Humanoid Robot HanSaRam: Recent Development and Compensation for the Landing Impact Force by Time Domain Passivity Approach Yong-Duk Kim, Bum-Joo Lee, Seung-Hwan Choi, In-Won Park, and Jong-Hwan Kim Robot Intelligence Technology Laboratory, Dept. of EECS, KAIST, Guseong-dong, Yuseong-gu, Daejeon, 35-71, Republic of Korea {ydkim, bjlee, shchoi, iwpark, johkim}@rit.kaist.ac.kr Abstract This paper briefs on an overview of recent progress and development in humanoid robot, HanSaRam series. HanSaRam (HSR) is a humanoid robot undergoing continual design and development in the Robot Intelligence Technology (RIT) Laboratory at KAIST since. Currently HSR-VI, the latest version developed in 4, is under experiment for on-line walking. In this paper, a method to reduce the landing impact/contact force is proposed for a stable dynamic walking. Humanoid robot might become unstable during the walking due to the impulsive contact force from the sudden landing of its foot. Therefore a new control method to decrease the landing impact force has been required. In this paper, the ground and the foot of the robot are modeled as two one-port network systems which are connected and exchange energy with each other. In addition, the time domain passivity controller, which has the landing impact force as input and foot s position to trim off the force as output, is implemented. The small sized humanoid robot, HSR-VI, is developed to verify the proposed approach through dynamic walking experiments. 1 Introduction Recently, many researches have been focused on a development of humanoid robot, which is similar to human beings [1,, 3, 4]. Current research, which is conducted in collaborating operations with human beings [5], has progressed far beyond studies in walking pattern generation [6] and an online (real-time) balance control [7] during walking. But the standard and the most important function of the humanoid robot is the ability to walk safely in the real environment. Since a legged robot can be unstable during fast walking, one of the essential research topics is to reduce the contact impact force that is created between the foot and the surface during walking. So far several approaches have been established to reduce impact/contact force, which is created from the ground surface. By using a heuristic approach, the method has been introduced by Huang [8] and Silva [9] to shift the foot position once it reaches the surface. However, there are problems in changing the foot position and controlling coefficients voluntarily. Several researchers have studied the hybrid impedance and computed torque control, and the hybrid position and force control for the impedance adjustment of the leg [1, 11]. However in these situations, the complex dynamics of the robot must be known, besides it is difficult to find control parameters. This paper describes an overview of recent progress and development in humanoid robot, HanSaRam (HSR) series. Currently HSR-VI, the latest version developed in 4, is under experiment for on-line walking. This paper also proposes a method to reduce the landing impact force of a humanoid robot. Time domain passivity approach [1] is implemented for this purpose. The robot s foot is modeled as a one-port network system. By calculating the energy input into the one-port network based on the landing force and the foot position, the foot of the robot is controlled to be passive. Unlike previous works, the proposed control method can guarantee the stable dynamic walking without any model information, and requires very little additional computation. The validity of the proposed control method is confirmed through dynamic walking experiments. The remainder of this paper is organized as follows: Section II shows the HanSaRam series. Section III proposes the time-domain passivity controller for reducing the landing impact force. Section IV presents the experimental results with the proposed controller. Finally, concluding remarks follow in Section V.
The HanSaRam Series This section describes an overview of recent progress and development in humanoid robot, HanSaRam series, and presents a control scheme on how to improve its mobility. HanSaRam (HSR) is a humanoid robot that is undergoing continual design and development in the Robot Intelligence Technology (RIT) Laboratory at KAIST since. The aim for developing HanSaRam (HSR) was to participate in HuroSot competition for FIRA Cup (www.fira.net). However, this robot will be a good test bed to test the walking algorithm and the control method due to its small size and light weight..1 HanSaRam-I, II and III In HSR series, HSR-I developed in, which lacked of torque for proper walking and of DOFs for lower body for turning motion, as it consisted of 1 RC servo motors without any sensor feedback. It was controlled by three micro-controllers. They could walk only forward and backward. The walking pattern was a periodic one. Based on the experience of HSR-I, HSR-II was developed to have a shape of a humanlike body. (a) HSR-III (b) HSR-IV Figure : HSR III and IV it into HSR-III. The walking patten planner was useful in generating the periodic gait. But an aperiodic motion planner was needed for the aperiodic motion, which is described in Section.4. Figure 3: The simulation program (a) HSR-I (b) HSR-II. HanSaRam-IV Figure 1: HSR I and II HSR-III was developed in 1 to provide more torques and DOFs as shown in Fig. (a). It had DOFs; 1 DOFs for a lower body using 1 geared DC motors and 1 DOFs for a upper body using 1 RC servo motors, to mimic a human body. But it did not have a sensor feedback system. Fig. 3 shows a walking pattern planner, which generated 3-D gait motion and motor trajectories by applying cubic spline interpolation to representative points such as feet, hip, head and hands. The ZMP stability test, provided in the walking pattern planner, could be done for the generated gait before applying In the walking pattern planner, the ZMP stability could be checked from the generated gait. But HSR- III did not have sensors to measure the ZMP. To test the sensor feedback mechanism HSR-IV was developed in. To measure the ZMP, 4 FSRs were equipped on each sole of HSR-IV s foot, as shown in Fig. (b). HSR-IV consisted of 1 servo motors so that it could make a turning motion. Two micro-controllers were used. One was a master controller and the other was a slave controller. The master controller was used for communication with the host PC and for sensor interface. The slave controller was used for controlling the RC servo motors. The ZMP compensation algorithm was implemented in the master controller [13].
.3 HanSaRam-V HSR-V, developed in 3 as shown in Fig. 4(a), had 8 DOFs and consisted of 1 DC motors for a lower body and 16 RC servo motors for a upper body. Its height and weight were 45 cm and 4.5 Kg respectively. The design concept for the lower body was focused on sufficient torque and zero backlash. Thus the lower body consisted of DC motors and harmonic drives. In the design of the upper body, 16 RC servo motors were used to reduce weight and a simple control method. humanoid by the inverted pendulum, walking pattern could be generated on-line, and moreover turn and stop motions can be easily generated. Since the width between two z-axis of pelvis was designed to be narrow, it could walk properly with less shaking of hip compared with the previous HSRs walking. Moreover, initial positioning of its posture was automatically set up by using a photo interrupter and a revolving disk for the DC motor control. RTLinux was also used for the control of HSR-VI, and 4 FSRs per foot sole were used to measure the ZMP. HSR-VI also had the ability for fully independent sensing, processing, and locomotion. (a) HSR-V (b) HSR-VI Figure 5: Aperiodic Gate Generator Figure 4: HSR V and VI RTLinux was used in the on-board PC, which contained the motion data and executed a high level motion logic. The motion data were provided by the walking pattern planner. The stand-alone vision board was equipped to find out three colors in real time. Six IR detectors were used to avoid the obstacle collision during walking. To measure the ZMP of the robot, 4 FSRs were equipped on each foot sole. Because HSR- V included all computational and power parts, it had the ability for fully independent locomotion..4 HanSaRam-VI HSR-VI, developed in 4 as shown in Fig. 4(b), had 5 DOFs and consisted of 1 DC motors for a lower body and 13 RC servo motors for a upper body. Its height was 5 cm and its weight was 4.6 Kg. The design concept for its lower body was also focused on sufficient torque and zero backlash with DC motors and harmonic drives like HSR-V. The main difference of HSR-VI compared with HSR-V was the design of lower body. It was simplified by designing the harmonic drive and DC motor as a single module. Its walking motion was generated on-line through threedimensional linear inverted pendulum mode [14]. Because it effectively represented the whole dynamics of Aperiodic motions could be generated by using an aperiodic motion planner as shown in Figure 5. By using the planner we could make various motions such as bowing, kicking, hand shaking, etc. 3 Compensation for the Landing Impact Force using Time-Domain Passivity Control 3.1 Review of the time-domain passivity In this section, we briefly review time-domain passivity concept before it is implemented for absorbing the landing impact force. First, we define the sign convention for all forces and velocities so that their product is positive when power enters the system port. Also, the system is assumed to have initial stored energy at t = of E(). The following is widely wellknown definition of passivity which is then used. Definition 1 The one-port network N with initial energy storage E() is continuous time passive if and only if E(t) = t f(τ)ż(τ)dτ + E() t (1)
for force f and velocity ż. Equation (1) states that the energy supplied to a passive network must be greater than negative E() for all time [15]. 3. System modeling Ground z + f - Foot Controller Foot f f 1 PO/PC δz pc z 1 z + Σ + Planner Foot Foot Figure 7: One-port network of robot s foot with PC. z f x z Ground (a) Robot s foot and ground surface. z + f - Ground (b) One-port network models of the robot s foot and ground surface Figure 6: Robot s foot system modeling. To implement the time-domain passivity approach, the robot s foot and the ground are modeled as a network system. Both systems can be modeled as oneport network systems, which are connected each other. Fig. 6 shows the real and the modeled network system, respectively. In Fig. 6(b), the force and the velocity are positive in the upper direction. The sign convention for force and velocity is defined so that the energy is positive when the power enters the system port of the robot s foot. From Definition 1 and using a sampled-time notation of time index j, passivity of the robot s foot system can be described as follows: E(k) = = k F (j 1)(z(j) z(j 1)) + E()() j= k F (j 1) z(j) + E() (3) j= where z is the height position of the foot. 3.3 Time-Domain Passivity Control The proposed time-domain passivity control consists of a passivity observer (PO) and passivity controller (PC), which monitors and controls the input/output energy flow between the robot s foot and the ground. f(= f 1 = f ) is the landing impact force, which can be measured by the force sensors on the robot s foot. z is the difference between two consecutive sampled data of z. The modified foot position z 1 is obtained from the originally planned trajectory (z ) and the output of the PO/PC (δz pc ). z is a planned height position of walking trajectory from the planner, which did not consider the landing impact force from the ground. PO computes the energy flow using the landing force and the foot position as follows: W (k) = W (k 1) + f 1 (k 1)(z 1 (k) z 1 (k 1)) (4) W o (k + 1) = W (k) + f 1 (k)(z (k + 1) z 1 (k)) (5) where W (k) is the total energy output from to k, and W o (k + 1) is the prediction of the one-step-ahead energy output. The last term of (5) is the estimation of the one-step-ahead energy output, which is the output energy from k to k + 1. If the PO can predict whether the system at the next step will be passive or not at the current step k, the PC can modify the desired position at the next step (k + 1) to make the system passive. The PC absorbs exactly the net energy output (if any) measured by the passivity observer at each time sample. Using the PO (steps 4 and 5 below), the following PC algorithm is developed for the one-port robot s foot: 1. f 1 (k) = f (k) is the input;. z 1 (k) = z 1 (k) z 1 (k 1) z (k + 1) = z (k + 1) z 1 (k); 3. z (k) is the output of the one-port network; 4. W (k) = W (k 1) + f 1 (k 1) z 1 (k) is the energy output at step k 5. W o (k + 1) = W (k) + f 1 (k) z (k + 1) is the prediction of the energy at step k + 1
6. The PC output for making the system passive is calculated as follows: { Wo(k+1) f δz pc = 1 (k), if W o (k + 1) <, if W o (k + 1) 7. The modified desired height position (z 1 (k + 1)) can be calculated from z 1 (k +1) = z (k +1)+ δz pc (k). Note that the PO/PC is for achieving the stable landing of humanoid robot. Once the stable landing is achieved, the robot s walking trajectory should be modified to follow the originally planned walking trajectory. The walking trajectory, changed by the PC, is interpolated to the originally planned one by using the polynomial method. 4 Experiments Using HSR-VI, dynamic walking experiments were performed to verify the proposed time-domain passivity control approach. The results were compared with those without PO/PC. In the experiments, the biped robot walked with a speed of 4. cm/s and a step length of 3. cm. The double and single support phases of a step were.15 s and.6 s respectively. The contact started when the foot landed on ground surface. It lasted for.1 s ( sampling times) while the system s energy kept positive. Foot position (cm) Force (Kgf) Energy (Nmm) 1 4 Right foot position without PO/PC Right foot force without PO/PC Right foot Energy without PO/PC -1 - -3 Figure 8: Results without PO/PC. First, Fig. 8 shows the experiment s results which were performed without PO/PC. When robot s foot was landing, there was a big landing force. The maximum value of the right foot s force was 5.7 Kgf and it was 1.4 times of the robot s weight. Even after the robot s foot landed on the ground, it was bounced back from the ground instantaneously due to the big landing force such that it disturbed stable dynamic walking. When the foot kicked the surface, the energy becomes negative, and the robot s foot was no longer passive. It means that the robot might be unstable due to this active energy output from the foot. Foot position (cm) Force (Kgf) Energy (Nmm) 1 4 Right foot position with PO/PC Right foot force with PO/PC Right foot energy with PO/PC 15 1 5 Figure 9: Results with PO/PC. Fig. 9 shows the results when the proposed timedomain passivity approach applied. Since the passivity controller modified the desired foot trajectory to satisfy the passivity condition, foot was slightly moved upward on each landing time, and the foot could absorb the landing impact force. Although the energy became negative during one sampling time, just after the foot was landing, it was recovered to positive immediately with the help of the PO/PC. 5 Conclusion This paper has presented an overview of research development in humanoid robot, HanSaRam. The HanSaRam project was initiated in and since then, six successive versions have been designed, developed and experimentally assessed. The results of the overall experiments indicate that the proposed controller decreases the impulsive landing impact force at the ground surface and makes stable foot landings possible. It is important to note that system dynamic equations are not used anywhere in
the proposed method. Moreover, control parameters for the PO/PC are not required at all. Acknowledgements This work was supported by the Ministry of information & Communications, Korea, under the Information Technology Research Center (ITRC) Support Program. References [1] K. Hirai, M. Hirose, Y. Haikawa, and T. Takenaka, The development of honda humanoid robot, in Proc. of IEEE Int. Conf. on Robotics and Automations, Leuven, Belgium, May 1998, pp. 131 136. [] J. Yamaguchi, A. Takanishi, and I. Kato, Development of a biped walking robot compensating for three-axis moment by trunk motion, in Proc. of IEEE/RSJ Int. Conf. on Intelligent Robots and Systems, vol. 1, Yokohama, Japan, July 1993, pp. 561 566. [3] K. Nishiwaki, T. Sugihara, S. Kagami, F. Sanehiro, and M. Inaba, Design and development of research platform for perception-action integration in humanoid robot: H6, in Proc. of IEEE/RSJ Int. Conf. on Intelligent Robots and Systems, vol. 3, Takamatus, Japan, Oct., pp. 1559 1564. [4] J.-H. Kim, D.-H. Kim, Y.-J. Kim, K.-H. Park, J.- H. Park, C.-K. Moon, K. T. Seow, and K.-C. Koh, Humanoid robot hansaram: Recent progress and development, J. of Advanced Computational Intelligence & Intelligent Informatics, vol. 8, no. 1, pp. 45 55, Jan. 4. [5] K. Harada, S. Kajita, F. Kanehiro, K. Fujiwara, K. Kaneko, K. Yokoi, and H. Hirukawa, Realtime planning of humanoid robot s gait for force controlled manipulation, in Proc. of IEEE Int. Conf. on Robotics and Automation, vol. 1, New Orleans, LA, Apr. 4, pp. 616 6. [7] B.-J. Lee, Y.-D. Kim, and J.-H. Kim, Balance control of humanoid robot for hurosot, in Proc. of IFAC World Congress, Prague, Czech, July 5. [8] Q. Huang, K. Kaneko, K. Yokoi, S. Kajita, T. Kotoku, N. Koyachi, H. Arai, N. Imamura, K. Komoriya, and K. Tanie, Balance control of a biped robot combining off-line pattern with real-teim modification, in Proc. of IEEE Int. Conf. on Robotics and Automation, vol. 4, San Francisco, CA, Apr., pp. 3346 335. [9] F. Silva and J. Machado, Position/force control of biped walking robots, in Proc. of IEEE Int. Conf. on System, Man, and Cybernetics, vol. 5, Nashville, TN, Oct., pp. 388 393. [1] J.-H. Park, Impedance control for biped robot locomotion, IEEE Trans. on Robotics and Automation, vol. 17, no. 6, pp. 87 88, Dec. 1. [11] H.-O. Lim, S. Setiawan, and A. Takanishi, Position-based impedance control of a biped humanoid robot, Advanced Robotics, vol. 18, no. 4, pp. 415 435, 4. [1] B. Hannaford and J.-H. Ryu, Time-domain passivity control of haptic interfaces, IEEE Trans. on Robotics and Automation, vol. 18, pp. 1 1, Feb.. [13] J.-H. Kim, K.-H. Park, J.-S. Jang, Y.-D. Kim, B.-J. Lee, and K.-P. Kim, Humanoid robot hansaram: Schemes for zmp compensation, in Int. Conf. on Computational Intelligence, Robotics and Autonomous Systems, 3. [14] S. Kajita, F. Kanehiro, K. Kaneko, K. Fujiwara, K. Yokoi, and H. Hirukawa, Biped walking pattern generation by a simple three-dimentional inverted pendulum model, Advanced Robotics, vol. 17, no., pp. 131 147, 3. [15] A. J. van der Schaft, L-Gain and Passivity Techniques in Nonlinear Control, ser. Communications and Control Engineering Series. Berlin, Germany: Springer-Verlag,. [6] Q. Huang, K. Yokoi, S. Kajit, K. Kaneko, H. Arai, N. Koyachi, and K. Tanie, Planning walking patterns for a biped robot, IEEE Trans. on Robotics and Automation, vol. 17, no. 3, pp. 8 89, June 1.