Engineering Solutions to Build an Inexpensive Humanoid Robot Based on a Distributed Control Architecture Vítor M. F. Santos 1 Filipe M. T. Silva 2 1 2 Department of Mechanical Engineering Department of Electronics and Telecommunications University of Aveiro, PORTUGAL Centre for Mechanical Technology and Automation TEMA Institute of Electronics Engineering and Telematics IEETA http: //www. mec. ua. pt/robotics
UNIVERSITY OF AVEIRO, PORTUGAL Centre for Mechanical Technology and Automation Institute of Electronics Engineering and Telematics Overview q Introduction q Initial considerations q Mechanical conception q Actuators, power and batteries q Servomotor issues q Sensorial issues and force sensors q The control architecture q Some preliminary results q Conclusions and open issues
UNIVERSITY OF AVEIRO, PORTUGAL Centre for Mechanical Technology and Automation Institute of Electronics Engineering and Telematics Project framework q Motivation Ø Develop a humanoid platform for research on control, navigation and perception. Ø Offer opportunities for under & pos-graduate students to apply engineering methods and techniques Ø The utopia of Man to develop an artificial being with some of its own capabilities… q Why not a commercial platform? Ø Versatile platforms imply prohibitive costs! Ø Reduces the involvement at lowest levels of machine design q Current status Ø So far, it is only a development engineering approach Ø The platform prototype already performs some motion Ø Early studies on control just began
UNIVERSITY OF AVEIRO, PORTUGAL Centre for Mechanical Technology and Automation Institute of Electronics Engineering and Telematics To build or to buy? q Several commercial platforms already exist: Ø Only a few offer great versatility, DOFs, possibilities of control, … Kawada HRP 1 S ZMP Nuvo Sony QRIO Fujitsu Sumo Bot Honda Asimo q Good platforms (e. g. , Fujitsu) have high costs (tens of thousands of Euros); others are not even for sale q Commercial platforms favour mainly high level software development q Developing a platform from scratch allows using hardware more oriented to the desired approach: Ø Distributed control, special sensors, alternative central units … q Developing a platform from scratch takes longer, but hopefully can be done at lower costs…
UNIVERSITY OF AVEIRO, PORTUGAL Centre for Mechanical Technology and Automation Institute of Electronics Engineering and Telematics Initial Considerations q Main Objectives Ø Build a low-cost humanoid robot using off-the-shelf technologies, but still aiming at a fully autonomous platform Ø Have a working prototype capable of participating in the Robo. Cup humanoid league (Germany 2006) q Design Concerns Ø Consider a distributed control architecture due to the expected complexity of the final system Ø Assume modularity at several levels to ease development and scalability Ø Provide rich sensorial capabilities Initial design considerations: − Robot dimensions − Mobility skills − Level of autonomy
UNIVERSITY OF AVEIRO, PORTUGAL Centre for Mechanical Technology and Automation Institute of Electronics Engineering and Telematics The needed DOFs q 6 DOFs per leg Ø Universal joint at the foot (2 DOFs) Ø Simple joint on the knee (1 DOF) Ø Spherical joint on the hip (3 DOFs) • Total DOFs for the Legs = 2 x 6 = 12 q Trunk with 2 DOFs Ø To envisage better balance control q 3 DOFs per arm without hand wrists Ø Universal joint on the shoulder (2 DOFs) Ø Simple joint on the elbow (1 DOF) • Total DOFs for the arms = 2 x 3 =6 q Neck/head accounts for 2 DOFs Ø To support a camera for vision perception q Total proposed: 22 DOFs
UNIVERSITY OF AVEIRO, PORTUGAL Centre for Mechanical Technology and Automation Institute of Electronics Engineering and Telematics Kinematics simulations q Four Denavit-Hartenberg open kinematics chains: Ø One leg on the floor up to the opposite foot not on the ground Ø Starting on the hip, a 2 nd chain goes up to the neck. Ø A 3 rd and 4 th chain for left and right arms. q Allows the analysis of: Ø Static torques Ø Path of Co. M and its projection on the ground q Opens the way to simulate: Ø Dynamics in higher speeds Ø ZMP paths q Inputs for the model in Matlab: Ø Ø Links’ DH parameters Links’ masses Links’ centers of mass Path planning at joint level
UNIVERSITY OF AVEIRO, PORTUGAL Centre for Mechanical Technology and Automation Institute of Electronics Engineering and Telematics Mechanical conception Upper hip Lower hip Final platfrom Head base 3 D model with 600+ components and 22 DOFs Neck Upper leg Trunk Shoulder Lower leg Trunk joints Arm Ankle Forearm Foot Hip
UNIVERSITY OF AVEIRO, PORTUGAL Centre for Mechanical Technology and Automation Institute of Electronics Engineering and Telematics Summary of mechanical properties q Complete humanoid model Ø 22 degrees of freedom Ø Weight - 5 kg Ø Height - 60 cm Ø Max width - 25 cm Ø Foot print - 20 8 (cm 2) q Materials used for the body and accessory parts Ø Aluminium (2. 7 g/cm 3) Ø Bronze (8. 9 g/cm 3) Ø Steel (7. 8 g/cm 3) Ø Nylon (1. 4 g/cm 3)
UNIVERSITY OF AVEIRO, PORTUGAL Centre for Mechanical Technology and Automation Institute of Electronics Engineering and Telematics Actuators q Static (and some simplified dynamic) simulations were carried out to estimate motor torques in a simulated step q Best low cost actuators in the market are Futaba RF servos or similar (HITEC, …). q Available models best suited for our application are: Application Arms & small torque joints Legs & high torque joints Model Mass (g) Torque (Nm) HS 85 BB ~20 0. 35 HS 805 BB 119 2. 26 q Additional mechanical issues for motors Ø Use gear ratios up to 1: 2. 5 to rise torques Ø Use tooth belt systems for easier tuning Ø Use ball bearings and copper sleeves to reduce friction
UNIVERSITY OF AVEIRO, PORTUGAL Centre for Mechanical Technology and Automation Institute of Electronics Engineering and Telematics Power requirements and batteries q Motors Ø Max current: 1. 2 – 1. 5 A per motor (big size model) q Electronics and control Ø Estimated to less than 200 m. A per board with a total of ca. 1. 5 A. q Voltage Levels Ø 5 V for logic; 6. 5 V for motors q Two ion-lithium batteries were installed (from Maxx Prod. ) Ø Ø 7. 2 V/9600 m. Ah per pack Maximal sustained current of 19 A Each pack weights circa 176 g Confined to a box of 37 37 65 (mm 3)
UNIVERSITY OF AVEIRO, PORTUGAL Centre for Mechanical Technology and Automation Institute of Electronics Engineering and Telematics Servomotor velocity “control” q Servomotors have an internal controller based on position Ø User cannot directly control velocity! Ø Either replace motor own control electronics or do some software tricks q Example: Two similar motors with different velocities Ø Dynamic PWM generation Ø Stepped target points Ø Without load, open-loop and feedback based actuation give similar results…
UNIVERSITY OF AVEIRO, PORTUGAL Centre for Mechanical Technology and Automation Institute of Electronics Engineering and Telematics First results with one leg in motion q Simple open loop actuation of the leg joints q PWMs generated by dedicated boards (shown further on)
UNIVERSITY OF AVEIRO, PORTUGAL Centre for Mechanical Technology and Automation Institute of Electronics Engineering and Telematics Envisaged sensorial capabilities Vision unit (on the head) Gyroscopes for angular velocity GYROSTAR ENJ 03 JA from MURATA Potentiometer for position feedback (HITEC Motor) Accelerometers for accelerations and inclinations ADXL 202 E from ANALOG DEVICE Motor electric current Serial power resistor Sensitive feet Strain gauges on a slightly compliant material
UNIVERSITY OF AVEIRO, PORTUGAL Centre for Mechanical Technology and Automation Institute of Electronics Engineering and Telematics The sensitive foot q A device was custom-made using strain gauges properly calibrated and electrically conditioned Ø Four strain gauges arranged near the four corners of the foot
UNIVERSITY OF AVEIRO, PORTUGAL Centre for Mechanical Technology and Automation Institute of Electronics Engineering and Telematics Force sensors and motor connection Servomotor PIC local board + Electric conditioning Foot sensor Servomotor reacts to differences on sensors located on the edges of the foot
UNIVERSITY OF AVEIRO, PORTUGAL Centre for Mechanical Technology and Automation Institute of Electronics Engineering and Telematics Control system architecture Main Control q Distributed control system Ø A network of controllers connected by a CAN bus Ø A master/multi-slave arrangement Ø Each slave controller is made of a PIC device with I/O interfacing. q Asynchronous communications Ø Between master and slaves: CAN bus at 1 Mbit/s Ø Between master and high level controller (currently serial RS 232 at 38400 baud) RS 232 Master CAN BUS Slaves 3 1 3 2 2 1 1 1 3 3 2 2 1 1
UNIVERSITY OF AVEIRO, PORTUGAL Centre for Mechanical Technology and Automation Institute of Electronics Engineering and Telematics Functions of the control level units q Main control unit Ø Global motion directives; high level planning. Ø Vision processing Ø Interface with possible remote hosts q Master CAN controller Ø Receives orders to dispatch to the slaves Ø Queries continuously the slaves and keeps the sensorial status of the robot • Currently does it at ca. 10 k. Hz q Slave CAN controllers Ø Generate PWM for up to 3 motors Ø Interface local sensors Ø Can have local control algorithms
UNIVERSITY OF AVEIRO, PORTUGAL Centre for Mechanical Technology and Automation Institute of Electronics Engineering and Telematics The set of local controllers q 7 slaves controllers for joints and sensors q 1 master controller Ø Interfaces slave controllers by CAN Ø Interfaces upstream system by RS 232
UNIVERSITY OF AVEIRO, PORTUGAL Centre for Mechanical Technology and Automation Institute of Electronics Engineering and Telematics Local control boards q All master and slave boards have a common base upon which a piggy -back unit can add I/O (sensors, additional communications, etc. )
UNIVERSITY OF AVEIRO, PORTUGAL Centre for Mechanical Technology and Automation Institute of Electronics Engineering and Telematics Examples of piggy-back boards q Accelerometers q Serial COM for master q Strain gauges conditioning
UNIVERSITY OF AVEIRO, PORTUGAL Centre for Mechanical Technology and Automation Institute of Electronics Engineering and Telematics First humanoid motion q The robot is able to stand, lean on sides, for/backward q Primitive locomotions have been achieved
UNIVERSITY OF AVEIRO, PORTUGAL Centre for Mechanical Technology and Automation Institute of Electronics Engineering and Telematics Low cost. . . How Low? q Servomotors Ø Big size: ~50 € x 14 -> 700 € Ø Smaller size: ~30 € x 8 -> 240 € q Miscellaneous electronic components Ø Total -> ~300 € q Aluminium gears and belts Ø Total -> ~300 € q Batteries Ø ~80 € x 4 -> ~320€ q Sensors (except camera) Ø Negligible (<100€) q Raw materials (steel, aluminium) Ø Negligible (<100€) q Total ~ € 2000 Ø Excluding manufacturing and development costs (software, etc. ) q Still missing: Ø Vision unit, central control unit (PC 104+), lots of software. . .
UNIVERSITY OF AVEIRO, PORTUGAL Centre for Mechanical Technology and Automation Institute of Electronics Engineering and Telematics On-going and open issues q Next concerns for the platform Ø Joint position feedback from dedicated sensor (not servo’s own!) Ø Safety issues to automatic cut of power on controller failure Ø Better adjustable tensors for belts Ø Selection and installation of central control unit (Embedded Linux) Ø Selection and installation of the vision unit (Fire. Wire. . ? ) q Research concerns Ø Localized/distributed control algorithms Ø Elementary Gait definition Ø. . .
UNIVERSITY OF AVEIRO, PORTUGAL Centre for Mechanical Technology and Automation Institute of Electronics Engineering and Telematics Hints for local control q The difficult relation between planning and stability opens the way to localised control Ø Minimal dependence on planned variables Ø Better adaptation to changing conditions (e. g. , load, ground) Ø Force-based perception is the key issue: • • Reaction forces Joint torques / motor currents q What local control can do Ø Accept a global directive and act locally based on an associated rule. Ø Example: • • Top order: keep standing immune to perturbations Local rules: try to actuate the joints you control in order to keep force balance (e. g. , try to have a distribution of forces as uniform as possible on the foot area)
UNIVERSITY OF AVEIRO, PORTUGAL Centre for Mechanical Technology and Automation Institute of Electronics Engineering and Telematics Conclusions q A highly versatile platform is possible to be built with constrained costs and off-the-shelf components. q The distributed control architecture has shown several benefits: Ø Easier development Ø Easier debugging Ø Provides modular approaches • The generic local controller using piggy-back modules is a confirmation of the modularity q Local controller capabilities include the possibility of localised control based simply on local perception and global directives. q A prototype system has been built and the selected technological solutions ensure a platform for research q A huge field of research issues can be addressed, mainly on control, perception and other autonomous navigation matters.
UNIVERSITY OF AVEIRO, PORTUGAL Centre for Mechanical Technology and Automation Institute of Electronics Engineering and Telematics Author’s Short Biography q Vítor Santos is Associate Professor at the Department of Mechanical Engineering of the University of Aveiro Ø He received his Ph. D from the University of Aveiro in 1994 … Ø His research interests include … Tel: +351 234 370 828 Fax: +351 234 370 953 Email: vsantos@mec. ua. pt http: //www. mec. ua. pt q Filipe Silva is Assistant Professor at the Dept. of Electronics and Telecommunications of the University of Aveiro Ø He received his Ph. D in Electrical Engineering from the University of Porto in 2002; modelling and control of biped locomotion systems Ø His main research interests are centred in the areas of Humanoid Robotics and Healthcare Robotics Tel: +351 234 370 531 Fax: +351 234 370 545 Email: fsilva@det. ua. pt http: //www. ieeta. pt
Скачать презентацию Engineering Solutions to Build an Inexpensive Humanoid Robot
33e5c69f3c5d5b9f70a0ce3f11af4c68.ppt