DEVELOPMENT OF JUMP ASSIST SYSTEM USING PNEUMATIC RUBBER MUSCLE

Transcription

1 DEVELOPMENT OF JUMP ASSIST SYSTEM USING PNEUMATIC RUBBER MUSCLE Kotaro TADANO*, Hiroshi ARAYA**, Kenji KAWASHIMA*, Chongo YOUN *, Toshiharu KAGAWA* * Precision and Intelligence Laboratory, Tokyo Institute of Technology 4259 R2-46, Nagatsuta-cho, Midori-ku, Yokohama, Japan ( tadano@pi.titech.ac.jp) ** Technology Development Division HQ, NSK Ltd. ABSTRACT In this paper, we develop a jump assist system that can generate instantaneous force using pneumatic artificial rubber muscles (PARMs) since such quick motion is required in the rescue activities. The device extends the knee joint using PARMs. Moreover, bi-articular mechanism between the knee and ankle joint is realized using PARMs. Assist timing with the device is detected with an acceleration meter installed in the device. The timing is investigated both theoretically and experimentally. The experiments are performed with three subjects to demonstrate the effectiveness of the device. KEY WORDS Key words, Pneumatic artificial rubber muscle, Power assist system, Bi-articular muscle, Instantaneous force INTRODUCTION A number of power-assist robots have been developed for amplify human muscle strength. These robots are widely expected to apply for various fields such as medical welfare, rescue, agriculture, physical labor in the factory and so on. Most of conventional wearable power-assist systems propose a method based on the biological signals such as EMG signals or hardness of skin surface [1][2]. However, the noises included in these biological signals make it difficult to identify the motions of the user accurately. Moreover, in the case of using electric motor at each joint, joint parts become heavy and bulky, and need to prepare an electric circuit. Pneumatic actuators have advantages such as high weight-power ratio, compressibility, low heat generation and clean energy. Therefore, the actuators have applied to drive power assist robots. Yamamoto et al. developed a power assist suit for nurse caring with pneumatic actuator utilizing pressure cuffs [3]. Load cells are used to detect the muscles to realized harmonic motion of the robot to the wearer. Pneumatic artificial rubber muscles are often used in the power assist robot since the characteristics of the muscles are well studied [4-7]. In order to support activity of daily living of aged or disable person safety and easily, a wearable power assist device for hand grasping have been developed [8]. By using the rubber muscles, the glove can assist various daily finger tasks owing to its flexibility and light weight [9]. In order to provide muscular support for the manual worker, a muscle suit has been developed. The muscle suit

2 consists of a mechanical armor-type frame and McKibben artificial muscle. Using a new link mechanism for the shoulder joint which consists of two half-circle links with for universal joints, all motion for upper limb has been realized [1]. Most of the power assist systems are focused on relatively slow motion such as walking or lift up motion. However, the systems are expected to use in various situations such as rescue activities. In the activities, quick motion is sometimes required. Therefore, we have intended to assist jumping with a power assist device which is one of quick motion. In this paper, we develop a jumping assist system that can generate the instantaneous force using pneumatic artificial rubber muscle (PARM). A servo control is applied to the system. The effective start timing of the assist detected by an acceleration sensor is investigated experimentally. JUMP ASSIST SYSTEM Design The developed jump assist system composed of an ankle and knee joint. Especially, we focused on the extension of the knee joint since the joint contribute highly for jumping. Figure 1 shows the designed jump assist device. A user wears the device on his/her lower legs. The device mainly consists of three parts, thigh, shin and foot frames as shown in Figure1. Both the user and arm unit are fastened with hook-and-loop at three points, thigh, second thigh and ankle. Pneumatic artificial rubber muscle (PARM) is known as an effective actuator for assist devices since it has flexibility and a high mass-power rate. Therefore, PARM is selected as actuators for the jump assist system. It is known from our previous research [11] that the generated force F by the PARM can be given as F = ap bε cε df (1) here, P indicates the inner pressure, ε is the contraction ration and f is the friction force of PARM. a, b, c and d are the constant parameters obtained from experiments. Joint mechanism Figure 2 shows a side view of the extension mechanism for the knee joint. The joint extend with the contraction of PARM. The extend torque τ gk with the device at the knee joint can be given as τ gk = Fr (2) here F is the generated force by the PARM and r is the radius of the pulley at the knee joint. The PARM is installed at the both side which suggest the knee joint is pulled with two PARMs. The PARM is actively controlled by pneumatic servo valves. In human, bi-articular muscles exist and act on the both end joints simultaneously in addition to the mono-articular muscles. The antagonistic pairs of bi-articular muscles could positively contribute to the compliant properties of the multi-articular extremity, leading to smooth, fine and precise movements [12]. Therefore, in the developed device, bi-articular mechanism is realized as shown in Figure3. The PARM for bi-articular mechanism (shown as bi-articular muscles in Figure 3) is passively controlled at a certain pressure to generate constant pulling force. As the mechanism is so called parallel mechanism, the torque at the knee and ankle joint τ k and τ a have the following relation with the length of the links. PAM τ k τ a = h k h a (3) here, k and a are the length defined in Figure 3. F Front view τ gk r Figure 1 Design of Jump assist device θ k Shin-frame Figure 2 Knee extension Mechanism Foot-frame Toe Thigh-frame r F Side view l t h k l s Center of gravity θ k Thigh-frame Knee Shin-frame F a Ankle Heel Bi-articlar muscle h a Figure 3 Bi-articular muscle mechanism l f h

3 F h [kn] a [m/s 2 ] The following equation is obtained from the balance of torques. τ k + τ a = τ gk (4) The force to the ground F a generated by the device can be written as F a = 1 (h k +h a ) τ gk (5) The photograph of the developed jump assist device is shown in Figure 4. The left photograph shows when a user mounted the device. As can be seen in the figure, the ankle and knee joint is assisted with four PARMs with the diameter of 2 mm. The weight of the device is 7.28 kg with both legs including the actuators. The size and specification of the device is summarized in Table 1. The device is designed to generate maximum torque of 9 Nm at the knee joint. Length [mm] Figure 4 Jump assist device Table 1 Specification of jump assist suit Available angle [deg] Thigh-frame Mono-articular muscle (Knee extention) Thigh-frame 2 Shin-frame 35 Foot-frame 17 Knee joint 117 Ankle joint 125 Maximum torque [Nm] Knee joint 9 Bi-articular muscle Shin-frame Foot-frame Weight [kg] Both legs 7.28 SIMULATION The timing to start assist for jumping is an important issue with the developed device. Therefore, we developed a simulation model and investigated the timing. Generated force by the subject for jumping The kinematic equation when a subject wearing the assist device but with no assist can be written as (m h + m a )a = F h (m h + m a )g (6) here, m h and m a are the weight of the subject and the assist device, respectively. a is the acceleration and F h denotes the generated force of the subject. Using the developed device as shown in Figure4, the experiments of jump with no assist were executed for two subjects to measure the generated force of the subjects F h. A rotary encoder is installed at the knee joint. The acceleration a was calculated by differentiating the position data obtained by the encoder. An example of experimental results is shown in Figure 5. The upper figure shows the calculated acceleration and the movement of the knee joint. The lower figure shows the calculated force F h from Eq.(6) using the acceleration data. The upward direction is defined as positive value for acceleration. The lateral axis shows the time. We selected the initial time s when the acceleration becomes 9.81m/s 2 (1G). It is clear from the upper figure that the subject bends the knee joint to generate the upward acceleration for jump Acceleration Knee angle Human force t [s] Figure 5 Experimental results of measurement of generated force of the subject for jump 5 θ k [deg]

4 E [J] Simulation model The kinematic equation with the assist device can be given as (m h + m a )a = F h + F a (m h + m a )g (7) 8 6 here F a is the generated force by the assist device. The relation between the contraction ratio of the PARM ε and the joint angle θ can be given in the following equation. dθ = L dε (8) r Here, L is the initial length of the PARM and r is the radius of the knee joint. The height from the ground to the center of gravity h as shown in Figure 3 has the following relationship with the knee joint angle. d = ( k + a )dθ (9) k and a are defined as shown in Figure 3. The following equation can be derived by substituting Eq.(4) and (5) to Eq.(3). (m h + m a ) h k+h a L r ε = F h + F a (m h + m a )g (1) The following conditions are assumed for the simulation. 1. The state change of air in the PARM is assumed to be adiabatic condition. 2. The static characteristics of the servo valve used in the experiment are measured in advance. The dynamics of the valve is fast enough compare with that of the PARM, only the statics of the valve is considered. 3. F h is assumed to keep the same value as that of no assist even the assist device generated force. Then, the assist force F a is calculated from Eq.(1) to (5) and (1). The assisted energy E can be given by integrated the assisted force F a until the knee joint extended. E = F a d (7) The simulations were performed by changing the assist start timing. Figure 6 shows the simulation results. The lateral axis shows the start timing. Initial time was as same as Figure 5. The longitudinal axis shows the energies calculated by Eq.(7). The simulations were executed with two subjects those weights were 55kg and 8kg. It is clear from Figure 6 that the most effective assist timing is around 8ms to 125ms after the acceleration becomes 1G. 4 2 Subject A Subject B T [ms] Figure 6 Simulation results (Energy) Table 2 Weight and height of jump Subject A B Wight[kg] 8 55 Height of jump without assist[mm] Calculated height of jump with assist[mm] The calculated jumping height is shown in Table 2 when the energy is totally used for jumping. The results suggest that theoretically the jump height can be increased more than 2% with the developed jump assist device. JUMP ASSIST EXPERIMENTS System configuration The system setup of the assist device is shown in Figure 7. An acceleration sensor installed at the waist of the subject is used for the trigger to start assist. The PARMs for the bi-articular mechanism are controlled at a constant pulling force by the servo valves. PC Pressure regulator Air tank A/D D/A Servo valve Switch Pressure sensor Encoder Mono-articular muscle Bi-articular muscle Pneumatic Artificial Rubber Muscle Figure 7 Configuration of the jump assist system A Acceleration sensor

5 h [mm] The PARMs to extend the knee joints are controlled with servo valves. The valves were full opened depended on the timing. The supply pressure was set at 6kPa abs. The knee angle is measured with a rotary encoder attached on the joint. The measured data were taken into personal computer through a counter. The height of jump was measured using a 3D position measurement system (POLARIS). Therefore, a passive marker was putted on the wrist of the subjects. Experimental results Three subjects were participated in the experiment. Subject A and B are the same as shown in Table 2. The height and weight of subject C is 158cm, 5kg. At first, the relation between the assist timing and the jump height was investigated. The subjects cross their arms in front of the chest and start jump motion from the standing posture. Therefore, arm motion has no effect for jump. Figure 8 shows the results with subject A. The lateral axis shows the assist start timing and the longitudinal axis shows the maximum jump height. It is clear from the experimental results that the maximum height can be achieved with the start timing of 75ms to 125ms. The tendency is almost the same as the simulation results shown in Figure 6. Same tendency can be observed with other subjects as well. Therefore, in the following experiments, we selected the timing to start assist as 1ms after the acceleration becomes 1G. The valves are full opened at the time. Figure 9 shows the photograph during the experiments. The left figure shows when the subject jumped using the assist device. The right figure shows the jumping mounted the device on the legs but without assist. Both figures show when the subject reached at the maximum height. The effectiveness of the jump assist system is clear from Figure 9. The assist system synchronizes well with the subject motion and felt no interference with the device. Figure 9 Experimental scenery during jumping Equiped (no assist) Not equiped Assisted SubjectA Subject B Subject C Figure 1 Experimental results of jumping height Figure 1 shows the experimental results with the subjects. Each subject executed the experiments under three conditions: equipped the assist device but without assist, not equipped the device (free jump), and equipped the device with assist. The subjects jumped five times with the same condition those were shown in the error bars in the figure. The effectiveness of the device is demonstrated from the experiments. Especially, the difference between with and without using the device can be clearly observed. However, the height slightly increased compared with the condition without mounted the device. The generated energy from the PARM was calculated using Eq.(1) and the experimental results. The energy is compared with difference energy between with and without assist. Figure 8 Experimental results of jump assist (Results of changing the timing to start)

6 E [J] Energy of PARM Difference between assist and not assist subjecta subjectb subjectc Figure 11 Effective energy used for jump The results are shown in Figure 11. The left side shows the generated energy with the PARM and the right figure shows the difference between with and without assist. This figure suggests that about 85% of the energy generated by the PARMs was used for jump. The amount is about 85J. The potential energy for example with subject A with 5mm is 392J. Therefore, the generated energy with the device is insufficient. However, the timing of assist became clear with the developed system. The generated energy can be increased by using PARMs with larger diameters which will be our future work. CONCLUSIONS In this paper, we developed a jump assist system that can generate instantaneous force using pneumatic artificial rubber muscles (PARMs) since such quick motion is required in the rescue activities. The device extends the knee joint using PARMs with the diameter of 2mm. Moreover, bi-articular mechanism between the knee and ankle joint is realized using PARMs. Assist timing with the device was detected with an acceleration meter installed in the device. The timing was investigated both theoretically and experimentally. It became clear that the most effective assist timing with the device is around 8ms to 125ms after the acceleration of the subject becomes 1G. The experiments were performed with three subjects to demonstrate the effectiveness of the device. REFERENCES 3. Yamamoto K., Hyodo K., Ishii M., Matuo T., Development of Power Assisting Suit for Assisting Nurse Labor, JSME international journal. Series C, 22, Vol.45-3, pp Schulte R. A., The Characteristics of the McKibben Artificial Muscle, In The Application of External Power in Prosthetics and Orthetics, publ.874, Nas-RS, 1962, pp Chou C. P. and Hannaford B., Statics and Dynamics Characteristics of McKinbben Pneumatic Artificial Muscles, IEEE International Conference on Robotics and Automation, 1994, pp Caldwell D.G., Gustavo A. Medrano-Cerda, Goodwin M., Control of Pneumatic Muscle Actuators, Control Systems Magazine, IEEE, 1995, Vol.15 1, pp.4 48, Klute G.K., Hannaford B., Fatigue Characteristics of McKinbben artificial muscle actuators, IEEE/RSJ 1998 International Conference on Intelligent Robots and Systems (IROS1998), 1998, Vol.3, pp Aragane M., Noritsugu T., Takaiwa M., Sasaki D., Naomoto S., Development of Sheet-like Curved Type Pneumatic Rubber Muscle and Application to Elbow Power Assist Wear, Journal of the Robotics Society of Japan, 27, Vol.26-6, pp Sasaki D., Noritugu T., Takizawa M., Yamamoto H., Wearable Power Assist Device for Hand Grasping Using Pneumatic Artificial Rubber Muscle, IEEE International Workshop on Robot and Human Interactive Communication, 24, pp Kobayashi H., Suzuki H., Nozaki H., Tsuji T., Development of Power Assist System for Manual Worker by Muscle Suit, IEEE International Conference on Robot and Human Interactive Communication, 27, pp Sasaki T. and Kawashima K., Development of Remote Control System of Construction Machinery for Rescue Activities with Pneumatic Robot, Advance Robotics, 26, Vol.2, No.2, pp Kumamoto M., Oshima T., and Yamamoto T., Control properties Induced by the Exsitence of Antagonistic Pairs of Bi-articular Muscles- Mechanical Engineerging Model Analyses, Human Movement Science, 1994, Vol.13, pp Kawamoto H., Sankai Y., Power assist method based on phase sequence driven by interaction between human and robot suit, Robot and Human Interactive Communication, 13th IEEE International Workshop, 24, pp Tsukahara A., Kawanishi R., Hasegawa Y. and Sankai Y., Sit-To-Stand and Stand-To-Sit Transfer Support for Complete Paraplegic Patients with Robot Suit HAL, Advanced Robotics, 21, Vol. 24, No. 11, p