Stability Improvement Using Soft Sole on Humanoid Robot Masanao Koeda 1, Tsuneo Yoshikawa 1, and Keita Nakagawa 1 1 Ritsumeikan University, College of Information Science and Engineering, Department of Human and Computer Intelligence Noji Higashi 1-1-1 Kusatsu, 2-877 Shiga, Japan. email: {koeda, yoshikawa, nakagawa}@robot.ci.ritsumei.ac.jp Abstract This paper shows the effect of a stability improvement of humanoid robot with a soft sole of feet which is aimed at walking on an uneven terrain. Generally, it was thought that the soft sole absorbs shocks but decrease stability at the same time. We conducted preliminary experiments of shifting center of gravity, stepping motions, and walking on an even/uneven terrain with and without soft sole of the feet. The experimental results show that the soft sole increases the stability in certain conditions. Keywords Humanoid Robot, Soft Sole, Stability, Uneven Terrain I. Introduction RECENTLY, various humanoid robots have been developed[1], [2]. For humanoid robots, walking skill is one of the most important and basically functions. However, at this time, it is difficult for humanoid robots to walk adaptively on the floor conditions such as a deformable floor like a grass plot or a sandy place, or an uneven terrain like a gravel road. Researches to walk on an uneven terrain by a humanoid robot can be classified into category; Active and Passive. The former aims for recognition of the condition of the floor using a sensor such as a stereo camera, or an ultrasonic sensor [3], [4], and generation of gait pattern depending on the floor condition [], [6], [7]. Yamaguchi et al.[8] developed mechanical foot with shock absorbing material and investigated its stability. Generally, it is known empirically that the soft sole absorbs shocks but decrease stability at the same time. We investigated the stability of a humanoid robot with a soft sole at the feet s bottom. Experiments were conducted in following 3 motions; shifting the center of gravity, stepping, and walking on an even/uneven terrain, and stability margin was compared each of them with the case of not using any rubber foam. Finally, we show that the stability was improved in a certain condition contrary to our expectations. II. System Overview Our system consists of a humanoid robot HOAP- 2(Fig.1) and a host PC. The system overview is shown in Fig.2. A. HOAP-2 HOAP-2 is a commercial humanoid robot which is developed by FUJITSU AUTOMATION Ltd. Its height is [cm] and weight is 7[kg] approximately. The size of foot is 98[mm] by 63[mm]. Detailed specification is shown in TABLE I. In each foot, 4 force sensors are mounted, and 3 axis acceleration/angular sensors are Fig. 1. Host PC RT-Linux Fig. 2. Humanoid robot HOAP-2 USB DC 24V System configuration HOAP-2 equipped in its body. Fig.4 illustrates the size of the feet and the position of the force sensors. CH., 1, 2 and 3 show the position of the force sensors. B. Host PC The robot is controlled in 1[ms] through a host PC which is running RT-Linux OS. They are connected by USB interface. The PC sends control commands to the robot, and the robot transmits acquired data from several sensors. A wireless network based on 82.11b can be used for communication between the PC and the robot. However, we selected the wire communication because the moving range of the robot is not so large in this study.
d1 d2 d3 ZMP d4 Fig. 3. Attached 1[mm] rubber foam Sole Fig.. Stability margin TABLE I Specifications of HOAP-2 Fig. 4. C. Soft Sole Foot size and sensors position 6 kinds of rubber foams of, 1, 1, 2, 2, and 3[mm] thickness were used as the soft material. The Young s modulus of the rubber foam is 4.3[Pa]. The foams are cut in the same size as the robot s foot and attached to the sole by a thin two-sided tape. Fig.3 shows the feet which is attached with 1[mm] rubber foam. When no rubber foam is attached, it is called [mm] thickness afterward. A. Stability Margin III. Evaluating Method To calculate stability of the robot quantitatively, Stability margin is defined as the following equation for evaluation criteria. (Stability Margin) = min {d i i = 1, 2, 3, 4} (1) where d i means the distance between ZMP(Zero Moment Point)[9] to each edge of the sole. In Fig., d 3 becomes stability margin because of the shortest distance from ZMP to the right edge. It can be said that the stability is improved when the stability margin becomes large. Height [cm] Weight 7[kg] DOF 6 in Leg 2 4 in Arm 2 1 in West 1 1 in Hand 2 2 in Neck 1 (Total 2DOF) Sensors Joint Angle Sensor Resolution:.1[deg/pulse] 3 Axis Accelerometer Measuring Range: ±2[G] Resolution:.[G] 3 Axis Gyrometer Measuring Range: ±6[deg/s] Resolution:.2[deg/s] Foot Load Sensor 4[ch/foot] PC OS: RT-Linux CPU: Pentium4 2[GHz] Memory: 1[GB] Sampling Rate: 1[ms] IV. Experiments and Results We conducted the following 4 set of experiments; shifting center of gravity, stepping motions, and walking on an even/uneven terrain with and without the soft sole of the feet. To compare the experimental results easily, all motions were generated by a complete open loop without feedback from any sensors. A. Shifting Center of Gravity A.1 Right and Left To move the center of gravity to the right and to the left, the leg was tilted slowly by 1[deg] without lifting the foot while the body was horizontal. This motion was operated for 2[sec]. Fig.7 shows the sequential images of this motion. This motion was operated 3 times in the
3 3 Stability Margin [mm] 3 2 2 1mm sole 1 mm sole mm sole 1 1 2 3 4 6 7 8 9 1 Time [sec] (a) Right foot Stabiity Margin [mm] 3 2 2 1mm sole 1 mm sole mm sole 1 1 11 12 13 14 1 16 17 18 19 2 Time [sec] (b) Left foot Fig. 6. Change of Stability Margin while shifting COG (a) Initial pose ([s]) (b) Tilt at 1[deg] to the right ([s]) Fig. 7. (c) Center (1[s]) Shifting COG to the right and left (d) Tilt at 1[deg] to the left (1[s]) (e) End (2[s]) (a) Initial pose ([s]) (b) Tilt at [deg] to forward ([s]) Fig. 8. (c) Center (1[s]) (d) Tilt at [deg] to backward (1[s]) Shifting COG to backward and forward (e) End (2[s]) conditions that 7 kinds of rubber foam (from [mm] to 3[mm] thickness) was attached. Fig.6 shows temporal changes of the stability margin while 3 kinds of sole (,, and 1[mm]). The horizontal axis shows time and the vertical axis shows the stability margin. In consideration of the measuring range of the force sensor, ZMP of the right foot in -1[sec] and ZMP of the left foot in 1-2[sec] were used to calculate the stability margins. The stability margins have decreased along with the shifting center of gravity. Moreover, the lowest value of the stability margin increases by attaching rubber foams. In the following experiments, the lowest values of the stability margin are compared. Fig.9-(a) shows the average of minimum stability margin in 7 kinds of soles. The stability margin became the maximum at 2[mm], and it tended to decrease in thicker rubber. A.2 Backward and Forward To move the center of gravity to backward and forward, the leg was tilted slowly by [deg] without lifting the foot while the body was horizontal. This motion was operated for 2[sec], and Fig.8 shows the sequential images of this motion. Fig.9-(b) shows the average of minimum stability margin in 7 soles. It shows that the rubber does not influence the stability margin so much in this motion. B. Stepping Motion Next, the stability margins while the spot stepping motion were compared. Experimental conditions were that one step took 8[sec] and each foot was lifted by 2[mm]. Fig.1 shows the sequential images of this motion. The averages of minimum stability margins while the motion was measured for 7 kinds of sole in times respectively. The results are shown in Fig.11. To calculate the stability margin in the measuring range of the force sensor, the ZMP of the landing foot was calculated in -2[sec], 6-8[sec], 8-1[sec], and 14-16[sec]. The results shows that the stability had decreased when over 2[mm] thickness rubber was attached, and the same tendency appeared as the previous experiments.
2 2 1 1 1 1 2 2 3 Minumum Stability Margin [mm] Minimum Stability Margin [mm] 3 3 2 2 1 1 (a) Right and left 1 1 2 2 3 (b) Backward and forward Fig. 9. Stability margin while shifting COG motion ([s]) (b) Shift COG to right (2[s]) (f) Shift COG to left (1[s]) (c) Lift left foot vertically (4[s]) (g) Lift right foot vertically (12[s]) (d) Lower left leg (6[s]) (h) Lower right leg (14[s]) (e) Shift COG to center (8[s]) (i) Shift COG to center (16[s]) Fig. 1. Stepping motion ([s]) (b) 1st step (4[s]) (c) 2nd step (1[s]) (d) 3rd step (16[s]) (e) 4th step (22[s]) (f) th step (28[s]) (g) 6th step (34[s]) (h) 7th step (4[s]) (i) 8th step (46[s]) (j) End (48[s]) Fig. 13. Even terrain walking with 1[mm] sole C. Walking Motion C.1 On Even Terrain Experimental conditions were that one step took 6[sec], each foot was lifted by 2[mm], the length of stride was 3[mm], and the robot took 8 steps forward. The experiments were conducted 7 times in each sole respectively, and the stability margin and the number of non-fall were compared. Fig.13 shows the sequential images of this motion. Fig.12 shows the average of minimum stability margin of each experiment. Same as in
Minimum Stability Margin [mm] Minimum Stability Margin [mm] 3 2 2 1 1 Fig. 11. 3 2 2 1 1 Fig. 12. 1 1 2 2 3 Stability margin on stepping motion 1 1 2 2 3 Stability margin on even terrain walking the case of the stepping motion, the ZMP of the landing foot was used. In whole of the conditions, the stability margin became small since the walking motion was unstable itself, and it shows a similar tendency as the previous experiments. Fig.17-(a) shows the number of non-fall while the robot takes 8 steps in 7 trials. The robot did not fall over when the rubber foam below 1[mm] thickness was attached on the sole. However, when the rubber was over 2[mm] thickness, the robot tended to become unstable and fall over. C.2 On Uneven Terrain Finally, we confirm the effect of the soft sole while walking on an uneven terrain. The motion is the same as the previous one. Fig. 14 shows the condition of the constructed uneven terrain for this experiment. Hemispheric convex objects of [mm] in diameter are arranged by every 8[mm] on a woody board. The character S in Fig.14 indicates the starting position and the 7 arrows which were drawn every 1[deg] mean the starting pose of the robot. Fig. 1 and 16 are the experimental snapshots when the robot which has or 1[mm] rubber sole is walking in the direction of the red arrow in Fig.14. The numbers of non-fall in 7 trials are indicated in Fig.17-(b). The experimental results shows that the excessive thickness of the rubber makes stability decrease, and the thickness about 1[mm] is suitable for walking on an uneven terrain. 4 18 8 8 8 8 9 6 8 8 8 8 8 8 8 Fig. 14. 1[deg] S Condition of uneven floor V. Conclusion [mm] In this paper, we described the stability improvement of a humanoid robot which was attached soft materials on its soles. We conducted experiments of shifting center of gravity, stepping motions, and walking on an even/uneven terrain with and without soft sole of the feet. As a result, comparing the soft sole with the hard sole in the point of view of the stability margin, soft soles increase stability in certain conditions. Proposed method is quite simple remodeling to improve the stability, and it is effective for an irregular terrain. In the future, we will analyze this effect mathiematically, and investigate the optimal thickness, hardness and shape of the sole, and we will also establish a suitable control method. References [1] Y. Sakagami, R. Watanabe, C. Aoyama, S. Matsunaga, N. Higaki, and K. Fujimura, The intelligent ASIMO: System overview and integration, In Proceedings of 22 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 2478-2483, 22. [2] K.Kaneko, F.Kanehiro, S.Kajita, H.Hirukawa, T.Kawasaki, M.Hirata, K.Akachi, and T.Isozumi, Humanoid Robot HRP-2, In Proceedings of 24 IEEE International Conference on Robotics and Automation (ICRA), pp. 183-19, 24. [3] T. Yoshimi, Y. Kawai, Y. Fukase, H. Araki, and F. Tomita, Measurement of Ground Surface Displacement using Stereo Vision and Mechanical Sensors on Humanoid Robots, In Proceedings of 23 IEEE International Conference on Multisensor Fusion and Integration for Intelligent Systems (MFI), pp.12-13, 23. [4] S. Kajita, and K. Tani, Adaptive gait control of a biped robot based on realtime sensing of the ground, In Proceedings of 1996 IEEE International Conference on Robotics and Automation (ICRA), pp.7-77, 1996. [] R. Ozawa, Y. Takaoka, Y. Kida, K. Nishiwaki, J. Chestnutt, J. Kuffner, S. Kagami, H. Mizoguchi, and H. Inoue, Using Visual Odometry to Create 3D Maps for Online Footstep Planning, In Proceedings of 2 IEEE International Conference on Systems, Man and Cybernetics (SMC), pp. 2643-2648, 2. [6] H. Takemura, A. Matsuyama, J. Ueda, Y. Matsumoto, H. Mizoguchi, and T. Ogasawara, Momentum Compensation for the Dynamic Walk of Humanoids based on the Optimal Pelvic Rotation, In Proceedings of 2 International Conference on Climbing and Walking Robots and the Support
([s]) (b) 1st step (4[s]) (c) 2nd step (1[s]) (d) 3rd step (16[s]) (f) th step (28[s]) (g) 6th step (34[s]) (h) Fall (38[s]) (e) 4th step (22[s]) Fig. 1. Uneven terrain walking with [mm] sole ([s]) (b) 1st step (4[s]) (c) 2nd step (1[s]) (d) 3rd step (16[s]) (e) 4th step (22[s]) (f) th step (28[s]) (g) 6th step (34[s]) (h) 7th step (4[s]) (i) 8th step (46[s]) (j) End (48[s]) 7 7 6 6 Number of Non-Fall Number of Non-Fall Fig. 16. Uneven terrain walking with 1[mm] sole 4 3 2 1 4 3 2 1 1 1 2 2 3 (a) Even terrain walking 1 1 2 2 3 (b) Uneven terrain walking Fig. 17. Number of non-fall while walking [7] [8] Technologies for Mobile Machines (CLAWAR), pp.48-492, 2. S. Kajita, K. Kaneko, K. Harada, F. Kanehiro, K. Fujiwara, and H. Hirukawa, Biped Walking On a Low Friction Floor, In Proceedings of 24 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 346-32, 24. J. Yamaguchi, A. Takanishi, and I. Kato, Experimental Development of a Foot Mechanism with Shock Absorbing Mate- [9] rial for Acquisition of Landing Surface Position Information and Stabilization of Dynamic Biped Walking, In Proceedings of 199 IEEE International Conference on Robotics and Automation (ICRA), pp.2892-2899, 199. P. Sardain and G. Bessonnet, Forces Acting on a Biped Robot, Center of Pressure Zero Moment Point, IEEE Transactions of Systems, Man, and Cybernetics, Part A. Vol. 34, No., pp. 63-637, 24.