Internet. Processor board CPU:Geode RAM:64MB. I/O board Radio LAN Compact Flash USB. NiH 24V. USB Hub. Motor controller. Motor driver.

Transcription

1 Architectural Design of Miniature Anthropomorphic Robots Towards High-Mobility Tomomichi Sugihara 3 Kou Yamamoto 3 Yoshihiko Nakamura 3 3 Department. of Mechano-Informatics, Univ. of Tokyo. 7{3{1, Hongo, Bunkyo-ku, Tokyo, 113{8656, Japan. sugihara@ynl.t.u-tokyo.ac.jp Abstract A design methodology to build miniature humanoid robots is discussed. Although light and small bodies would make aggressive types of motion experiments much safer and smoother, they would often cause self-collisions and even restrict the space to mount mechatronic components. In order to defeat some kinematic diculties including the former issue, a technique to modularize and assign joints is proposed through our prototyped robot. And, as a solution against the latter issue, a portable core control unit which stores a stand-alone electronic system is also introduced through the second version of our humanoid, whose system centers around it. 1 Introduction Greatly advanced technology in the eld of robotics have aroused people's expectation that robots will work in the human society. Some have begun to discuss the practical use of humanoid robots. In spite of that, those in the current stage disappointingly lack of mobility against the severity of real environment. Much deeper understand of motion control through experimental studies are required for their evolution to be our alternative bodies. At this starting stage, robots over one meter high [1, 2, 3] would be still threats for us if we shared the life-space with them. It's preferable to begin with a light miniature body for the safety and smoothness of the experiments. A small body, however, causes some constraints on the assembly. Self-collision, for example, could happen so easily that careful joint assignment is required to keep wide motion range of each. And, the space to mount mechatronic components and to wire the cables is restricted. There has been several attempts to build small lightweight humanoid robots. Remote-brained robot approach proposed by Inaba et al. [4] released the robot body from heavy computers. Their bodies consist of tens of small servo DC motor modules originally for radio-controlled toys, and have evoluted to humanoid types[5]. Using those series, Nagasaka et al.[6] realized various whole-body motions. On the same technology, Nordin et al.[7], Furuta et al.[8] and Yamasaki et al.[9] also developed lowcost anthropomorphic robots. However, low precision on the control of those actuators prevents a reliable implementation of dynamic motion controller. Kuroki et al.[10] developed SDR-4X. And, Fujitsu Automation[11, 12] produced HOAP-1/2 for the purpose of oering an open platform to researchers. They developed unique components including motors, gears, ampliers, I/O interfaces and so forth to be integrated in them. A large part of those technology is not revealed. On the other hand, small types of mechatronic products with high specication to be embedded especially into robots have recently become commercially available. Common developers have gained the choice of them for their own systems. This paper discusses a design methodology of miniature humanoid robots to overcome the technical hurdles through two instances we developed with such public components. Kinematic issues including the motion range are rstly dealt with in the prototype, using modularized ortho-axis-coupled joint units. And then, the issue of integration and wiring is solved on the architecture which features AnimatoCore, which is applied to the second model. 2 UT- { prototyped robot 2.1 Specications Fig.1 shows UT-:mighty, the prototype of our miniature humanoid robots. It is about 58[cm] tall, and weighs about 7[kg] including the battery and processor on the body. The total numberofjoints is 23; 3 for the neck are without actuation. 4 located at each arm, and 6 at each leg are actuated by coreless DC motors manufuctured by maxon motor ag., and decelerated with harmonic drive gears by Harmonic Drive Systems. The gear ratios are all 100:1,

2 Figure 1: Outer view of UT-:mighty and the output of motors are 4.5W(for elbow rotational joints), 6.5W(for shoulder abductional joints, elbow exional joints and knee rotational joints), and 11W(for the other joints), respectively. The choice of those high specications mounted on the arms and legs encourages aggressive types of the whole body motion, and also promises feeding the results of experiments back to life-sized robots. Although it causes an increase of weight at each joint, the thinshelled exoskeletal structure, made of magnesium alloy casting, is designed as curved surface for high stiness versus weight, so that the total weight meets the standard weight-per-size ratio. In addition, the total number of parts is reduced, and much higher unit cost of magnesium alloy casting than milling and bending is balanced. 2.2 Ortho-axis-coupled joint unit (A) an illustration (B) gear part Figure 2: Ortho-axis-coupled joint unit Each joint is modularized as a unit with coupled axes, named ortho-axis-coupled joint unit. The idea is basically in accordance with the fact that many of the couples of adjacent axes in anthropomorphic systems are mutually-perpendicular. The whole-body of the robot is assembled, connecting those units with magnesium exoskeletal structures. Fig.2(A) illustrates the unit and the structural connectivity around it. Each axisinvolves a pair of gear and motor. Fig.2(B) is the picture of the gear part. Adopting this technique, 1) the robot keeps maintainabil- Figure 3: Total joint assignment of UT-:mighty (A) hip (B) knee&ankle (C) shoulder Figure 4: Spotlighted unique joint assignments ity, getting attachment-and-detachment of eachjoint to be easy, and 2) the production cycle is accelerated, sharing common parts. Fig.3 shows the total joint assignment of the robot. All the units are located to avoid easy self-collision. Some unique features in it are as follows. i) Hip joint(fig.4(a)) Two axes of the hip joints are assigned so as to enlarge the motion range of exion-extension and abduction. The distance between the two hip joints is countered by inward oset of knee rotation joints to prevent the sideward perturbation of the trunk. ii) Flexional joint at the leg(fig.4(b)) The ankle exional joint is assigned with 5[mm] oset backward to the hip and knee in order to exclude singular posture out of the motion range. And, the mechanical stopper at the knee exional joint saves kinematic energy in the stretched-knee posture. iii) Shoulder joint(fig.4(c)) The root axis of each shoulder is slanted from the perpendicular axis to 45 degrees, aiming at a natural avor on its mixed motion of extension and abduction in a simple way.

3 four corners to calculate the resultant linear force and moment applied to each foot from the ground. 2.4 Experimental operations Figure 5: Postures to examine the motion range - + NiH 24V Coreless DC motor +Encoder DC-DC Motor driver Accelerometer Internet Processor board CPU:Geode RAM:64MB Motor controller Gyro I/O board Radio LAN Compact Flash USB A/D converter 3-axes force sensor USB Hub Fig.7 shows some motion examples of UT-. (A) is a rock'n-roll motion realized on the basis of Sugihara et al.[13], in which the projection of the center of gravity(cog) onto the ground is controlled through manipulation of Zero Moment Point (ZMP[14]). Each transportation of the COG projection from one sole to the other took about 0.6[sec]. (B) is a knee-bend motion. From the upright posture to the bent posture was taken about 0.6[sec]. COG stayed right upon the midpoint of the feet during the motion. (C) is a walking motion generated by an online gait planner [15]. Each step took 0.5[sec], and the step length was about 20[mm]. Through those experiments, we found some immature parts on UT- in terms of both mechanics and electronics. They have been fed back to the second version explained in the following section. 3 UT-2 { animatronic humanoid robot Figure 6: Electronics system of UT- Pictures in Fig.5 are some postures of the robot to examine the performance of the above feature. 2.3 Electronics system The electronics system architecture of the robot is shown in Fig.6. It is an independent system, having the processor board and battery within, and excluding any cables from outer resources which disturbs robot's motion. The processor board is CARD- PCI/GX(EPSON) with a customized I/O board (Fujitsu Automation), featuring Geode GX1(National Semiconductor). Communication with internet is ensured via wireless LAN card(melco), so that it can be remotely operated. It also has the internal USB LAN which branches at USB Hubs(Sanwa supply), and communicates with actuator drivers and A/D converters. The DC motors are controlled by the motor controller imcs01 + amplier imds03(ixs research corp.). The input signal from sensors, which includes the gyro sensor MG2(MicroStone), the accelerometer MA3(Microstone) and the 3-axis force sensor PicoForce(NITTA), are processed at A/D converter imcs03(ixs research corp.). 3-axis force sensors are mounted on the soles and wrists, 4 for each sole and 1 for each wrist. Those on the sole are located at Figure 8: Outer view of UT-2:magnum 3.1 Specication Fig.9 shows UT-2:magnum, our second version. Size, weight and joint assignment of the robot are almost the same with those of the prototype. It has some mechanical improvements that 1) thickness of the exoskeletal shell is controlled, 2) the hip exional joint is reinforced to two-point-support bearings while that of the prototype is supported by one-point bearing as Fig.10 shows, and 3) 22W motor was chosen at the hip abductional joint. The 3-axis accelerometer and gyro sensor are replaced with GSX01001T(Matsukyu Co.,Ltd.) and CRS- 03(Silicon Sensing Systems Ltd.), respectively.

4 (A) A fast rock'n-roll motion (B) A knee-bend motion. (C) A walking motion. Figure 7: Snapshots of example motion experiments 3.2 AnimatoCore The electronic components only except actuators, rotary encoders and force sensors are integrated to be a core control unit, independently from the robot body. We named it AnimatoCore. Fig.11 is its illustration and the outer view. On this technique, cable wiring in the limited space is simplied so that 1) fragile electric circuits are excluded from the extremeties, which frequently suer from vibration and perturbation during the motion (robustness), 2) the risk of self-destruction around mechanically complex parts such as wire-breaking is reduced (suppression of interference), 3) unmount-and-remount operation of components, especially in the early stage of the development, is less troublesome (maintainability), and 4) system modication with increase/decrease of actuators or sensors is easy (extensibility). In addition, it could function as a stand-alone portable controller of various space-saving robotic systems such as mobile robots, multi-legged robots, surgical robots and so forth. One constraint is that this architecture conditions mechanical design to have sizable space to mount the unit. Fig.12 is a diagram of the electronics system of UT-2 centering around AnimatoCore. The processor board is E!Kit-1100(Device Drivers) featuring Au MHz(AMD), 128MB RAM and two CF slots, one of which is for wireless LAN(corega) and the other for storage. In the cource of the development of UT-, we found that USB has some drawbacks for realtime communication, mainly due to the procedural complexity. Then, we chose

5 Processor board Analog Amplifier CUnet I/F board Battery DC-DC converter Motor driver & A/D convertor Figure 9: Joint assignment of UT-2:magnum Figure 11: Image and outer-view of AnimatoCore (A) original version (B) improved version Figure 10: Reinforced hip joint structure. A moving part hatched in the picture is supported by only one point in the prototype, while by two points in the second version (indicated by arrows). CUnet(StepTechnica), which realizes widely-ranged multi-cpu network on shared memory, and constructed the internal communication network between the main processor and distributed motor/sensor controllers. We also developed a small device controller board with H8 processor (Renesas Technology Corp.), two-channel DC motor drivers and eight-channel A/D converters on the cooperation with 3TEC Corp. Fig.13 shows the external view of the board. 3.3 Experiments Fig.14 shows a sequence of bowing, shadow-boxing and walking motions. All motions were navigated by pre-designed trajectories generated by [15]. 4 Conclusion Technical diculties in the design of miniature humanoid robots range from kinematical ones to those in the integration. Our developments proposed 1) joint unit modularization and easy construction of the robot body with them, and 2) an electronic system architecture centering around the core control unit to ease the troubles with integration and wiring in the limited space, and we innovated ortho-axiscoupled joint unit and AnimatoCore. Those products are practically included in two small humanoids, which encourage more aggressive types of motion and the evolution of motor control. Acknowledgments This research was supported partly by the \Robot Brain Project" under the Core Research for Evolutional Science and Technology program of the Japan Science and Technology Corporation, partly by Category \S" of Grant-in-Aid for Scientic Research, Japan Society for the Promotion of Science, and partly by Project for the Practical Application of Next-Generation Robots in The 21st Century Robot Challenge Program, New Energy and Industrial Technology Developemnt Organization. We'd like to make a cordial acknowledgment to Prof. Masafumi Okada, Dr. Ken'ichiro Nagasaka, Ms. Eri Totsuka and many other people. If not for their assistance, UT- series could hardly be completed. References [1] K. Hirai et al. The Development of Honda Humanoid Robot. In Proc. of the 1998 IEEE Int. Conf. on Robotics & Automation, pages 1321{1326, [2] J. Yamaguchi et al. Development of a Bipedal Humanoid Robot { Control Method of Whole Body Cooperative Dynamic Biped Walking {. In Proc. of the 1999 IEEE Int. Conf. on Robotics & Automation, pages 368{374, 1999.

6 Figure 14: Bowing, shadow-boxing and walking motion experiments Internet AnimatoCore - DC-DC + NiH 24V Processor board CPU:AMD Au MHz RAM:64MB Radio LAN Compact Flash Device control board FET counter A/D convertor CPU:H8 CUnet memory Extensible CUnet board Coreless DC motor +Encoder Accelerometer Gyro 3-axes force sensor Figure 12: Electronics system of UT-2 Figure 13: Device controller board [3] S. Kagami et al. Design and Implementation of Humanoid H6 and its Application to Remote Operation, pages 41{50. Experimental Robotics VII, Lecture Notes in Control and Information Sciences 271. Springer, [4] M. Inaba. Remote-Brained Robotics: Interfacing AI with Real World Behaviors. In Proceedings of The Sixth International Symposium of Robotics Research, pages 335{344, [5] M. Inaba et al. Design and implementation of a 35 d.o.f. full-body humanoid that can sit, stand up and grasp an object. Advanced Robotics, 12(1):1{14, [6] K. Nagasaka et al. Acquisition of Visually Guided Swing Motion Based on Genetic Algorithms and Neural Networks in Two-Armed Bipedal Robot. In Proc. of the 1997 IEEE Int. Conf. on Robotics and Automation, pages 2944{2949, [7] P. Nordin and M.G. Nordahl. An Evolutionary Architecture for a Humanoid Robot. In Proc. of the Fourth Int. Symp. on Articial Life and Robotics, [8] T. Furuta et al. Design and Construction of a Series of Compact Humanoid Robots and Development of Biped Walk Control Strategies. In First IEEE- RAS Int. Conf. on Humanoid Robots (HUMANOIDS 2000), [9] F. Yamasaki et al. PINO { The Humanoid that Walk. In First IEEE-RAS Int. Conf. on Humanoid Robots (HUMANOIDS 2000), [10] Y. Kuroki et al. A Small Biped Entertainment Robot Exploring Attractive Applications. In Proc. of the 2003 IEEE Int. Conf. on Robotics & Automation, pages 471{476, [11] products07.html. [12] products09.html. [13] T. Sugihara et al. Realtime Humanoid Motion Generation through ZMP Manipulation based on Inverted Pendulum Control. In Proc. of the 2002 IEEE Int. Conf. on Robotics & Automation, pages 1404{1409, [14] M. Vukobratovic and J. Stepanenko. On the Stability of Anthropomorphic Systems. Mathematical Biosciences, 15(1):1{37, [15] T. Sugihara and Y. Nakamura. A Fast Online Gait Planning with Boundary Condition Relaxation for Humanoid Robots. In Proc. of the 2005 IEEE Int. Conf. on Robotics & Automation, 2005.