History

A Framework for Simulation and Testing of UAVs in Cooperative Scenarios

Description
A Framework for Simulation and Testing of UAVs in Cooperative Scenarios
Categories
Published
of 23
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
Share
Transcript
  J Intell Robot Syst (2009) 54:307–329DOI 10.1007/s10846-008-9268-8 A Framework for Simulation and Testing of UAVsin Cooperative Scenarios A. Mancini · A. Cesetti · A. Iualè · E. Frontoni · P. Zingaretti · S. Longhi Received: 15 March 2008 / Accepted: 30 June 2008 / Published online: 21 August 2008© Springer Science + Business Media B.V. 2008 Abstract  Today, Unmanned Aerial Vehicles (UAVs) have deeply modified theconcepts of surveillance, Search&Rescue, aerial photogrammetry, mapping, etc. Thekinds of missions grow continuously; missions are in most cases performed by afleet of cooperating autonomous and heterogeneous vehicles. These systems arereally complex and it becomes fundamental to simulate any mission stage to exploitbenefits of simulations like repeatability, modularity and low cost. In this paper aframeworkforsimulationandtestingofUAVsincooperativescenariosispresented.Theframework,basedonmodularityandstratificationindifferentspecializedlayers,allows an easy switching from simulated to real environments, thus reducing testingand debugging times, especially in a training context. Results obtained using theproposed framework on some test cases are also reported. Keywords  UAVs · Simulation · Cooperative scenarios 1 Introduction During last years, in addition to ground vehicles, mobile robotics is broadening to in-novative branches as  Unmanned Surface/Underwater Vehicles  and  Unmanned Aerial Vehicles  (UAVs). Missions of various kinds are in most cases performed by a fleetof cooperating autonomous and heterogeneous vehicles. Interaction, cooperationand supervision are the core problem of these complex systems. The complexitycorrelated to today challenges in terms of missions and tasks sets up the necessityof simulating, debugging and testing. Simulation activities are fundamental becausedifferent methodological approaches can be easily implemented and evaluated toreduce developing times. This is particularly true in an educational context. A. Mancini ( B ) · A. Cesetti · A. Iualè · E. Frontoni · P. Zingaretti · S. LonghiDipartimento di Ingegneria Informatica, Gestionale e dell’Automazione,Università Politecnica delle Marche, Via Brecce Bianche, 60131 Ancona, Italye-mail: mancini@diiga.univpm.it  308 J Intell Robot Syst (2009) 54:307–329 In the case of ground robots a lot of simulation and test frameworks have beendeveloped.Probably,Player/Stage/Gazebo[1]isactuallythemostcompleteframe-work owing to advanced features like the emulation of 2D–3D environments, sensorsimulation (Laser Range Finder (LRF), sonar,...) and integration with commercialrobotic platforms (i.e., MobileRobots [2], irobot [3]). The framework proposed by Frontoni et al. [4] is particularly suited in an educational context allowing an easy switching from simulated to real environments. Other simulation environments aretaking the attention of the scientific community for the full integration with a lot of commercial platforms, for example Carmen [5], Microsoft Robotics Studio [6] and USARsim (for RoboCup) [7].For the UAV branch of robotics the state of the art is a bit different. First, itis more fragmented because the set of aerial vehicles is more heterogeneous. Inthe case of simulated ground vehicles, common robots are differential wheeled orcar like. On the contrary, among aerial vehicles there are blimps, gliders, kites,planes, helicopters, etc. Each vehicle has a particularity that makes the difference ina mathematical description of physical phenomena. Mathematical models of aerialvehicles are really complex because an aerodynamic description is necessary for arealistic modelling.In this paper a framework for simulation and testing oriented to rotary-wingsaerial vehicles is presented. The framework allows UAV simulation (as stand-aloneagents or exchanging data for cooperation) owing to a Ground Control Station(GCS) that supervises the tasks of each agent involved in the mission. The paperis organized as follows. Next session introduces our framework; a Unified ModellingLanguage(UML)representationisfirstintroducedtosyntheticallydescribeconceptsthat inspired our framework. The use of a UAV CAD modelling for parameterextraction and simulation aids is proposed in Section 3; the modelling activity is contextualized to the Bergen Twin Observer Helicopter. In Section 4, a test case involving take off, landing and navigation is presented; a cooperative scenariothat involves two helicopters in an exploration mission is included. In Section 5conclusions and future works are outlined. 2 Framework In setting up a framework for the simulation of complex and multi-agent scenarioswe identified the following major aspects:– high fidelity mathematical model of vehicles;– multi-agent management;– extended set of simulated sensors;– modularity;– reduced time for updating/adding new modules;– Virtual Reality rendering;– easy switching from simulated to real world and vice versa;– educationally oriented.Till today, game engines and flight simulators are the only available frameworksto simulate UAVs. Game engines (like FlightSimulator [8] or Flight Management System(FMS) [9]) areoptimal for visualization, while flight simulators (like JSBSim,  J Intell Robot Syst (2009) 54:307–329 309 YASim and UUIU [10]) are characterized by a high-fidelity mathematical model,but are lacking in high quality rendering. Most of them are developed for planes.A good, but expensive exception, is the RotorLib developed and commercializedby RTDynamics [11]; in the helicopter context, frameworks with the requirementslistedabovearealmostabsent[12].Theframeworkhereproposedaimsatovertakingthis lack.In Fig. 1 a graphical abstraction with the main modules of the developed frame-work is shown.The stratification of the framework permits to identify five layers: Supervision,Communication, Dynamics, Agent, User Interaction. All the tasks that involve oneormorevehiclesaremanagedandsupervisedbytheGCS,anddataaresenttoagentsusing the communication layer. A socket based interface allows the data exchangebetween GCS and agents in the case of simulated agents, while the communicationmakes use of a dedicated long-range radio modem if a real vehicle (e.g., helicopter)is used [13].Detailed descriptions of more relevant modules of the proposed framework arepresented in the following subsections. First relations among classes are easilyrepresented using an UML diagram and, then, agent structure, simulated dynamics,basic control laws, GCS and virtual reality and world representation are analyzed.Without loss of generality, each aspect is contextualized to a particular class of UAVs, i.e., helicopters.All the modules are implemented in Matlab/Simulink; the main motivation of thischoiceisthereducedcomplexityforcodedevelopment.Inparticular,theend-userof the framework can easily integrate his code for developing and testing an algorithm, Fig. 1  Framework for UAV simulation; a new simulator engine instance is generated for each agentto be simulated  310 J Intell Robot Syst (2009) 54:307–329 e.g., for obstacle avoidance, without the necessity of re-compiling other activities.An additional motivation for the adoption of Matlab is the capability to interface theAeroSim toolbox released by Unmanned Dynamics [14]. The AeroSim Blockset isa Matlab/Simulink block library which provides components for rapid developmentof nonlinear 6-DOF aircraft dynamic models. In addition to aircraft dynamics theblockset also includes environment models such as standard atmosphere, back-ground wind, turbulence and earth models (geoid reference, gravity and magneticfield). These blocks can be added to the basic framework to increase the realism of simulation.2.1 Framework Description by UMLBefore introducing the description of each layer, the proposed framework is pre-sented making use of an UML diagram, according to the Object ManagementGroup’s specification; UML allows to model not only application structures, behav-iors and architectures, but also business processes and data structures [15]. The mostsignificant diagram to model the proposed framework is the  class  diagram, whichis useful to represent hierarchy, networks, multiplicity, relations and more. Due tocomplexity of relations, first, main-classes (Fig. 2) are presented and, then, diagramsof sub-classes and objects (instances of classes) follow.The  Communication  class allows the connection among agents and GCS. Spe-cialized classes, as shown in Fig. 3, implement different communication methods by socket and radio modem.Agents can be also monitored and remotely controlled owing to the  User Inter-action  Class, as shown in Fig. 4.  Active and Passive  stand for the user implication;an active interaction implies the remote control of an agent; a passive one regardscockpit virtualization and virtual reality. Fig. 2  The class diagram. Only parent classes are shown; the GCS can supervise and monitor avariable number of agents. Each agent is then autonomous in terms of control and mission execution  J Intell Robot Syst (2009) 54:307–329 311 Fig. 3  Communication Class.Simulated agents use thesocket paradigm; realcommunication makes use of a long range radio modem The  Vehicle  class is modelled in Fig. 5 taking into account the heterogeneousagents. An interesting analysis of aerial vehicles can be found in [16]. Vehicles havea series of dependencies with other classes as actuators, sensors, dynamics and theFlight Management System (FMS) as shown in Fig. 2.Figures 6 and 7 show the complete class diagram including the parent and derived classes. Sensors are fundamental accomplishing successfully and secure autonomousmissions; environment must be sensed to execute autonomous tasks like take-off,navigation and landing. Specialized classes are necessary to interface sensors inproviding data. If a limited set of sensors is available in the simulated scenario, a newclass( Generic )forthemanagementofnewsensorscanbeimplemented.Sensorsandactuators form the  Avionic System . In general, each agent can have a set of   i  sensorsand  j  actuatorswith i possiblydifferentfrom  j  ;forexampleinthecaseofahelicopter,five classes are derived from the  Sensor   class and three classes (or four if the  ServoMotor   class is splitted into  Analog  and  Digital  ) from the  Actuator   class (see Figs. 6and 7).The diagram of the  Dynamics  class is proposed in Fig. 8. This class makes senseonly in a simulation scenario. Considering the membership class of an agent, themathematical model used to describe the dynamics varies significantly. In the case of a helicopter, models as  Disk Actuator   and  Blade Element   are widely adopted [17]. The FMS, formed by a set a classes, is described in more details in the nextsub-section.2.2 Agent StructureIn a simulated or real case, the structure of an agent in the context of UAVs isbased on a complex interaction of different specialized modules. In the real case, theFMS is implemented as real-time code running on high performance architecturesas PC104+, FPGA, DSP; in the simulation environment, FMS is a complex set of  Fig. 4  User Interaction Class.User(s) can interact withagents owing to activeinterfaces as joystick(simulated scenario) ortransmitter (real case)
Search
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks