance errors are always occurred and should be increased directly proportional to both the wheelchair speed and scan speed. If the error is occurred, the system may wrongly estimate the location of the obstacle from OBJ to OBJ' which has incorrect distance and direction as shown In Fig. 1.
METHOD
Simultaneous ultrasound emission method The proposed method emits simultaneous ultrasound signals of the same frequency and intensity from multiple emitters, and in the process detects the nearest obstacle. Each ultrasound sensor has an effective emission angle of 60o, and the searching system collects information from multiple sensors for the front of the wheelchair. For instance, the 3-sensor configuration shown in Fig. 2a, which considers the wheelchair motion and the overlapping area, can cover an angle of about 120o. Assuming that all the sensors emit signals at the same time, the reflected signal will be detected first at the closest sensor as shown in Fig. 2b. For an object at a longer distance, each sensor may capture reflected waves from other sensors that have longer flying time. The traditional method regards any reflected signal from other emitting sources detected at a sensor as crosstalk noise. However, the proposed method considers crosstalk as additional information. In the proposed method, the distance between each sensor and the obstacle is calculated as follows:
Error occurrence of conventional method The ultrasound emission method was widely used in object searching because it is very simple and cheap. Generally, ultrasound sensors sequentially emit and detect the reflected ultrasound waves. The durations between the emission of the signal and the detection of the reflection in each sensor are used in the distance calculation. The distance from the sensor to the detected object could be accurately estimated using the TOF method. If more than three distances between each sensor and the object are obtained, the 3-dimensional location could be calculated by the spherical trigonometry. But, in the moving condition at autonomous wheelchair, the estimation error should be occurred due to the alteration of each sensor location. [13] In searching system, some delay time between each sensor emission exists to avoid interferences between the waves from each sensor. In Fig. 1, ultrasound sensor U1 emits the wave and detects the reflected wave then, U2 sensor emits the wave. There is some delay time (Td) to avoid interference between each sensor. Assuming
Fig. 1. Erroneously estimation of the obstacle location.
Tn,n Ln = c -------2
(1)
Tn,n – 1 + ( Tn,n – 1 – Tn,n ) Ln – 1 = c ---------------------------------------------------2
(2)
Fig. 2. The obstacle detection with ultrasound sensors. (a) Simultaneous emission and reflection pattern. (b) Distance calculation model.
Biomed Eng Lett (2013) 3:47-50
49
where, c is the speed of ultrasound wave, Ln is the distance between sensor ‘n’ and the obstacle, Ln-1 is the distance between sensor ‘n-1’ and the obstacle, Tn,n is the TOF for a wave emitted from sensor ‘n’ to return to sensor ‘n’ and Tn,n-1 is the TOF for a wave emitted from the sensor n to return to the sensor ‘n-1’. Experiments The proposed method was validated and compared with the traditional method using sensors installed on a wheelchair. The scan rate of the sensors used in the experiment was 30 msec, and the center frequency of the emitted ultrasound signal was 40 kHz. Five sensors were placed in the front portion of the wheelchair, 15 cm apart from each other, and a 56 cm × 62 cm wooden panel was used as the obstacle. Experimentally, we had found out that the ultrasound sensor could stably detect any shape of object bigger than 20 cm × 20 cm size. The first measurements were taken using both the methods for a stationary wheelchair with the obstacle placed 2 m and 3 m ahead. Next, the obstacle was placed 4 m ahead, and the distance was measured from the moving wheelchair as it approached 2 m and 3 m distances from the obstacle at 60 cm/s, 90 cm/s and 120 cm/s speed. The test speed of the wheelchair was determined based on the speed during actual operation, which is around 1 m/s. Each test segment was repeated three times, and the average distance was considered. For the scan method, ultrasound signals were sequentially emitted from the U1 sensor on the left to U2 - U5 at an interval of 30 ms, and the distance was calculated using the measured TOF at each corresponding sensor. For the simultaneous emission method, distances from each sensor to the obstacle were calculated using the measured TOF of simultaneously emitted signals from each sensor. An 8-bit, ATmega 128 microprocessor was used to control the sensors, and to calculate the distances using the measured times. The microprocessor calculated the distances in real time and saved the data on a computer. Fig. 3 shows the percentage errors of the calculated distances to the measured values.
DISCUSSION AND CONCLUSION As shown in Fig. 3, both methods yielded errors within 1– 2% with no significant difference for the stationary case, except for the 5th sensor that resulted in a larger error for the scan method. The moving wheelchair cases yielded larger errors with a little correlation with the speed. However, the errors still stayed within 2% for the simultaneous emission method, while the scan method yielded generally larger errors, with some sensors showing much larger error
Fig. 3. Percentage error of calculated distances to measured values for various wheelchair speeds: (a) Stationary, (b) 60 cm/s, (c) 90 cm/s, (d) 120 cm/s.
jumps. The error jumps were significant on U1 and U5, and Fig. 3b shows the jump on U4 as well. It is inferred that the cause for the larger errors is the inherent time delay in the scan method. In general, the proposed method shows better results in terms of the error in distance calculation than the
50
traditional scan method. If the distances from three or more sensors to the obstacle are known, the obstacle can even be located in 3-D space. Assuming that each sensor is at the centre of a sphere of which the radius is the distance between the obstacle and sensor, the intersection of the spheres that the sensors generate denotes the location of the obstacle. Therefore, with three sensors solving three sphere equations yields the location of the obstacle, which is an efficient method for an autonomous wheelchair. Since accurate estimation of distances is crucial in 3-D calculations, the proposed method is suitable for application to 3-D mapping of obstacles [13]. Actually, the estimation errors are within 10 centimeters assuming the wheelchair speed is 1 m/s, delay time is 30 ms and 2 meters distance of obstacle. It may be considered as a small thing. But, in autonomous wheelchair, the safety is one of the most important things and the tiny error can bring about serious trouble because the user of the wheelchair is a person who has some disability. Therefore, the smaller error the searching system has, the safer autonomous wheelchair is. However, the proposed method is superior to the scan method in every aspect. The proposed method detects only the closest object, while the scan method can detect multiple objects at the same time. Searching for multiple obstacles is beneficial because a wheelchair encounters many in a practical environment. On the other hand, from a user’s perspective, information about the closest object is most important. Therefore, it is critical for the safe and convenient operation of autonomous wheelchairs to detect precisely and avoid the closest object. In summary, the proposed obstacle detection method based on a simultaneous emission strategy exhibits excellent performances on autonomous wheelchairs.
ACKNOWLEDGMENTS This work was supported by the National Research Foundation of Korea Grant funded by the Korean Government (NRF2010-013-D00091).
Biomed Eng Lett (2013) 3:47-50
CONFLICT OF INTEREST STATEMENTS Kim CG declares that s/he has no conflict of interest in relation to the work in this article. Song BS declares that s/he has no conflict of interest in relation to the work in this article.
REFERENCES [1] Brabyn J, Gerrey W, Fowle T, Aiden A, Williams J. Some practical vocational aids for the blind. Conf Proc IEEE Eng Med Biol Soc. 1989; 1502-3. [2] De Acevedo Renato LM. Electronic device for the blind. IEEE AERO El Sys Mag. 1999; 4-7. [3] Tachi, Komorya K, Tanie K, Ohno T, Abe M. Guide dog robotfeasibility experiments with Meldog Mark III. Proc Int Symp Ind Robot. 1981; 95-102. [4] Tachi S, Tanie K, Komoriya K, Abe M. Electrocutaneous communication in a guide dog robot (MELDOG). IEEE T BioMed Eng. 1985; 32(7):461-469. [5] Kim CG, Lee HG, Kang JH, Song BS. Research of wearable walking assistive device for the blind. Korean J Vis Impair. 2007; 23(1):147-64. [6] Borenstein J, Ulrich I. The guidecane-a computerized travel aid for the active guidance of blind pedestrians. Proc IEEE Int Conf Robot Automat. 1997; 20-5. [7] Shoval S, Ulrich I, Borenstein J. The Navbelt-A computerized travel aid for the blind based on mobile robotics technology. IEEE T Bio-Med Eng. 1998; 45(11):1376-86. [8] Kang JH, Kim CG, Lee SH, Song BS. Development of walking assistance robot for the blind. J Korean Sensor Soc. 2007; 16(4):286-93. [9] Rentschler AJ, Cooper RA, Blasch B, Boninger ML. Intelligent walkers for the elderly: Performance and safety testing of VAPAMAID robotic walker. J Rehabil Res Dev. 2003; 40(5):42332. [10] Shoval S, Ulrich I, Borenstein J. Robotics-based obstacleavoidance systems for the blind and visually impaired, NavBelt and the GuideCane. IEEE Robot Autom Mag. 2003; 9-20. [11] Moon CS, Do YT. Design of range measurement systems using a sonar and a camera. J Korean Sensor Soc. 2005; 14(2):116-24. [12] Borenstein J, Koren Y. Error eliminating rapid ultrasonic firing for mobile robot obstacle avoidance. IEEE T Robotic Autom. 1995; 11(1):132-8. [13] Kim Chang-Geol, Song Byung Seop. Object search using synchronous ultrasonic wave emission for the blind guide system. J Biomed Eng Res. 2008; 29(5):384-91.