Interactive Virtual Simulator (IVS) of Six-Legged Robot Katharina

Transcription

1 Interactive Virtual Simulator (IVS) of Six-Legged Robot Katharina U. SCHMUCKER, A. SCHNEIDER, V. RUSIN Fraunhofer Institute for Factory Operation and Automation, Magdeburg, Germany ABSTRACT The six-legged walking robot Katharina (fig. 1), developed in Fraunhofer IFF / Magdeburg, is equipped with a supervisory control system and force sensing system to realize various types of locomotion with parallel independent body motion control, and automatic leg adaptation to small irregularities of the surface and providing uniform distribution of support reactions [1,2]. Despite of the presence of the physical prototype it was necessary to develop some tools which allow a real-time simulation of the Fig. 1 Six-legged robot behavior of robot interacting with its environment, an Katharina interaction with user, and user-friendly visualization of the results. Thus, we developed the Interactive Virtual Simulator (IVS) of Katharina. 1 SPECIFICATION OF INTERACTIVE VIRTUAL SIMULATOR To develop an IVS decomposition of the complex prototype into the following mutuallyinfluencing subsystems is carried out: mechanical structure: geometry, DOF, constraints; actuators: size, power conditions, strength requirements; control system: control magnitudes, motion accuracy; feedbacks: internal sensors of the robot, its body scheme (limbs position/speed) and sensors of the environment information (collision sensors, force, tactile sensors, triangulation ranging and others ); interactive environment: 3D modeling of contact and interacting forces.

2 One of the aims of the work consisted in choosing the unified software tools for development and testing of the above mentioned subsystems by the criteria "convenience of use - availability implemented capabilities". The optimum from our point of view is the combination of Matlab (now Release 13) and Blaxxun VRML engine. Matlab consists of a collection of integrated products for data analysis, visualization, application development, system modeling, simulation and code generation. In combination with Blaxxun VRML Matlab provides an engine for real-time visualization and simulation of rigid-body dynamics, collision detection, contact modeling, and collision response. Matlab is unsurpassed for any 3D interactive simulation requiring stable and accurate physics. Matlab/SimMechanics allows programmers to simulate natural behavior of objects in the physical world. By using the Matlab toolboxes it is possible to create interactive 3D simulations of the objects with constraints under any circumstances. The functionality provided by physics simulation software is particularly needed for training and prototyping purposes. 3D models of articulated machinery must also mimic the behavior of the actual machines. The above named systems are cross-platform, available on Windows and Unix-family platforms. 2 USING OF INTERACTIVE VIRTUAL SIMULATOR The main ways to use the IVS are engineering in the loop and operator in the loop. Engineering in the loop : - optimization of the existing or planning prototype construction; - testing, optimization or/and updating of the proposal drive system; - optimization of the control system properties; - development and/or improvement of applied control algorithms with the purpose of expansion of capabilities of the robot (complex service operations, interaction with an unknown environment and/or a ground, complex locomotion and navigation); - analysis of the dynamic processes in the physical prototype, especially in absence of some physical sensor systems. Operator in the loop : The number of adjustable parameters of such mechanism as legged robot (direction, velocity, gait art, step length, body position and orientation, size of automatic avoided obstacles etc.) especially with tools mounted on the robot s body (referred position, orientation, interaction force) is much more than the DOF of the conventional control device (i.e. three-axis industrial joystick with max buttons). Thus, the IVS can be used for the interactive training before the operator controls the real robot. 3 EXAMPLES OF THE IVS USING By dint of IVS the following tasks were realized:

3 Walking with various gaits A basic function of the robot control system is the maintenance of robot gait and realization of transition from one gait to another during movement. The necessity of change of gait can arise from rational use of power/velocity opportunities of robot limbs for overcoming obstacles and even for switching-off of one of the limbs in case of its failure or with the purpose of its use as manipulator. This task can be solved on the basis of information about the robot configuration (i.e. potentiometers, incremental encoders, gyroscopes). The following gaits were realized and tested: wave gait, tripod gait and gallop gait. Locomotion over uneven terrain with automatic avoiding of small obstacles For walking task it is not sufficient to move the legs only forwards and backwards. Because the robot has to move generally on an uneven surface, the task of the generation of gait cycles must be solved without any a-priori information about the surface properties, therefore without precise selection of support points. This task can be solved on the basis of the information about the reaction forces (force sensors build-in in the legs). In IVS were realized the algorithms of legs and body motion control for walking over uneven terrain with adaptation to small roughness of a surface. Locomotion upstairs Walking upstairs is an essential expansion of the robot working area especially for industrial indoor applications and for human indoor robotics, as well for entertainment robotics. This task can be solved on the basis of the information about both the reaction forces (force sensors build-in in the legs) and the robot configuration (i.e. potentiometers, incremental encoders, gyroscopes). In IVS were realized the algorithms for the up-/downstairs walking of the robot. Locomotion with navigation / avoiding of big obstacles Overcoming big obstacles cannot be done automatically without information about the environment/surface. This task can be solved on the basis of the information about the reaction forces (force sensors build in the legs), robot configuration (i.e. potentiometers, incremental encoders, gyroscopes), and first of all about the form and dimensions of the obstacles (video sensors), and also the tactile information for restoration of the obstacle properties (force sensors build in the

4 tools). In case of the partial or whole absence of the needed physical sensor information about environment the IVS can be used for the complete decision of navigation tasks. Robot body stabilization on the moving support area One of the basic advantages of walking platforms in comparison with wheel and track vehicles consists in the more comfortable moving of the luggage or the equipment located on the body. It s obvious when the support surface is not absolutely rigid or when the surface is subject of fluctuations and displacement. With the help of leg control it is possible to stabilize position of the body in the environment. The realized algorithm provides the robot s body stabilization on the moving support area. Independent body motion as a basis for service operations Most research in the area of the development of walking robots and related applications is mainly associated with the ability to walk. But walking alone does not use the whole variety of operations that can be done by such robots. They can also be used as universal body platform for assembly, repair and security tasks. The necessary positioning and orientation of the body can be achieved by controlling the movement of the legs, which stand on the soil. So the controlled robot body can be understood as a platform with manipulator, tool, repair devices etc. installed. Examples of using adaptive body platforms by performing different tasks are: - handling of large building constructions (assembly and disassembly of bridges, road construction, offshore-work etc.); - work within area, which is dangerous for humans; - outdoor works (agricultural work, mine search, inspection and etc.). In IVS was realized the body motion with 3 rotational and 3 translational DOF. Service operation drilling For drilling operations the pressing force determines the cutting force, which has to be controlled. The cutting force has to be constant at all time in the direction of normal force. For avoiding jam and tool breakage the control system can compensate the lateral components of interacting forces, which appear during the drilling operation. Force control of legs in locomotion and motion of body for technological operations were developed and experimentally tested. For the orientation of the drill tool touching the working surface it is necessary to control the body movement in such a way that the longitudinal axis of the drill is

5 collinear to the normal direction of the surface. The coordinates of the point of contact must be constant in time. Data from the force sensor were used for evaluation of the normal direction of the working surface. While the drill touches the working surface the three force components are measured. From this data the normal direction to the working surface can be controlled in such a way that the lateral components of the force vector are going to minimal values. The longitudinal component of the contact force must be parallel to the axis of the drill. Information about the main force and the torque vectors that acts on the body vehicle coming into contact with an external object and maintaining the given contact force, are used for the assembly and drilling operations. Main algorithms are developed and experimentally tested. Cenv Denv 0 influence Cenv Denv x contact Force/Impedance control of the robot interacting with environment/surface In many cases of technical applications of robot systems tasks of co-operation between different mechanical systems must be solved. In walking robots this problem occurs when legs are getting into contact with the soil and a determined interaction of forces is required besides the requirements to be covered by the speed and/or position control system additional criterions for the quality of the mechanism of getting into contact must be fulfilled [3]. It must be emphasized that the work conditions for walking robot are usually unknown, time-dependent and a-priori not linear. The actual simulation work is focussed on optimization of motion of legged robot with constraints under the condition of getting into contact with an object in order to cover given contact quality requirements (minimal tendency to oscillations, contact stability, damping, and contact force limitation), and on the development of an intelligent control system, that enables desired robot compliance under complex conditions. Joysticks with force-feedback In the development of the remote control system for locomotion over uneven terrain and for service operations, two different feedback methods for human-operator are used: visual - feedback and force-feedback. The visual information is not sufficient for a human-operator to estimate the interaction between legs and support surface, as well as between robot body/technology equipment and environment. For successful control of forces and moments acting on the robot and its tools, the operator should have the appropriate feedback. A kinesthetic force reflection feedback can be used to extend the perception of the environment. Two different types of joysticks with force feedback are possible to achieve remote control: isometric and

6 isotonic. With an isometric joystick, control signals are formed due to measurement of the deformations arising in elastic elements of a design. With an isotonic joystick, the control signals are proportional to linear or angular displacements of the joystick. The basic disadvantage of isometric joysticks consists in the lack of information transfer about the acting forces to the operator. Isotonic joysticks allow the creation of an active control system with bilateral effort reflection. Thus, the information about magnitude and direction of force and moment vectors can be transferred to the human-operator [4]. The actual version of IVS is utilized to model interaction between the virtual robot and the virtual environment with forcereflecting of the mechanical properties of the support surface. 4 CONCLUSIONS The main advantages of the Interactive Virtual Simulator (IVS) of the six-legged robot Katharina are the following: IVS helps to carry out the most significant tasks of the Engineering in the loop -area for the simulated robot: construction optimization, investigation of the drive system, optimization of the control system properties, development and/or improvement of applied control algorithms, analysis of the dynamic processes proceeding in the physical prototype. Operator in the loop -modus aids in effective and rapid operator training before real robot control. Development time for the most important tasks such as design, optimization and updating operations are significantly reduced in comparison with using physical prototypes. IVS is universal and can be used for the simulation of various types of robot constructions, drives, and control systems. REFERENCES [1] A. Schneider, U. Schmucker: Force legged platform Katharina for service operations // 4th Intern. Conference on Climbing and Walking Robots, CLAWAR 01, Karlsruhe, [2] U. Schmucker, A. Schneider, V. Rusin, Y. Zavgorodniy: Control signals generation for limbs and body of walking robots // 5th Intern. Conference on Climbing and Walking Robots, CLAWAR 02, Paris, [3] F. Palis, V. Rusin: Self-learning impedance controller // 5th Intern. Conference on Climbing and Walking Robots, CLAWAR 02, Paris, [4] A. Schneider, L. Slutski: Force-reflecting joysticks and cutaneous displays for teleoperated robots // 8thTopical on Robotics & Remote Control, Pittsburg, USA, 1999.