MECHATRONICS STRUCTURE OF THE CENTAUR LIKE WHEELED SERVICE ROBOT. Sami Ylönen, Aarne Halme

Transcription

1 MECHATRONICS STRUCTURE OF THE CENTAUR LIKE WHEELED SERVICE ROBOT Sami Ylönen, Aarne Halme Helsinki University of Technology, Automation Technology laboratory P.O.Box 5400, HUT, Finland, Tel , Fax Classification code according to topics: E3 Abstract: WorkPartner is a new type of lightweight robotic working machine designed mainly for outdoor use. Mobility is based on a hybrid system, which combines benefits of both legged and wheeled locomotion. This system provides good terrain negotiating capability and a large velocity range. The working tool has two hands, like human, which can be used for the manipulation of or handling of tools. The robot is called WorkPartner because the goal is to make an adaptive and learning robot that can carry different tools and work interactively with a human. Copyright 2002 IFAC Keywords: autonomous mobile robots, robot control, robotic manipulators, actuators, electric power systems, energy storage, engine. 1. INTRODUCTION WorkPartner is the next generation of interactive service robots for outdoor tasks. The ultimate goal is a highly adaptive service robot. Some possible work tasks for the WorkPartner: garden work, cutting down trees, guarding, picking up trash, transferring lightweight obstacles, and environment mapping. Art designer's model drawing of the WorkPartner is shown in Fig.1. WorkPartner is a large-scale mechatronic research and development project. It includes a large software development part. The leading idea in managing the R&D work is to make the design as modular as possible in mechanics, electronics and software. The project is public and can be followed on the Web-site: Fig. 1 WorkPartner robot, shown here as the art designers model drawing.

2 GPS-antenna Leg-control unit CCD-camera Navigation unit Main computer Laser pointer Manipulator 2.1 The wheeled leg The wheeled leg, which is shown in Fig. 3, consists of 3DOF mammal type leg and an active rubber wheel. One leg weighs, including the wheel, about 21 kg. It is capable to produce about 70 kg continuous and 100 kg peak force upwards in the nominal driving position. The maximum stride length when walking is about 0,7 m. The leg-wheel mechanism has been optimised for use as a hybrid propulsion device. The main parameters of the leg can be found in Table 1. Engine Body joint Fig. 2 WorkPartner-robot. Elbow-controller Laser scanner In the Fig. 2 are shown subsystems of WorkPartner, which are explained more in the chapters below. 2. PLATFORM The WorkPartner robot is built on a mobile platform called Hybtor (Hybrid Tractor) which is shown in Fig. 2. The platform has four legs equipped with wheels and an active body joint. Each leg has three actively controlled joints and an actively controlled wheel. It weighs about 250 kg and the payload is about 40 kg. The actuation system is fully electrical and the hybrid power system operates with batteries and 3 kw combustion engine. The locomotion system allows motion with legs only, with legs and wheels powered at the same time or with wheels only. With wheels, the machine can obtain 7-km/hour speed on a hard ground. The purpose of the hybrid locomotion system is to provide a rough terrain capability and a wide speed range on variable ground at the same time. Compared to biped human-like mobility system, a four-leg system is statically more stable and easy to control when working with the manipulator. Motion control of the platform is considered in more detail in another papers (Ylönen et al., 2001; Halme et al., 2001). The platform is equipped with a two-hand manipulator system as illustrated in the Fig 2. Fig. 3 Side view of the legs. The leg is controlled in different modes depending on the state of the robot and the needs of controlling the body motion. The modes allow force, velocity and positional control of the joints. In addition, brakes in the motors are used actively to lock the joints when feasible. All these functions are taken care by the leg controller that is commanded from the upper level of the overall control system. The design allows changes of control modes automatically so that the robot can change e.g. from velocity control to force control when stopping moving and starting working. Legcontrol system is considered in more detail in (Ylönen, 2000). Joint Hip (inclination) Thigh (rotation) Angle α [ ] Table 1 Parameters of the leg Max. angular velocity ω [ /s] Max. torque M [Nm] Max. moment of inertia J [kgm2] Links l [mm] ±20 28, , Knee , , , , Wheel ,7 0, (radius)

3 2.2 The wheel of the leg The wheel has two functions, a round rubber wheel works as a foot in the walking mode and as a wheel in the driving mode. The rubber wheel absorbs shocks generated in fast walking. With gear reduction 84,2, the wheel reaches up to an angular velocity of 462 deg/s. This means that the vehicles maximum speed is approximately 7 km/h. Later to improving the speed of the machine, an EC250W motor can be replaced by the stronger 500W motor or the one stage gearbox can be rebuilt into a two stage gearbox. Because of the commercial heavy rubber tire and the wheel disk (mass of the rubber tire and the wheel disk is 6,7 kg), the moments of inertia of the joints are quite high, although the moments of inertia are calculated without rotors of the motors and tooth wheels (see Table 1). When moments of inertia are especially high, it has a negative effect in fast walking. Using the advanced lightweight wheel disk and the optimised rubber tire, the mass of the wheel can be reduced to 4 kg. 2.3 The main actuator The muscles of the machine are identical linear actuators, shown in the Fig. 4. Each of them consists of a Maxon EC250W 48V electric motor, a gear tailor-made by Rover LTD and a ball screw from SKF (CCBR32x100). Inside the rod of the ball screw a bolt s axial tension measuring foil strain gage has been mounted (Fig. 4). With the force sensor, forces can be detected also when the actuator is not actively powered. The main performance values of the actuator are presented in Table 2. Force tests of the actuator. The developed linear actuator with a force sensor has been tested using exterior force sensor (Kistler 9067 force sensor with a charge amplifier 5041b). The sensors were connected to the computer by a DaqBook measurement card. The theoretical force of the linear actuator is calculated as F= current * torque constant of motor * gear ratio * constant of ball screw F=I * 71mNm/A * * 1250(1/m) = 540 N/A * I According to Fig. 5, where the corresponding measurements are shown, the force produced by the actuator is quite linearly related to the current of the motor. Because of a static friction, the force calculated from the current is growing faster that the actual one. The same effect can be observed in reducing force. The actual force stays longer in higher level. The step-shaped force graph is derived from static friction. Comparing the mounted strain gage force sensor with the exterior one, forces observed are quite similar. Only the set point of the strain gage is little erroneous Kistler force sensor -8 Strain gage force sensor Force calculated from current Fig. 5 Force of the linear actuator under changing load. Input of the servo amplifier (Pic 10/100) was sinusoidal signal with frequency 0,1 Hz. 3. MANIPULATOR Fig. 4 The main linear actuator used in WorkPartner robot. Table 2 Parameters of the linear actuator Weight 2,4 kg Length of stroke 100 mm Gear ratio 6,084 Modulus of ball screw 4 mm/revolution Max velocity (no load 70,90 mm/s speed) Max force (continuous) 2500N (I=4,6A) Self-locking force with brake 3042N (0,4 Nm brake) The platform is equipped with a two-hand manipulator system. The simulator skeleton of the robot is illustrated in Fig. 6. Simulator studies are considered in more detail in (Aarnio and Koskinen, 2001). The manipulator is made of aluminium and weighs only about 30 kg. The manipulator can handle loads up to 10 kg. A two-hand human-like and size manipulator was chosen because similarity to human tasks and close co-operation with people are demanded.

4 CAN bus to the main control system of the robot. Working tasks are performed with the help of camera information autonomously or interactively with the operator. Direct teleoperation is used in teaching or when performing complicated manipulation operations. 4. HYBRID POWER SYSTEM Fig. 6 Simulator-picture of the WorkPartner's manipulator. The manipulator consists of a 2-degrees of freedom body, two 3-degrees of freedom arms and a 2-degrees of freedom camera and distance measuring laser pointer head. The manipulator s body is jointed to the platform with two joints, which allow orientation to horizontal and vertical directions. Eight joints of the robot body are equipped with harmonic type gearing, mechanical break and potentiometer. Motor drives are based on DSP processors. Drives are equipped with CAN interface, analog inputs and digital inputs and outputs. With the manipulator installed into the front of the body, WorkPartner looks like a centaur. Instead of humanoid, it may be called centaur, see Fig. 7. The operation time of the WorkPartner robot must be long enough, preferably several hours, to make its services acceptable in outdoor missions. The wellknown problems related to carrying electricity in the on-board energy system is solved by using a hybrid one. The energy is carried in the form of fuel and transformed to the form of electrical power for robot actuation. The system includes a combination of a small lightweight combustion engine with a generator and batteries. The batteries are capable of providing the maximum peak power needed and the generator provides the long-term energy conversion from fuel. The system is controlled with a microcontroller to obtain maximum efficiency. The computer control optimises the energy efficiency by starting the generator when the recharging of the batteries is effective (from fuel to coulombic form) and stops it when it is not. 5. SENSING SYSTEMS Sensor fusion of a passive vision and active laser range finder sensors are used to observe the local environment. The sensors are assembled to the manipulator system of the robot, the camera and distance measuring laser pointer in the head, and the scanning range finder in the chest of the torso of the robot. The motion control system of the robot measures the actuator motor currents in the legs and arms. These signals are used to control forces whenever possible. This reduces the need of direct force sensors. The robot needs to know its body inclination angles in order to maintain its stability on rough terrain. The body roll, pitch and heading angles are used also to estimate the shape of the ground under legs. Measurements are done by inclinometers and gyros. Fig. 7 WorkPartner robot making work task, using the manipulator. With two arms manipulator the robot is to be able to operate with large objects, the sizes of which exceed the arms grippers size. This is important in many service tasks. Motion of the manipulator and platform is coordinated by the overall motion control system. Working tasks require an advanced control system (e.g. force control), therefore the manipulator has a separate control system which is connected via the The perception of the environment is necessary in order to perform working tasks commanded by the operator. It is mainly done by the aid of a laser scanner. It is able to detect obstacles up to 80 meters with directional accuracy equal to 0,25 degrees. The whole horizontal scanning range is up to 180 degrees. The vertical scan is realised by installing the device in the tilting body of the manipulator system. Vision system with colour camera is used to support the recognition of objects in the environment and in the interactive HMI to communicate with the operator.

5 When the environment map is reused for the first time, it must be linked to the global map. This is done in an outdoor environment with the aid of standard GPS. After deactivation of the Selective Availability disturbance, the standard deviation of the position error is decreased to 3 meters. This accuracy together with the speed measurement accuracy, about 0,1 m/s, make it possible to link previously perceived environment data to the presently perceived data from the same area, to set the robot on the map, and to start more accurate navigation. 6. NAVIGATION SYSTEM The navigation system computes the pose (position and attitude) of the robot relative to the objects found in the close working environment. Navigation methods used are based on simultaneous mapping and localisation algorithms, which utilise fused information from inertial sensors, a laser range finder and GPS. The basic navigation task is done by detecting features in the environment and using them as beacons. The camera and laser range finding scanners are used as the perception sensors. The robot also estimates change of its pose by using dead reckoning navigation based on odometer and a compass or gyro. GPS-receiver is used when the signal is available in an outdoor environment. All available data is fused to get the best possible estimate for the current pose. The underlying idea in designing the navigation system is that the robot adapts very soon and automatically, without the help of the operator, to a new environment where it enters to work. Navigation system is considered in more detail in (Selkäinaho et al., 2001). 7. CONTROL SYSTEM The computer system is distributed around CAN-bus as illustrated in Fig. 8. Each leg has one controller based on Siemens 80C167 Micro-controller and PHYTEC mini-module 167. Other nodes, demanding more computing resources - like those taking care of motion and locomotion control, user interface or perception system devices - are based mainly on PC-104 card technology. Additional computer power can be used, via wireless local area network, WLAN. The main computer is a 586 PC- 104 board and is running on QNX operating system. The two-hand manipulator has a separate similarly distributed control system around another CAN-bus. The electronics include both commercial and tailor made digital servo controllers for the actuator motors and amplifier cards for force sensors in legs and shoulder actuators. All the hardware is modular and easy to maintain. In a fault situation, the computer control system can be by-passed and the robot can be driven manually with an extra control box. The manipulator hands can be repositioned manually by releasing the joint brakes. WLAN Main PC- 104-pentium QNX Dual-CAN PLATFORM MANIPULATOR CAN-Network HEAD turning unit Middle controller Ene.sys ctrl CAN-Network Elbow 2 Arm 1 Body Turn Tilt 1 Arm 2 Elbow Fig. 8 Overall schema of the on-board control system.

6 8. HUMAN-MACHINE INTERFACE An efficient and user friendly operation interface is one of the most important functions of the WorkPartner robot. The final goal is to develop an interactive cognitive level interface that utilises voice and gestures as well as symbolic representation as the main communication method. In addition, the user may use simple accessories like a hand held controller with a laser pointer. Cognition means here that the user is able to utilise symbolic presentations and objects existing in the working environment when communicating with the robot. The HMI is designed for operators working parallel to the robot and in some cases, co-operating very closely with it. However, the remote control mode is also possible when the robot is accessible via Internet. The HMI is designed to be multimedia based and highly interactive. Symbolic representation in communication is based on the underlying idea that both the operator and the robot perceive the same environment and interpret it through a commonly understood virtual model. The model is a simplified 3D description of the environment which include the objects relevant for performing tasks. The basic map of the 3D description is done by the aid of a laser range camera. After geometrical mapping, the relevant objects are added to the map by the operator. The second underlying idea of HMI is learning. Similar tasks are performed with less human interaction after learning the sequence of the unit operations needed. The advanced features of the HMI are presently under development and will be mostly realised in the later parts of the program. WorkPartner robot is wirelessly connected both to the user and Internet. Servers in the Internet can serve as knowledge storage shared by many users who are using the same kind of service robot. In the future, the user may find the basic sequence for a new task from Internet rather than teaching it to the robot. Humanmachine interface is considered in more detail in another paper (Suomela and Halme, 2002). REFERENCES Aarnio, P. and K. Koskinen (2001). Using simulation during development of combined manipulator and hybrid locomotion platform. In: Field and Service Robotics FSR2001 (Halme, A., R. Chatila, E. Prassler. (Ed)), pp Yleisjäljennös-Painopörssi, Helsinki. Halme, A., I. Leppänen, M. Montonen and S. Ylönen (2001). Robot motion by simultaneous wheel and leg propulsion. In: 4th International Conference on Climbing and Walking Robots (Berns, K., R. Dillmann (Ed)), pp Professional Engineering Publishing Ltd, London. Selkäinaho, J., A. Halme and J. Paanajärvi (2001). Navigation system of an outdoor service robot with hybrid locomotion system. In: Field and Service Robotics FSR2001 (Halme, A., R. Chatila, E. Prassler. (Ed)), pp Yleisjäljennös-Painopörssi, Helsinki. Suomela, J. and A. Halme (2002). Novel Interactive Control Interface for Centaur-like Service Robot. In: 15 th IFAC World Congress on Automatic Control. Ylönen, S., I. Leppänen, S. Salmi and A. Halme (2001). Hybrid locomotion of a Service Robot. In: Field and Service Robotics FSR2001 (Halme, A., R. Chatila, E. Prassler. (Ed)), pp Yleisjäljennös-Painopörssi, Helsinki. Ylönen, S. Palvelurobotin sähkömekaanisen jalan ohjausjärjestelmä (Control System of the Electromechanical Leg of a Service Robot) (2000). Master's thesis, Helsinki University of Technology. 9. CONCLUSIONS WorkPartner is a highly ambitious research and development project that has many challenges in designing mechanical, mechatronical and software components. WorkPartner is versatile, robust and interactive robot, which works with a human. The best properties of humanoid and centaur robots are united in the WorkPartner. WorkPartner robot integrates in one system almost all the themes and aspects, which are actual today in service robotics, research. WorkPartner project will continue to the end of the year 2005.