An article by:
Dipl. Inf. Dirk Spenneberg
Robotics Group
University of Bremen

The SCORPION

The focusin the SCORPION project is the development of a bio mimetic eightlegged outdoor-capable walking robot. A possible future field ofapplication for robots based on legged locomotion is the work indangerous, highly unstructured, rough and unpredictable environments,where mobility is critical, for example search and rescue missions incollapsed buildings. The mobility of present wheeled and also trackedvehicles is too limited for such tasks. Therefore it is essential todevelop better means of locomotion. The "SCORPION" project deals with building an eight legged all-terrainrobot, which is controlled by an bio mimetic control approach. Thecontrol of these robots is based on models of two biological controlprimitives: Central Pattern Generators and Reflexes. The CentralPattern Generator Model is controlled from a higher central controllevel by means of Rhythmic Motion Patterns (RMPs) and posture controlprimitives (PCPs). With this approach omni directional walking and asmooth and fast crossing between different motion patterns ispossible. Additionally the posture and the speed of the robot can bechanged while walking. The robot were successfully tested in rough,sandy and rocky terrain. One copy of this robot is currently tested atthe NASA Ames Research Center.

TechnicalData

The"SCORPION" has eight legs and measures 65cm from front to back. Thewidth depends on the posture of the legs and varies between 20cm and60cm. In a typical M-shape walking position the robot is 40cm wide.The robot weights 11.5 kg including the 3.0 Ah batteries.

Each leg has 3 degrees of freedom, a thoracic joint for protractionand retraction, a basal joint for elevation and depression and adistal joint for extension and flexion of the leg. The joints areactuated by 24V, 6W DC-Motors. The leg also features a spring elementin the distal segment to reduce the mechanical stress and formeasuring the ground contact force by an integrated linearpotentiometer. The robot senses the following proprioceptiveinformation: the position of each joint, the amount of current drawnby each motor, the tilt in 2 dimensions (pitch, roll) and the load oneach foot tip. Furthermore it is equipped with a compass and a forwardfacing ultrasound distance sensor. For teleoperating the system andfor acquiring data during an experiment, the system has abi-directional communication link for video transmission with a PALCCD camera and for data exchange between the robot and an externallaptop with a communication and control interface. Thus it is possibleto use the robot as a semiautonomous system. The control hardware is onboard and features a Motorola MPC555microcontroller and an XILINX Virtex E FPGA. The MPC555 is used forthe processing of the non-leg sensors and for the behavioural controlof the system. The FPGA is used for the processing of all leg-sensors(position and drawn current of each joint, foot-tip) and for the 20KhzPID-control of the joints.

The biomimetic Control Approach

To describethe dynamics of such a system using a standard kinematical model (asused for industrial robotic arms) is very difficult. For a system withn degrees of freedom at least n coupled equations are needed. Forrotary joints these equations are nonlinear. Computations with such aset of nonlinear equations are very complex, thus this technique isvery unattractive for online computation in autonomous systems.

In contrast biological systems like spiders or insects are very wellable to solve this complex nonlinear problem. Given the fact that theyare using in comparison to current computers a comparatively slowcomputing nervous system, there must be other solutions to the problemof robust walking through rough terrain. Biologist have proposeddifferent models in dependency of their experimentally gained data andobserved animals. What all these models have in common is that theyrely heavily on decentralized control. The control models can bedivided in two main ideas. The first is control based on so called"Central Pattern Generators" (CPG) wherein the actuators arecontrolled by an endogen pattern produced by a central oscillatorymechanism. The second is reflex-driven control where, in contrast tothe first one, the state of the actuator is only a function of theinteraction with its environment. Both approaches have in common thatthey need only a low amount of computational power and thus aresuitable for autonomous walking robots. A pure reflex based controlacts only on the basis of sensory input, whereas the CPG-based controlis able to produce rhythmic motion without the need of sensoryfeedback. Thus in case of false, unreliable or insufficient sensordata (e.g. at high speeds) the CPG-approach is more robust. On theother hand in a very unstructured and dynamically changing environmentthe endogen produced motion pattern might be highly inadequate.

Scheme ofthe Motion Pattern Control Approach

Thereforethe control approach we developed for the "SCORPION" robot aims atusing the best of both worlds. The idea is to have rhythmic motionbehaviours (RMBs), which can be activated at different strengths bythe central control level. The RMBs control the motion of the systemlike a CPG-based system when the disturbances from the environment arerather small. These behaviours simultaneously influence the amplitudeand the frequency of the thoracic, basal and distal oscillators (OST,OSBand OSD). The oscillators are connected to a common clock which isused for local and global (in relation to the other legs)synchronization purposes. The oscillators output is a rhythmic signalwhich can be described with splined sinusoid waves. It describes thetrajectory of the corresponding joint in its angle space. Thus itrepresents the desired behavioural motion, which is translated via themotoric end path into pulse width modulated (PWM) signals to drive themotors.Inline with this rhythmic input to the motoric end path are a set ofperturbation specific reflexes, which are implemented as 'watchdogs'.If greater disturbances are sensed these reflexes (e.g. a "StumblingCorrection reflex") are triggered and override the signals of theoscillators with precompiled motion signals to stabilize the system.The RMBs enable the system to move forward, backward and lateral andto turn with different radii. By activating more than one of these theRMBs' effects can be combined through a overlaying process. Forexample, if the 'move forward' and the 'move lateral left' RMBs areactivated at the same strength the resulting motion is a diagonalforward motion to the left. The overlaying process ensures that thetransition from one motion to another is a smooth and quick motion,thus ensuring that the system keeps stable. An advantage of this isthat the change of the motion direction doesn't require to stop therobot first. In addition to the RMBs the architecture provides the higherbehavioural level with so called Posture Motion Behaviours (PMB) asmeans of control of the posture of every leg. For example the heightand the tilt control behaviours on the central control level are usingthese PMBs to stabilize the system while it is walking. Again theoverlay process combines the influence of the PMBs with the influenceof the RMBs on the actuators. Thus it is possible to change the heightof the main body, while walking, by just outstretching the legs. Allthese mechanisms together enable the system to walk at quite constantspeeds through rough terrain. Thereby it is possible to walk with thesame software architecture over a high variety of different substrateslike rocks, sand, mud, grass, concrete and asphalt. Its maximum speedover flat terrain is half of a body length (30cm/sec). The system isable to overcome obstacles as high as a full outstretched "SCORPION"and it can climb up ramps up to 35% and still overcome smallobstacles, like 8cm high pipes. Furthermore the additional postural control, gives the user inaddition to the rhythmic locomotion the exact control over everyjoint, which can be used for deliberative control. This can be usedfor a wide variety of interesting future behaviours, for example theposture of a single leg can be changed in the way that it can be usedto carry simple objects on the back of the robot. We implemented andsuccessfully tested a behaviour to grab a beam with one leg and thenwalk away with it on the remaining seven legs.Other experiments have shown, that is it possible by only changing theposture of the system and using the unchanged RMB for forward walking,up-side down brachiation along a beam can be executed. The onlydifference between brachiation and forward walking in our approach isthe activity level of the forward RMB on the legs and the posture ofall legs. Thus the locomotion control level provides the programmer with asimple but very powerful interface for locomotion control. For moreinformation on the "SCORPION" robot please contact Dirk Spennebergor take a look at the publications section.



Bitte klicken Sie auf ein Foto für eine größere Ansicht

 
 

More Information