IROS2019国际学术会议论文集2461

1 Abstract We introduced the disturbance degree of ground and proposed an evaluation method to measure the mobile performance of a crawler on soft ground during direction switching. First, we developed a planar omnidirectional crawler, which had a configuration with two left and right unit crawlers for performing turning motion, as the target for evaluation. Second, by utilizing the proposed disturbance degree of ground evaluation method, we investigated how the turning and translational motions of the crawler mechanism affected soft ground by measuring the flow of sand on a horizontal surface. It was quantitatively shown that translational motion switched the travel direction with lesser disturbance to the road surface compared to turning motion. We confirmed that ground disturbance could be evaluated during direction switching using the proposed method. I. INTRODUCTION We introduced the disturbance degree of ground and proposed an evaluation method to measure the mobile performance of a crawler on soft ground during direction switching. We developed a planar omnidirectional crawler, which had a configuration with two left and right unit crawlers for performing turning motion, as the target for evaluation. We have developed an omnidirectional mobile mechanism with surface contact on the ground, as shown in Figure 1. Holonomic omnidirectional mobile mechanisms with point 1-5 or line contact 6-9 with a running surface have been reported in previous studies. However, the transmission of driving force to small crawlers surrounding a large crawler is difficult. Therefore, a planar omnidirectional mobile mechanism is being developed 10,11. The conventional planar crawler mechanism is widely adopted for search and rescue robots 12-14. However, it is extremely difficult to avoid obstacles on stairs because the switching of the travel direction on stairs by turning motion is dangerous. In contrast, as the planar omnidirectional crawler mechanism can exert driving force in any direction, the travel direction can be switched on stairs. Further, it is difficult for the conventional planar crawler to perform turning motion on a grating area because grousers become trapped in the holes of gratings. In contrast, the omnidirectional crawler can easily travel on gratings because it can perform translational motion. In addition, the planar omnidirectional crawler does not experience the wall-deadlock problem 15 that occurs in the conventional planar crawler. The movement accuracy of the Figure 1. Planar omnidirectional crawler with configuration using two left and right unit crawlers for turning motion. Large Diamater Crawler Longitudinal Direction Motion Deformable Crawler with Uneven Terrain Omnidirectional Movement Sideways Motion Small Diamater Crawler ( a ) Conventional crawler.( b ) Omnidirectional crawler. Damage the ground Stuck in sand 【 Top view 】 Figure 2. Advantage of the planar omnidirectional crawler. conventional planar crawler along a path planning route is poor because turning motion causes large slippage on the ground 16. In contrast, the planar omnidirectional crawler can achieve high accuracy along a path planning route because it does not slip in an ideal state. The planar omnidirectional Rack chain Spur gear Taper part AA battery for size comparision Eri TAKANE*, Kenjiro TADAKUMA*, Tori SHIMIZU, Sosuke HAYASHI, Masahiro WATANABE, Shingo KAGAMI, Keiji NAGATANI, Masashi KONYO, and Satoshi TADOKORO Basic Performance of Planar Omnidirectional Crawler during Direction Switching using Disturbance Degree of Ground Evaluation Method This work was supported in part by Japan Science and Technology Agency “Impulsing Paradigm Change Through Disruptive Technologies Program: Tough Robotics Challenge.” *E. Takane and K. Tadakuma contributed equally to this work. E. TAKANE, K. TADAKUMA, T. SHIMIZU, S. HAYASHI, M. WATANABE, S. KAGAMI, M. KONYO, and S. TADOKORO are with the Graduate School of Information Sciences, Tohoku University, Sendai 980-8579, Japan (e-mail: takanerm.is.tohoku.ac.jp; tadakumarm.is.tohoku.ac.jp; shimizu.torirm.is.tohoku.ac.jp; hayashi.sosukerm.is.tohoku.ac.jp; watanabe.masahirorm.is.tohoku.ac.jp; swkic.is.tohoku.ac.jp; konyorm.is.tohoku.ac.jp; 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) Macau, China, November 4-8, 2019 978-1-7281-4003-2/19/$31.00 2019 IEEE2732 crawler can move efficiently because it is not necessary for the crawler to turn back several times to move laterally. The planar omnidirectional crawler mechanism facilitates traveling on soft ground. The turning motion by the conventional planar crawler exerts large frictional force on the ground and crawler because slipping area is large. This results in scratches on the ground and causes the failure of the crawler 17,18. In addition, the turning motion on soft ground causes the crawler to dig into the earth and thus become trapped in the ground. Owing to this, traveling especially spin turning on soft ground is difficult. (Figure 2 (a) 19,20. In contrast, the planar omnidirectional crawler can switch the travel direction without requiring the crawler to turn. Thus, the ground is not damaged and the crawler does not become trapped in sand (Figure 2 (b). Research on how robots affect the surface on which they run has been performed in the field of terramechanics 21-24, mainly for planetary exploration, volcanic exploration, and soft ground investigation after a tsunami. In 25-28, the flow movement of sand in the depth direction was analyzed by observing wheels and crawlers from the side cross section when a robot moved forward on soft ground. In addition, studies were conducted to observe the imprint of the planar crawler mechanism on soft ground and evaluate the turning property 29. Furthermore, the three-dimensional flow movement of sand under mobile robots was measured using DEM, X-rays, and radiation 30-35 without the problem that the behavior of sand near a boundary surface is different from actual soft ground was not considered 36,37. Even though studies have experimentally shown that translational and turning motions disturb the road surface on soft ground, particularly that close to an actual environment, the quantification of the disturbance and flow of sand is still unclear. The quantification of the disturbance and flow of sand make a contribution to development of robots that travel on soft ground. The quantitative evaluation of the degree of disturbance caused to a running surface can be used as a design criterion for mobile mechanisms, and future crawler designs can be improved by reducing this degree. In addition, various evaluation methods have been proposed to guarantee robot mobile performance 38. We aim that mobile performance can be guaranteed when the disaster response robot is operated in an actual environment by evaluation standard for ability on the soft ground using our proposed method. Herein, mobile performance means that ability enables mobile robots to travel without damaging the running surface and themselves. Therefore, in this study, we proposed a method to measure the sand flow in a horizontal plane when a robot rotated on a spot in real time. In particular, we attempted to quantify and compare the influences of the turning and translational motions of the planar crawler mechanism when the crawler changed its travel direction by investigating the disturbance degree of the running surface. We defined this degree as the movement distance of sand based on optical flow. We proposed a method for the standard evaluation of switching the travel direction. By utilizing the method, we quantitatively determined that the planar omnidirectional crawler mechanism could switch the travel direction without damaging the ground, as compared to the conventional mechanism. The rest of the paper is organized as follows: In section II, we describe the prototype configuration developed for realizing turning motion. The configuration consists of two left and right unit crawlers with rack chain mechanisms to transmit driving force to small crawlers surrounding a large crawler. Section III explains the experimental method of observing and analyzing the state of gravel in the horizontal plane from the bottom while the crawler is traveling. Section IV summarizes the experimental results, and we quantitatively confirm that it is possible to switch the travel direction by only causing a slight disturbance to the road surface in translational motion compared to turning motion. Finally, Section V provides the conclusions. TABLE I. PROTOTYPE MACHINE SPECIFICATIONS. Sprocket diameter of large crawler165 mm Sprocket diameter of small crawler24 mm Depth640 mm Height310 mm Width795 mm Weight48.9 kg Motor Velocity of translational motion 85 mm/s Velocity of turning motion 4.2 and 10.5 rad/min 150W RE40 + HEDL554 + 203123 (maxon motor co.) Motor rotational speed is 50% of its max rotational speed. ( ( Motor rotational speed is 20% and 50% of its max rotational speed respectively. ( ( II. CONFIGURATION WITH TWO LEFT AND RIGHT UNIT CRAWLERS We realized the configuration with two left and right unit crawlers enabling a turning motion equipped with a rack chain mechanism 11, which can transmit the driving force to small crawlers surrounding a large crawler. The appearance of the planar omnidirectional crawler prototype is shown in Figure 1, and the prototype machine specifications are listed in Table 1. To improve their step-climbing ability, the small diameter crawlers have a tapered shape. Two motors are arranged for the large crawler, so a totally four motors are mounted inside two large crawlers. The difference between the rotational speeds of the left and right large crawlers allows the robot to make a turn. Depending on the application, the turning motion is not required, and only the translational motion is enough. For example, in a factory work robot, when an arm with a pivot axis is mounted, work can be performed without changing the posture around the yaw axis through mobile mechanisms. However, in this study, we realized the planar omnidirectional crawler mobile mechanism in a search-and- rescue robot. Depending on the crawler diameter, length, and center of gravity, the step-climbing ability changes in the front and side directions. In addition, the road width changes depending on the approach angle. In disaster sites, the turning motion is effective because various situations can be encountered. It is better for the rescue robot to have a turning motion. When using it in actual disaster site, guarantee is necessary to satisfy specifications. In other words, it is necessary to evaluate the turnability on the soft ground. 2733 III. VERIFICATION OF REDUCTION IN THE DISTURBANCE DEGREE OF THE GROUND DURING DIRECTION SWITCHING A. Proposed Method and Experimental Purpose We proposed the disturbance degree evaluation method to measure the mobile performance of the omnidirectional crawler during direction switching on soft ground. We experimentally examined how the road surface is affected by the movement of the crawler. Furthermore, we measured the amount of surface disturbance on the road when the crawler switched the travel direction. In addition, we experimentally showed that the planar omnidirectional crawler mechanism allowed for movement with less slippage on the ground compared to the conventional crawler mechanism. B. Experimental Equipment We prepared gravel on a transparent acrylic stage to observe the sand flow from the bottom of the robot when the crawler moved. Two types of gravel were used with diameters of 2430 mm and 2 mm. Figure 3 shows the schematic of the experimental apparatus, and Figure 4 shows the actual apparatus. Gravel movement was captured by a camera placed under the acrylic stage; the crawler switched its travel direction by 90 through turning and translational motions. A stage thickness of 60 mm with sufficient rigidity was used so that the acrylic stage did not deflect. The gravel comprised white stones with sizes of approximately 2430 mm, which were sufficient for image processing. 14-kg gravel was arranged in an area of approximately 700 750 mm2 so that it did not overlap (a single layer). A camera for measurement (GoPro Hero 6 4k 30 fps) was installed at a position 700 mm below the acrylic stage. To measure the state of the gravel through optical-flow-based image processing, the background in the cameras angle of view was covered with a black cloth to easily distinguish the white stones to be measured. The gravel was illuminated from the side during the measurement. After the crawler moved into the measurement range from the start area, it switched its travel direction by 90 and exited to the finish area. This movement was performed in four manners, as described below, and it was measured five times in each direction. Figure 3 shows movement (iii). (i) front right spin turning forward (ii) backward left spin turning backward (iii) forward right (iv) left backward Furthermore, experiments were conducted in which white gravel with a size of approximately 2 mm in diameter was arranged along a height of approximately 10 mm. The detailed movements can be observed in the attached movie. C. Analysis of Sand Movement For the initial gravel state, OpenCV was used to detect 5000 feature points using Shi-Tomasis corner detection method. Then, gravel movements (feature points) were tracked when the crawler entered the measurement area, when it switched its travel direction by 90, and when it exited the area by employing the iterative LucasKanade method, which calculates optical flow with pyramids. We also calculated the analysis area Start area Final area Camera Swithing direction area Stone Transparent stage 【 Front view 】 【 Top view 】 Gravel 1 2 12 Planar omni-crawler at the initial position Planar omni-crawler at the final position X Y Figure 3. Schematic of the experimental apparatus. Figure 4. Actual appearance of the experimental apparatus. instantaneous velocity of tracked feature points. The amount of movement at the feature points was calculated in the X and Y directions after the crawler reached the finish area. The moved feature points were counted for every movement measuring 0.44 mm (1 pixel). Herein, the n- th coordinate values of the feature points at the initial state and final state are (1，1) and (，), respectively. and indicate the amount of movement at feature points on the X- axis and Y-axis, respectively, and they are calculated as follows: = | 1|, = | 1| . (1) The average distance of the moved feature points ( ) can be expressed as follows: = 1 2+ 2 =1 (where2+ 2 ). (2) where the number of the moved feature points is . In the case that the distance of a moved feature point was extremely small and less than , the movement of the feature point was considered close to zero and the feature point was regarded as a stationary Planar omnidirectional crawler X Y 2734 1 0.0 s 2 3 32.18 s 4 1 0.0 s 2 3 37.09 s 4 【Bottom view】 Analysis area 1 2 3 4 ( a ) Experimental movement of the crawler. ( b ) Result of optical flow ( c ) Appearance of the movement on stoned of 24-30-mm diameter. of 2-mm-diameter stone. X Y X Y X Y Figure 5. Gravel appearance during turning motion (bottom view). point and not included in the calculation. was calculated by measuring stationary gravel. The counted feature points were represented by a cumulative frequency curve for every 10 pixels (class interval is 10 pixels), and turning motion (conventional method) and translational motion (omnidirectional movement) were compared. Herein，the maximum class value is M, the k-th class value is , and the k-th frequency is . The k-th cumulative frequency, , is expressed as follows: = 1+ =1 . (3) Figure 6. Instantaneous velocity of the feature points (bottom view). 0 0.2 0.4 0.6 0.8 1 1.2 0 100 200 300 400 500 020406080100120140 Cumulative frequency Number of feature points Distance of movement of feature point mm Thenum.offeaturespts.inX-axisdir.(20%speed) Thenum.offeaturespts.inY-axisdir.(20%speed) Thenum.offeaturespts.inX-axisdir.(50%speed) Thenum.offeaturespts.inY-axisdir.(50%speed) CumulativefrequencyinX-axisdir.(20%speed) CumulativefrequencyinY-axisdir.(20%speed) CumulativefrequencyinX-axisdir.(50%speed) CumulativefrequencyinY-axisdir.(50%speed) ( a ) Result of crockwise rotation motion. ( b ) Result of countercrockwise rotation motion. 0 0.2 0.4 0.6 0.8 1 1.2 0 100 200 300 400 500 020406080100120140 Cumulative frequency Number of feature points Distance of movement of feature point mm Thenum.offeaturespts.inX-axisdir.(20%speed) Thenum.offeaturespts.in Y -axisdir.(20%speed) Thenum.offeaturespts.inX-axisdir.(50%speed) Thenum.offeaturespts.inY-axisdir.(50%speed) CumulativefrequencyinX-axisdir.(20%speed) CumulativefrequencyinY-axisdir.(20%speed) CumulativefrequencyinX-axisdir.(50%speed) CumulativefrequencyinY-axisdir.(50%speed) Figure 7. Relationship between the number and the moving distance of feature points during turning motion. IV. EXPERIMENTAL RESULTS A. Turning Motion (Conventional Method) We measured the degree of disturbance during turning for two cases, i.e., 20% and 50% of the motors maximum rotational speed. The gravel appearances at 20% speed in the experiment and the results of image analysis are shown in Figure 5, which illustrates movement (i) described in Section III.C. Figures 5 (b) and (c) show the situation after the initial state, after entering in the forward direction, after turning motion, and after exiting in the forward direction. Figures 5 (b) and (c) show that turning motion resul