This paper gives an overview of the modeling and simulation capabilities of the Servicing Test Facility (STF) developed by DLR in Oberpfaffenhofen. The facility has the main purpose to provide a simulation environment for real-time dynamic simulations, hardware component testing and human teleoperator training. This will give confidence in novel telecontrol technologies and provides the capability to investigate the support needs of the teleoperator in order to perform his task. This paper outlines the general setup of the facility with user interface, computational resources, software architecture and hardware equipment of the laboratory. The main focus lies on the overall software architecture and the capabilities of the simulation facility. Emphasis is given to the presentation of the efficient data flow management between the different processes running on a total of 4 computer systems. In this context, within the simulation environment, the integration of the software tools used is shown. This includes the task of running complex real-time telerobotic operations and simulations commanded interactively by an operator.
Mechanical activities in space covering the areas of in-orbit assembly and servicing as well as experiment handling are associated with high costs, especially if human presence is needed. These operations have to become cheaper with increasing demand of routine operations. A possible solution to this problem is the application of automation and robotics (A&R) for use in space inside and outside of pressurized modules. Besides the expected cost reduction, particularly as a replacement of expensive extravehicular activities (EVA) by astronauts, the high precision and repeatability of tasks performed by robotic devices is one driving force of the increasing importance during the last decade. In the future the use of robotic devices will not be limited to operations in low earth orbit, where they can replace human activities in the hostile environment of space. As already used in some planetary missions, A&R will gain greater importance for missions where human presence is not possible due to programmatic reasons such as one-way journeys, long mission durations or immense costs which cannot be covered by any space budget.
The reasons for use of A&R space applications is outlined above, but up to now only one space robotics experiment developed in Europe has proven its operability. This was the Robotics Technology Experiment (ROTEX) with high on-board intelligence /1/. It was successfully flown in April 1993 on board of US-Shuttle Columbia. The operational modes were: an automatic mode, a telesensor programming mode and a telemanipulation mode. The latter was commanded by either an astronaut on-board or an operator on the ground. The ground-based operation of space manipulators with assistance of sensor devices on the operational hardware will gain more importance in future as it can be utilized not only in environments with human co-presence as in low earth orbit but also in higher orbits unreachable for manned space flight at the moment. Currently there are a number of projects covering ground based teleoperation such as the Spanish minisatellite program (MINISAT / MINIMAN), the German Experimental Servicing Satellite (ESS), the ESA-led study of the Geostationary Servicing Vehicle (GSV) or the Japanese experimental satellite ETS-7. All these projects include use of robotics in either an automated or in an teleoperational mode.
Simulations of space manipulators have to be performed during the early definition and design phase. Due to the light-weight structures they are not limited to kinematic investigations as dynamic effects can occur and may have a significant influence on the behavior of the manipulator. These effects have to be examined by suitable simulations in these early project phases leading to optimum manipulator systems. In addition to an ideal hardware and controller layout, a perfect interaction between operator and manipulator via user interface is essential for successful ground-based teleoperation of space manipulators. Therefore the selection of adequate input devices as well as a predictive simulation and visualization environment are one of the most significant influences on the effectiveness and success of the teleoperational task to be performed. These tasks outlined above are part of the Servicing Test Facility (STF) at DLR in Oberpfaffenhofen /2/. The cooperation of its individual hard- and software components will be described to demonstrate the capability of the facility.
Real-time operations of space manipulators, especially if operated from the ground, have to face the following difficulties:
These difficulties have to be considered carefully as an optimal solution to these problems is the key for successful operations of space-based manipulator systems. The intention of the STF is to provide an opportunity not only for manipulator layout verification but also for the design of adequate user interfaces for remotely operated manipulators as well as operator training. To achieve these goals all of the difficulties outlined above were included in the considerations for the setup of the STF. Thus the facility will provide a realistic simulation of the space-based system to be simulated and controlled from ground. The STF consists of a system of computers capable of simulating the manipulator, the space environment, the communication between the space and the ground segment. In addition the teleoperator interface consisting of input device and visualization is controlled by another computer system. The implementation of an industrial robot gives the opportunity not only to simulate operations but also to perform hardware-in-the-loop simulations for verification and testing of hardware devices.
The basic approach of the STF is to guide and control the space manipulator using computer graphics animation. Within this telerobotic philosophy the space manipulator acts like a slave with the computer simulated manipulator as the master. Since the space robot will react later depending on the time delay due to the transmission time, the software simulated robot will serve as a predictive model. Thus the quality of the simulation depends strongly on the knowledge and implementation of the manipulator system to be simulated. The depth of the model details to be implemented have to be optimized with respect to accuracy on the one hand and computational efficiency on the other hand to fulfill the demands of real-time simulations. These real-time requirements together with the need for efficient user interfaces were the driver for the current set-up of the STF in this pilot phase which consists of:
In addition to these elements combined in the operating environment for teleoperation a software tool for flexible user interface design and implementation is used (TAE+).
The basic components of the teleoperator interface are a hardware input device to control the manipulator and an interface to inform the operator about the actual condition of the system to be commanded. This interface must give the operator the ability to continuously control the position of the end-effector, individual joints or the functions of the end-effector. It is usually realized as a display presenting all informations necessary for remote operation including animated display of the simulated system /3/. Additionally, any other real-time information like individual joint states, forces, moments or data obtained by other sensors in either numerical or graphical format can be displayed. If this information system is screen-based with support of effective design tools a high flexibility is given to adapt the teleoperater interface to different operation tasks and manipulator layouts. For that reason the STF is equipped with the graphical interface builder TAE+ (Transportable Application Environment), originally developed by NASA but now available commercially.
Controlling 6 degrees of freedom is basically possible using different approaches like keyboard input or a combination of different 2D or 3D standard input devices like mouse or joysticks. A better approach was demonstrated in the ROTEX-experiment where a 6D sensor ball took advantage of the 3 translational and the 3 rotational degrees of freedom. In addition to the 6 degrees of freedom the DLR 6D sensor ball used has 8 programmable buttons which enables the operator to easily control the absolute position and orientation of the end-effector (global mode) as well as the individual joint states (local mode).
To ensure an efficient environment for on-line simulation and commanding of the laboratory hardware a distributed computer environment is necessary. These computer systems are connected via Ethernet cable to ensure an effective cooperation to achieve the goal of a efficient simulation environment. The current hardware environment implemented in the STF is displayed in Fig. 1.
Fig. 1: Current Hardware Environment
The computers in use are as follows:
Silicon Graphics (SGI) workstations were selected for the animation display (AD) due to their excellent graphics performance and the existence of a 3D animation system specially suited for robotics (KISMET). For the TI machine the choice of workstation is generally not as critical as for the AD. At the moment a dual-processor SGI computer is in use where the time critical functions are running on the second CPU that is not burdened with operating system tasks. For the simulation computer a more powerful SGI machine was selected to fulfill the needs of on-line simulation of the dynamic multibody systems.
In addition to these computer systems a 6 DOF industrial robot (REIS RV12L) is a significant element of the STF capable of serving as a representation of the relevant flight segment. It enables not only the verification of the proper software simulation by comparing the simulated results with those commanded to the hardware but also the testing of hardware components such as end-effector tools and their connection elements. The laboratory robot hardware is not restricted to the simulation of kinematically identical space hardware but rather to perform end-effector movements as commanded by the operator and simulated for manipulators different in layout.
This equipment together with its modular architecture provides the STF with a high flexibility with respect to individual user or simulation requirements. This is achieved by the possibility to combine the different tasks outlined above into the preferred simulation configuration. The cooperation of the individual elements forming the STF is displayed in Fig. 2. The dynamic possibilities of the STF are listed in Table 1, but the facility could also be run in a pure kinematics mode with or without hardware involvement. However, the primary objectives are investigations and simulations concerning the dynamic behavior of lightweight space manipulators.
Fig. 2: Elements of the Servicing Test Facility (STF)
|
|
|
|
|
|
|
|
|
|
|---|---|---|---|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Efficient and successful simulation and operation of space-based manipulators depend strongly on the ideal integration of the different subsystems mentioned above. The software architecture implemented in the STF is displayed in Fig. 3.
Fig. 3: Command and data flow management implemented in STF
The heart of the software architecture is the Control and Interface Software for Commanding (CISCOMM). This software runs on the second CPU of the TI-computer system and is responsible for most of the communication tasks during a simulation run. The input commands of the operator are picked up by CISCOMM in a definite time interval currently set to 100 ms. This cycle can be changed to adapt to different simulation configurations where a shorter time interval means better simulation results, but also the computational time for each simulation step has to be considered and must not exceed the stepwidth of the cycle. If the facility is operated in global mode, i.e. the position and orientation of the end-effector are commanded, the input values from the 6D sensor ball are transformed to joint positions via inverse kinematics. The formalism of this inverse kinematic is exchangeable and has to be individually developed depending on the configuration of the manipulator to be simulated. If operated in the local mode, the input values are used directly as the joint positions depending on the axis to be moved.
The software performing the dynamic simulation task is the commercially available software package SIMPACK. It is a multibody simulation software that allows off-line modeling of the system to be simulated /4/. This task is performed via an easy-to-use Motif-based user interface, the resulting multibody system is then exported as a program segment using a highly optimized programming code. This code has to be compiled and integrated in the STF as a task running on the Simulation Computer System (SCS). In order to achieve realistic simulation results, not only the position of the joints as obtained by the input device have to be transferred from CISCOMM to CISS (Control and Interface Software for SIMPACK), but also the corresponding velocities and accelerations. These joint state derivatives are obtained using a differential method with the time and the position of the last two cycles. The obtained accelerations, from which the consistent joint positions and velocities can be obtained, are the input values for the integration of the system. The output values from SIMPACK are the result of a time integration where there is a choice of different integration methods suitable for the individual system to be simulated. During implementation problems occurred with the correspondence of the simulated results with the commanded values. This is due to computational inaccuracies and to timing problems of the CISCOMM process. To avoid discrepancies due to computational inaccuracies, not only the position but also the simulated velocities are returned to the controlling process CISCOMM. Here these results are used for the generation of the acceleration commands for the next simulation step. This procedure leads to a closed loop simulation of the system.
After the simulation CISCOMM forwards the obtained results to the CISK (Control and Interface Software for KISMET) process running on the Animation Display System (AD). This process uses shared memory to ensure a suitable integration of the visualization software package KISMET which displays the actual simulated position. This commercially available software enables different views of the system depending on the operators choice and virtual camera positions as needed for the manipulation task to be performed. Using special glasses a user friendly three dimensional visualization can be realized. The received data are the joint positions of the simulated manipulator as well as position and orientation of additional elements to be displayed. This visual display serves as the primary feedback to the operator. If the obtained data are displayed properly, the CISK software requests the next data do be displayed from CISCOMM.
The communication process as described up to here represents a pure simulation mode with no involvement of the laboratory hardware. If hardware-in-the-loop simulations have to be performed, the simulation results are passed from CISCOMM to the CISM (Control and Interface Software for MACOSO) process running on the MLC computer system in parallel to the data exchange with CISK. Unlike the CISK process, the MLC-system does not get the individual joint states but the position and orientation of the end-effector. The data exchange between CISM and MACOSO (Manipulator Control Software) is performed using the shared memory area of the MLC-computer. The individual joint states of the laboratory manipulator are obtained by using an inverse kinematic procedure dedicated to the kinematic layout of the laboratory hardware and implemented in MACOSO. The reason for this second generation of joint positions in addition to the one performed within the CISCOMM process lies in the possibility of different kinematic configurations of the real space manipulator to be simulated and the industrial robot REIS RV12L used within the laboratory. This enables the simulation of nearly every thinkable manipulator layout while the laboratory robot's end-effector exactly behaves like the one of the space manipulator including dynamic effects due to bending and elasticity. These obtained joint states are checked by the collision avoitance software (CiS) running on the second CPU of the MLC computer. In case of a potential collision the laboratory hardware is stopped, otherwise the data are transferred to a control algorithm optimized for the laboratory hardware. This algorithm produces a smooth path of the end-effector trajectory in spite of fixed time steps of the whole simulation environment. The actual joint states are returned from the hardware via CISM to the TI. These data are displayed there to ensure the supervision of the actions of the laboratory robot in comparison to the given commands.
First experiences with the used 6D sensor ball and Man-Machine-Interface (MMI) were gained with the Teleoperations Demonstration Experiment (TDE) between DLR and the Spanish Aerospace Institute (INTA) in Madrid /5/, /6/. During this experiment performed in November 1994 a ground version of the MINISAT/MINIMAN flight segment located in Madrid was remotely controlled from a ground station in Oberpfaffenhofen via the HISPASAT telecommunication satellite. The communication was established using the SLIP protocol over the satellite link. In this teleoperated mode of the remote hardware the STF configuration and the layout of the MMI have proven their functionality.
A near-future space application is expected to arise with the establishment of geostationary servicing satellites, such as the German Experimental Servicing Satellite ESS, which is currently under investigation in Phase A/B with participation of DLR. Due to the lack of a fixed basis, any movement of the manipulator arm can have significant effects on the attitude of the free floating satellite. These effects can be amplified by a possible spin of the servicing satellite. In addition to the potential of STF to simulate not only the complete manipulator system but also the described effects, the possibility of testing and verifying the capture tool and other manipulative hardware devices being connected to the manipulator arm makes it an ideal simulation facility for all kinds of remote manipulation systems.
The setup of the Servicing Test Facility (STF) has proven its capability not only of kinematic but also of dynamic simulation of remotely manipulated systems. This is performed via the integration of commercially available together with individually created software. First test runs have shown that nearly any virtual manipulator system can be simulated while the laboratory hardware provides the operator with corresponding movements of the end-effector.
The future activities will concentrate on a further development of the Man-Machine-Interface (MMI) to ensure an optimum control of the simulated manipulator. This includes not only the display, where higher frame rates are achieved, but also the input device. Further investigations concerning the effectiveness of and alternatives to the 6D control ball will be performed. Even advanced control devices like virtual reality and data gloves are discussed as an operator interface since this will probably improve operating capabilities. Furthermore, the current approach to generate an effective teleoperator interface has to be considered as an prototyping exercise to define the requirements on such an interface. In the future this approach has to be extended with regard to ergonomic aspects to achieve an optimum user-friendly interface. Additionally, the integration of individual manipulator layouts combined with the generation of dedicated software elements like inverse kinematic have to become more effective. Finally an aspect of future work will cover the idea of giving the operator the ability to perform predefined tasks like the replacement of an end-effector mounted tool with another one from a tool-box.