A Simulation Architecture For Model-based Systems Engineering and Education

Transcription

1 A Simulation Architecture For Model-based Systems Engineering and Education 1 Quoc Do, 2 Todd Mansell, 1 Peter Campbell and 1 Stephen Cook 1 Defence and Systems Institute University of South Australia Mawson Lakes Campus, South Australia. Ph , fax Quoc.Do, Peter.Campbell and Stephen.Cook {@unisa.edu.au} 2 Defence Science and Technology Organization Organisation, Edinburgh, South Australia. Ph Todd.Mansell@dsto.defence.gov.au Abstract. Traditional systems engineering is based on a document-centric approach in that it is underscored by a central system design, that may be expressed as a suite of mental models that capture the customers capability expectations, system requirements, architectural design, design implementation decisions, system verification and validation, and system acceptance in static document form. The reality is that large system design and development is a dynamic and adaptive process that can be better executed and understood by employing a federation of executable models that can respond to the dynamics of the processes involved giving rise to the need for Model Based Systems Engineering (MBSE). Microcosm is a joint program between the Australian Defence Science and Technology Organisation (DSTO) and the Defence and Systems Institute (DASI) of the University of South Australia that is being designed to explore MBSE concepts, foster research and development in MBSE and in systems engineering education. Microcosm uses real and simulated unmanned ground robotic platforms with a suite of real and simulated sensors operating in both real and simulated environments such that any sensible configuration of real or simulated components can be used in a defined experimental scenario to achieve these goals. This paper describes the architecture and initial implementation of the simulation system of Microcosm with a focus on the use of an agent-based approach to enable the abstraction of behavior that provides the flexibility required by the simulation system. The outcomes of initial trials will be discussed. 1. INTRODUCTION Traditional systems engineering is based on a documentcentric approach in that it is underscored by a central system design, that may be expressed as a suite of mental models that capture the customers capability expectations, system requirements, architectural design, design implementation decisions, system verification and validation, and system acceptance in static document form. The reality is that large system design and development is a dynamic and adaptive process that can be better executed and understood by employing a federation of executable models that can respond to the dynamics of the processes involved giving rise to the need for Model Based Systems Engineering (MBSE). Microcosm is a joint program between the Australian Defence Science and Technology Organisation (DSTO) and the Defence and Systems Institute (DASI) of the University of South Australia that is being designed to explore MBSE concepts, foster research and development in MBSE and in systems engineering education. Microcosm uses real and simulated unmanned ground robotic platforms with a suite of real and simulated sensors operating in both real and simulated environments such that any sensible configuration of real or simulated components can be used in a defined experimental scenario to achieve these goals. The architecture we have chosen for the simulation component of Microcosm is an object-oriented, agentbased design which most readily supports the requirement that users in both the teaching and research areas be able to quickly and easily setup the connections between various components needed to execute a selected scenario. The use of software agents has been identified as a good approach for assisting in solving compatibility issues that arise in simulation systems built from components (Boloni et al. 2000). The underlying design idea is that each component has an abstract set of behaviours which can be implemented by one or more different models (Christiansen 2000). For Microcosm there will be a minimum of two such models one which is the actual physical system, the other which is the simulation component that corresponds to the physical component. There may be more than one such simulation component, depending on the scenario chosen, this capability being an attractive element of the design. In addition to the models corresponding to the physical components of Microcosm, the simulation will also have a representation of the environment that the robots and sensors operate within. Initially this will provide a replication of the environment description used to operate the physical components, but by using the developed architecture it s possible to use other environmental models that stimulate the system with different simulated characteristics. An important element of the architecture is the concept of a context manager that allows users to specify the

2 components that they wish to use in a given scenario and then allow the simulation setup software to select the appropriate interfaces and make all the necessary connections. We view this capability as a key element for the successful use of Microcosm, but also as a key requirement for the successful application of agent-based methods to MBSE in general. A high level view of the architecture is shown in Figure MICROCOSM DESCRIPTION The initial focus of the Microcosm project is (1) research into the application of Model Based Systems Engineering (MBSE) in complex SE&SI programs, and (2) establishment of a practice based SE&SI research and training environment (Shoval et al. 2008). Central to the Microcosm project is the Microcosm sandpit environment where the physical elements (humans, computers, unmanned vehicles, deployable sensors, laboratories, and simulators) are housed. The Stage 1 implementation of the Microcosm sandpit (delivered Dec 2008) involved the procurement of unmanned ground vehicles, ground based sensors, communications infrastructure, a mission control segment, and a modelling and analysis environment. Microcosm is a fully instrumented open-system that employs a customised systems engineering process based on the Defence Capability Development Cycle utilising the Spiral development model to foster MBSE research and education. This Sandpit is to be used by stakeholders as a facility to stage demonstrations, conduct research in MBSE within a defence context, evaluate systems configuration and operation, and investigate process improvement (Cook et al. 2008). Most importantly, this Sandpit creates an environment where the challenges and pitfalls of complex systems integration can be learnt (often through careful crafting of credible engineering scenarios.) It is planned that Microcosm will be used as a focus for collaborations with pertinent programs nationally and internationally. It is the intention to engage with similar practiced based SE&SI programmes such as the Systems Engineering Innovation Centre (SEIC), at Loughborough University in the UK. SEIC has established a similar environment as part of its ConSERT 1 program. As a result, DSTO, UniSA and the SEIC are discussing opportunities to share environments, research initiatives and potentially collaborative systems engineering and systems integration programs. 3. ARCHITECTURE REQUIREMENTS The Microcosm facility has been developed to explore innovative approaches to Systems Engineering and Systems Integration (SE&SI) practice, initially by addressing six use-cases: Simulation and Analysis, Human-Agent-Based Modelling for Systems Engineering, Education Activities, Autonomous Vehicles Research, Systems Engineering Approach to Model Development, and Systems Enhancement Research. These use-cases were reported in (Mansell et al. 2008). 1 The architecture is required to be an open architecture, object-oriented and supports the plug- and-play paradigm for ease of system evolutionary development, where system components consist of real and synthetic counterparts. The architecture supports a number of complex military scenarios, which are assembled from either real or synthetic components, or a combination of real and synthetic components. Each scenario is described by a set of DAF views (DoDAF 2007; MoDAF 2008). This ensures the compatibility of system s description and representation with defence customers. 3.1 DAF Views of the Microcosm Stage One Operational Scenario The operational view (OV1) for the Microcosm stage one operational scenario is showing in Figure 1. It consists of two unmanned ground vehicles (Pioneer 3DX robots), and two fixed global external sensors: vision and SICK LMS 291 laser sensors. Each robot has a sensor suite, which consists of a laser, ultrasonic, compass, odometry and vision sensors to perform specific tasks within a given scenario. Figure 1. Operational view (OV1) of the Microcosm stage one scenario. The OV1 is extended to achieve a lower level of abstraction, representing the system using a set of system views (SVs). There are eleven DAF system views but SV2 and SV5 are most suitable for adequately representing the Microcosm operational scenario. The SV-2 diagram represents the Microcosm system as a collection of nodes and the information flow between them, which is illustrated in Figure 2. There are five nodes: two fixed sensor nodes, a node for each robot, and a ground station. Figure 2. System view (SV-2) of the Microcosm stage one operational scenario

3 System view (SV-5) on the other hand, represents the system from a functional point of view. This is achieved by a system functional follow diagram as shown in Figure 3. The Microcosm system has a Master-Runtime Control and multiple robots, denoted R 1 to R n. The Master- Runtime Control senses, processes and fuses information from global sensors, collects data from the robots to create and maintain an operational picture and performs intruder detection. It instructs the simulation environment to simulate the robots and intruder motions, and sends the intruder s position to the robots for tracking and intercepting. The robot performs path planning and tracking based on it s current position (determined by onboard odometry) and the intruder s position, and generates and executes motion commands. 5. THE DESIGN ARCHITECTURE The Microcosm Sandpit high-level architecture has three distinct parts: Microcosm Information Management System (MIMS), Modelling and Simulation Control System (MASCS), and the Microcosm Physical System (MPS). This is further depicted in Figure 4. The MIMS is an integrated information management system that stores all the systems engineering products associated with the project through each spiral-development cycle. The SE products of the previous stage, especially the lessonslearnt, and system s capabilities versus its constraints are critical for the Microcosm facility to evolve to meet Defence s demand for systems integration and MBSE research and development. The MASCS is a simulation and control subsystem that contains synthetic models of Microcosm s components including environment models, simulated autonomous vehicles, and a suit of onboard and off-board sensors. Control algorithms are also being developed to provide simulation capabilities for various defence operational scenarios, where aspects of Microcosm s performance improvement due to a sensor or subsystem upgrade can be investigated. The system also provides the capability for hardware-in-the-loop (HIL) simulation through the use of a common interface between simulated and physical components that provides seamless interactions between these components in a given operational scenario. Finally, the MPS consists of all the physical components of the Microcosm facility, that is the autonomous mobile robotic vehicles and external sensors. Figure 3. System view ( SV-5) system functional flow diagram. 4. APPLICABILITY OF THE AGENT BASED APPROACH FOR MBSE As discussed in the introduction, agent-based modelling offers many advantages for the development of MBSE tools. These advantages include: The ability to federate models of heterogeneous types within a single simulation structure, and arising from this, the ability to federate models of physical processes with those for behavioural processes, The ability to employ different models for the same function, depending on the context of the simulation, leading to flexibility in simulation approach, and also in the granularity with which various components are represented, again depending on context, and The ability to employ an architecture that allows the incorporation of existing models of the components of a system without major changes to the simulation tool s structure. Figure 4. The high level architecture of the Microcosm Sandpit. The conceptual design of the Microcosm is shown in Figure 5. In the figure the largest box represents the physical environment within which the real components of Microcosm act. The robot, sensor and effector objects can each be instantiated as real objects in the Microcosm domain. The environment object will either be the specification of the real environment locations of walls, obstacles and beacons, etc. or a model of some other environment as called for by other scenarios. Such a scenario might then have the robots and sensors perceiving that they are acting on a sea surface, rather than a concrete floor. Finally, the ad hoc controller object can be instantiated either by a real person, or by an agent based model of a person. The objects, real robot, real sensor and real effector, are instantiated in the simulation system either as software that is supplied with the physical equipment or as bespoke software that has been written by the Microcosm project team. Microcosm is designed to run in an autonomous mode with the human acting as an observer, or as an observer who can issue instructions to the robots,

4 sensors or effectors in response to the signals received from them. The corresponding agent-based model will be built to respond to the same information. independently, pass messages and run in parallel within a single process. The DSS on the other hand extends the CCR capability across processes and machines. Both CCR and DSS are provided within the Microsoft Robotic Development Studio (MRDS) (Johns and Taylor 2008). The Microcosm facility software service-oriented architecture is depicted in Figure 6. All entities in the architecture were implemented as services. Each service has one or multiple sub-services. Communications between services were low-weight message-passing, which support both the pull-push, and publish-andsubscription methods. All physical components in the system have their synthetic counterparts in the form of services. Synthetic services were controlled by the Simulation Orchestrator, which acted as a gateway to interact with physical hardware components via the Master Runtime Control. Hardware services (device drivers) were controlled by the RobotControl1 and RobotControl2, both interacting with the simulation environment via the Mater Runtime Control. Figure 5. The modeling and simulation control system (MASCS) architecture. The user interface runs the Microcosm environment through a context manager object. The context manager is a rule-based component which allows a user to select the various options that go to comprise a scenario and makes the necessary connections between the components, while at the same time preventing connections that are not supported. For example, a user will be able to select which real components and which simulated components are to be used for a given run of the Microcosm system and the software will then automatically assemble and connect all the required components. Information about all aspects of the system is assembled and held in the database. The context manager will contain the meta-information needed to allow rapid assembly of the environment according to user choices and is essential to support the rapid setup of different scenarios to meet both teaching/learning and research needs. Finally, the interface objects allow the plug-and-play of either the models of the system components or the real software modules managing actual data from the hardware. 6. SYSTEM IMPLEMENTATION AND RESULTS 6.1 Service-oriented Architecture The Microcosm Sandpit stage one system implementation was based on a service-oriented architecture. It has been implemented using the Decentralised Software Service (DSS) (Nielsen and Chrysanthakopoulos 2008) and the Concurrent Control Runtime (CCR) (Morgan 2008) from Microsoft. The CCR is a managed library that provides classes and methods for concurrency, coordination and error handling. It enables segments of code to operate Figure 6. Service-oriented architecture for the Microcosm stage one implementation. 6.2 Microcosm Stage One Hardware Configuration The Microcosm Sandpit stage one was intended to be used by stakeholders as a facility to stage demonstrations, conduct research in MBSE, evaluate systems configurations and operations, and investigate process improvement (Cook et al. 2008). The physical system setup was based on the operational scenario described in Figure 1. The hardware decomposition is presented in the SV2 diagram showing in Figure 7. Each robot had a suite of sensors that includes the Hokuyo laser and ultrasonic rangefinders, wireless GigE camera, wheel odometers, magnetic compass, wireless Bluetooth and wireless LAN. The ground station hosted the simulation environment, the Master-Runtime Control, and the vision system, which were located in the Desktop computer, Laptop VI, and Laptop III respectively.

5 architecture shown in Figure 5. The Master Runtime Control was responsible for intruder detection, sensor fusion, creating and maintaining an operational picture, and requesting positional update from the robots. It also instructed the Simulation Orchestrator to emulate the robots and intruder motions. Figure 7. System view (SV2) hardware decomposition of the Microcosm stage one physical hardware setup 6.3 Simulation Environment The simulation environment was implemented using the simulation capability of the Microsoft Robotic Development Studio (MRDS). It uses the AGEIA PhysX physics engine to simulate physical interactions within the virtual environment, such as fiction, gravity and collision etc (Johns and Taylor 2008). The Microcosm Sandpit physical environment and the robots were modelled to scale in the simulation and sensor accuracy was modelled with only a low-level of fidelity. The modelled environment is showing in Figure 8. The synthetic robots were configured to replicate the motions of their real counterparts in the real environment, performing the operational scenario depicted in Figure 1. A synthetic intruder was also modelled and its motion was emulated based on the intruder s position calculated by the laser sensor in the physical environment. Intruder and robots motion data was supplied to the simulation environment by the Master-Runtime Control. 6.4 Real-time Demonstration Stage one of the Microcosm Sandpit has been successfully tested through real-time demonstrations for the Defence customer. The demonstration was based on the operational scenario depicted in Figure 1. The physical setup is shown in Figure 9. In the scenario, Robot1 was tasked to follow and intercept an intruder upon detection. Robot2 was tasked to pivot at a fixed location and track the intruder using the onboard camera. The ground station housed the Master-Runtime control, vision system and the simulation environment, which were implemented on the Laptop III, Laptop VI and Desktop computers, respectively. Stage one of the Microcosm Sandpit employed a centralised control architecture, as depicted in Figure 6. The Master-Runtime Controller was considered as an instance of the Controller Entity in the general Figure 8. Aerial view of the synthetic environment of the Microcosm Sandpit. Figure 9. Aerial view of the Microcosm stage one physical system setup. Upon intruder detection, the robots received the intruder s position from the Master-Runtime Control. Robot2 pivots and tracked the intruder using the onboard camera, while Robot2 performed its own path planning to follow and intercept the intruder using a finite state machine, which is illustrated in Figure 10. After the Initialisation state it transited automatically into the Standby state, upon intruder detection it progressed to the Follow-Intruder state, in which it performed path planning and followed the intruder. While in the Follow- Intruder state, if an obstacle appeared in the way (detected by the onboard laser sensor), it transited to the Obstacle-Avoidance state, and resumed when the obstacle was cleared. The robot stopped within one meter in front of the intruder, and transited to the Return-To-Base state if the intruder vanished. The robot was equipped with text-voice capability, producing a voice message every time the robot moved from one state to another for ease of understanding in the real-time demonstration.

6 7. SUMMARY Microcosm is being developed as a tool for teaching and researching aspects of MBSE by means of demonstration and experiential learning methods. Our goals require that we have a tool with considerable flexibility that can be easily configured to represent a large number of different scenarios, for either purpose. In the case of the teaching applications there is also the requirement that Microcosm be easy for students to setup within the time limits imposed by class work. To this end we have designed an agent based simulation architecture that will provide the context responsive system setup procedures to meet these goals through the use of a context manager that operates on a set of flexibly executed agent models that represent the several components of the Microcosm system. Mansell, T., P. Relf, S. Cook, P. Campbell, S. Shoval, Q. Do and C. Ross, Microcosm- A Systems Engineering and Systems Integration Sandpit. Asia- Pacific Conference on Systems Engineering - APCOSE, Japan. MoDAF Link: Morgan, S Programming Microsoft Robotics Studio, Microsoft Press, US. Nielsen, H. F. and G. Chrysanthakopoulos "Decentralized Software Services Protocol DSSP/1.0." Link: E8-494A-BB8C-3D49850DAAC1/DSSP.pdf. Shoval, S., A. Hari, S. Russell, T. Mansell and P. Relf, Design of a Systems Engineering Laboratory Using a Scenario Matrix. Conference on Systems Engineering Research, Los Angeles, USA INCOSE. Figure 10. Robot one state transition diagram. 8. ACKNOWLEDGEMENTS The authors would like to acknowledge the funding and in-kind support from the Defence Science and Technology Organisation (DSTO), ref no: REFERENCES Boloni, L., D. C. Martinescu, J. R. Rice, P. Tsompanopoulou and E. A. Vivalis " Agent based simulation and modeling." Concurrency: Practice and Experience V12:pp Christiansen, J. H., A flexible object-oriented framework for modeling complex systems with interacting natural and societal processes. 4th. International Conference on Integrating GIS and Environmental Modeling (GIS/EM4) Problems, Prospects and Research Needs, Banff, Alberta Canada. Cook, S., P. Campbell, S. Shoval, S. Russell, Q. Do and T. Mansell, Microcosm: A Systems Engineering Educational Environment. Systems Engineering Test and Evaluation Conference, canbera, Australia. DoDAF Link: Johns, K. and T. Taylor Professional Microsoft Robotics Developer Studio Wiley Publishing Inc.