EXPERIMENTAL RESULTS ON COMMAND AND CONTROL OF UNMANNED AIR VEHICLE SYSTEMS

Transcription

1 EXPERIMENTAL RESULTS ON COMMAND AND CONTROL OF UNMANNED AIR VEHICLE SYSTEMS Pedro Almeida (1), Ricardo Bencatel (1), Gil M. Gonçalves (1), João Borges Sousa (1), Christoph Ruetz (2) (1) Faculdade de Engenharia da Universidade do Porto {pinto.almeida, ricardo.bencatel, gil, (2) Universitat Karlsruhe, Abstract: Experimental results from operational deployments of Unmanned Air Vehicle (UAVs) systems are presented and discussed. This is done in the context of the Apollo command and control framework developed by the AsasF project from Porto University. Apollo is targeted at multi-vehicle operations in mixed initiative interactions (allowing intervention of human operators). Field tests have been carried out using a modified radio-controlled plane. Successful Apollo integration in a real UAV operational environment has also been demonstrated. This closes the integration cycle between the software architecture and the hardware platform. Results concerning aircraft performance, future capabilities and experimental steps of the UAV system under developed are also described. Copyright 27 IFAC Keywords: UAV control, UAV operations, Hybrid Automata, Middleware, Networked vehicle system, Command and Control and Systems Engineering. 1. INTRODUCTION The design and deployment of multi-vehicle systems with mixed initiative interactions in a systematic manner and within an appropriate scientific framework requires an interdisciplinary approach from computation, control and communication sciences. The major challenges come from the distributed nature of these systems and from the human factors. This is why we need to couple the development of command and control frameworks with field tests with human operators. Here we present the results from the first operational deployment of the Apollo command and control system (Almeida et al., 26a). Apollo was developed in the context of the AsasF project from Porto University. AsasF is an interdisciplinary research and development project at the Faculty of Engineering from the University of Porto in Portugal. The main objectives of the project are: 1) to design and build light unmanned air vehicles to test concepts for coordinated missions; 2) to develop concepts of operation for multi-vehicle systems; 3) to design and build modular controllers for unmanned air vehicles; and 4) to evaluate and test the developments in operational deployments. This is done in the framework of Systems Engineering, following the approach presented in the IEEE standard (IEEE, 25). The AsasF project builds on expertise, tools and technologies developed at the Underwater Systems and Technology Laboratory (USTL) from Porto University. Researchers at USTL have been designing and building ocean and air going autonomous and remotely operated vehicles with the goal of deploying networked vehicle systems for oceanographic and environmental applications. The concepts behind the AsasF control architecture are built on experience in the modular design of distributed control hierarchies described in (Sousa et al., 23). AsasF uses the Neptus command and control framework (Dias et al., 25; Dias et al., 26) and the Seaware middleware publish/subscribe framework for distributed real-time systems (Marques et al., 26). Apollo is modelled in the framework of dynamic networks of hybrid automata. This facilitates the mapping of the command and control concepts onto the software architecture that implements it. The software implementation and the system integration are done with the help of Seaware and Neptus frameworks. The systems integration allows for multi-vehicle coordination and control as well as to test and evaluate new control concepts. Recent advances in UAV research include one system where in-the-air effective task allocation and collaboration is demonstrated (Ryan et al., 26). The capabilities displayed by this system include single-user control of a fleet of aircraft, distributed task assignment, and vision-based navigation. Research in (Lee et al., 23) shows a strategy of path-planning for an UAV to follow a ground vehicle. This ground vehicle may change its heading and vary its speed. Here, the UAV will maintain a fixed airspeed and will manoeuvre itself to

2 track the ground vehicle. (Girard et al., 24a) uses selected case studies as the motivation to examine emerging results in networked multi-vehicle systems. A control architecture for a UAV system that does border or perimeter patrol is proposed in (Girard et al., 24c). This paper is organized as follows. Section 2 describes the AsasF UAV system. Section 3 presents our command and control system, with special emphasis on the Apollo system. Section 4 discusses experimental results from the first Apollo operational deployment. We discuss the conclusions and future work in section ASASF UAV SYSTEM The System Breakdown Structure (SBS) for the AsasF Unmanned Air Vehicle System (UAVS) is depicted in Fig. 1. onboard autopilot (Piccolo). A single user can control a number of Piccolo-equipped UAVs using the ground station and accompanying Operator Interface (OI) software. The OI displays all onboard information from each UAV and sends high-level commands to the Piccolos over a 2.4 GHz radio link Piccolo includes all necessary sensors and communication links to perform an accurate control of the vehicle. It is really as small as it looks (Fig.3). The Piccolo autopilot includes: 3 gyroscopes, 2 two axis accelerometers. Static and total pressure sensors. Temperature sensor (used with static and total pressure sensors to measure altitude and airspeed). GPS. Radio. Control of up to 1 servos. Figure 3: Piccolo avionics Figure 1: SBS of the Asasf UAS Airframes Currently, we are using three UAV platforms: Lusitânia, Brutus v1 and Brutus v2. Lusitânia (Fig. 2) is based on a commercial remotely controlled model airframe. Brutus (fig 4b) is based on a carbon fibre airframe developed under the NAAM project by INEGI 1 and FEUP for the Air Cargo Challenge 2. Brutus and Lusitânia share the same computational and sensor configurations. Engine The same engine is being used in the three airframes: a 15cc, 2.9HP, 2 stroke engine, OS 91-FX. This will allow us to have a takeoff payload weight larger than 5kg. The maximum flight time for Lusitânia is 2 minutes with a payload of 5kg (includes a wireless video camera). Figure 2: a) Lusitânia b) Brutus v1 Avionics The UAVs are equipped with the Piccolo autopilot. Piccolo is an avionics system provided by CloudCap Technology (Vanglienti et al., 24) and consists of a central controller (ground station) and an 1 Instituto de Engenharia Mecânica e Gestão Industrial Ground Station The ground station (Fig. 4) comes with the Piccolo Autopilot, and handles communication between the operator interface and the avionics. Figure 4: Ground Station Payload The UAV carries a very small camera (36mm x 36mm x 33mm) (Fig 5) (video system) and a Humidity, Temperature, and Light Node based on a Telos mote 3. This node is a measurement source of meteorological data optimized for use on a UAV platform. The camera can be remotely controlled, and provides the operator with a video feed in real-time. This is done through a reliable wireless transmission system in the 2.4GHz frequency, with a range of 8Km Figure 5: Video Camera 3

3 3. COMMAND AND CONTROL FRAMEWORK 3.1 EXECUTION CONCEPTS We use the concept of manoeuvre a prototype of an action/motion description for a vehicle as the atomic component of all execution concepts. We abstract each vehicle as a provider of manoeuvres and services. A simple protocol based on an abstract vehicle interface governs the interactions between the vehicle and an external controller: the external controller sends a manoeuvre command to the vehicle; the vehicle either accepts the command and executes the manoeuvre, or does not accept the command and sends an error message to the controller; the vehicle sends a done message or an error message to the controller depending on whether the manoeuvre terminates successfully or fails. This protocol facilitates inter-operability with other platforms. We have adopted an object-oriented approach in the design of manoeuvre controllers. Manoeuvre controllers are implemented through the specialization of a toplevel class (manoeuvre) containing most of the member-functions and member-objects that manoeuvres will need. Each manoeuvre controller inherits the structure of this manoeuvre class. The toplevel class Manoeuvre controller is depicted in Figure 7. It is a 4 state hybrid automaton. A detailed description of the Manoeuvre Controller automaton can be found in (Almeida et al., 26b). In the current implementation we have three specializations of manoeuvre controllers: tele-operation, goto and Loiter. 3.2 APOLLO Apollo has a layered structure depicted in Fig.6. The Manoeuvre controller supervises the execution of a vehicle manoeuvre. It has feedback loops to control the actuators. The Vehicle supervisor supervises all of the UAV operations. It receives manoeuvre specifications by the Mission Supervisor during a mission execution. It then launches the corresponding manoeuvre controller, monitors its execution and accepts configuration commands. If there is no mission to be executed, the Vehicle supervisor commands the execution of a default manoeuvre (in our case, loiter manoeuvre). The Mission Supervisor supervises the execution of a mission and puts forward a suitable error tackling strategy. Apollo is modelled in the framework of dynamic networks of hybrid automata DNHA (Girard et al., 24b). Hybrid automata model systems exhibiting hybrid behaviour consisting of continuous-time phases separated by discrete-event transitions (Henzinger, 1996). DNHA are used to model dynamic interactions. Informally, DNHA allow for interacting automata to create and destroy links among themselves and for the creation and destruction of automata. Figure 6: Piccolo Apollo Seaware Neptus Shared Data Figure 7: Manoeuvre controller top-level class We have coded an error masking capability in the manoeuvre controllers. In practice, we want to be able to enable/disable transitions to the Error state. This is done by the conjunction of the error condition with a mask variable in each transition. The operator is allowed to set and reset this variable. 3.3 TOOL SET The AsasF C4I framework is depicted in Fig. 6. It consists of Apollo (Almeida et al., 26b), the UAV autopilot and external controllers. The Neptus/Seaware tool set, developed at USTL, is being used to support the implementation of the AsasF C4I framework. Communications between the UAV and the C4I will implement the NATO standard STANAG 4586 (NSA 26). These tools are described next. Seaware is a middleware framework that addresses the problem of communications in heterogeneous environments with diverse requirements (Marques et al., 26). Seaware adopts publish/subscribe based messaging, defined by anonymous message exchange between data subscriptions and publications (Fig. 8). Each application dynamically registers itself, specifying the topics it wishes to publish and subscribe, without the need to know in advance who its peers are or where they are located. Neptus is a distributed command, control, communications and intelligence framework for operations with networked vehicles, systems, and human operators (Dias et al., 25). The interactions

4 with human operators are classified according to the phases of a mission life cycle: world representation; planning; simulation; execution and post-mission analysis. This networked system is represented in Figure 1, where all communications between vehicles and their external supervisors/controllers are done through Seaware. Figure 8: Interactions under Seaware Neptus supports concurrent operations. Vehicles, operators, and operator consoles come and go. Operators are able to plan and supervise missions concurrently. Additional consoles can be built and installed on the fly to display mission related data over a network. Figure 9 depicts a UAV operator console built with the Neptus Console Builder (CB) application. This facilitates the addition of new vehicles with new sensor suites to Neptus. Figure 9: Neptus UAV operator console Neptus supports the control of several UAVs, AUVs and ASV concurrently. There is a Seaware node per vehicle and per operator console (one per vehicle). Each vehicle node is characterized by a topic domain identifying the vehicle to allow for a set of messages to be exchanged with the corresponding operator console. Apollo is easily configured for deployment in heterogeneous UAVs to carry out coordinated missions. This is because of the modular design and of the use of the publish/subscribe framework for communication. In future releases, the user will be able to easily add manoeuvres to the Apollo system; this includes manoeuvres for coordinated missions. The abstract vehicle interface protocol facilitates interactions and inter-operability with the Seaware publish/subscribe framework. Seaware defines the message topics and syntax for coordinating vehicles in a network. This allows us to build a network system composed of different types of entities such as control consoles, heterogeneous (AUVs, ASVs, UAVs) physical or simulated vehicles and fixed systems, which may come and go in a transparent way to the network. Figure 1: Network vehicle system 4. AUTONOMOUS FLIGHT FIELD TESTS Autonomous flight field tests with the Lusitânia UAV took place at OTA Air Base in December 26 and January 27. This was done in an incremental fashion. First we flew Lusitânia in a Remotely Controlled configuration. Afterwards we mounted the Piccolo autopilot onboard and remotely piloted Lusitânia in this configuration for preliminary tests. It was only then that we calibrated the Piccolo control loops, while switching between manual and autonomous flight modes. Finally, Lusitânia was ready for the autonomous execution of a mission plan. Once in autonomous mode, the UAV was controlled by a single user through the Piccolo Operator Interface. The hierarchical control architecture for UAS, Apollo, was successfully field-tested. A mission was hard coded at the Mission Supervisor level. The execution in the framework of the architecture described in section 3.2 was proven to be correct. Neptus was used for vehicle monitoring only (Figure 9). In what follows we present an analysis of the telemetry from the first autonomous flights. Figure 11, represents the commanded task (red line shows the trajectory formed by the waypoints) and the vehicle trajectory (blue line). Airspeed command was held constant during the whole task, whilst the altitude changed for every waypoint. North [m] East [m] Figure 11: Unleveled box path It is noticeable that the UAV overshoots at each waypoint, with the exception of the top left one. This is

5 mainly because the Piccolo preturn option (UAV starts turning to next waypoint before reaching commanded one) was turned on only at that waypoint. We remark the correct operation of the vehicle even in the presence of wind coming from Northeast (see Figure 14). Sharper turns against the wind contrast with the softer ones when flying in the other direction (right to left). Turn rate [º/s] mode). This phugoid mode (exchange of airspeed for altitude) is always present although its effects are dissimulated by the pitch damper action. This indicates a need to increase the integral term to minimize the steady-state error. Wind estimation is provided by the Piccolo autopilot and is based on the aircraft s airspeed and groundspeed. The estimated wind conditions show an average wind speed of 4m/s and a direction between 4º and 6º, which means that the wind is coming from Northeast. The error in estimation was.24m/s, the minimum acquisition time was 6s, while the maximum was 45s ,5 4 4,5 5 5,5 6 6,5 7-4 Time [min] r [º/s] Turn_cmd [º/s] Figure 12: Turn rate performance Figure 12 depicts an overall well behaved turn rate controller, given that the vehicle is able to track the commands; the step response of the system is quick enough (less than 1s). Nevertheless, there is a slight problem with the steady-state response. It presents high oscillations. This suggests that performance can be improved by decreasing the proportional turn rate gain in the Piccolo autopilot Wind speed [m/s] ,,5 1, 1,5 2, 2,5 3, 3,5 4, 4,5 5, 5,5 6, 6,5 Time [min] WindS [m/s] WindD [º] Figure 14: Wind estimation Figure 15 shows a sample image captured with the on board camera, where the red circle indicates the location of the ground station Wind Direction [º] Airspeed [m/s] Altitude [m] 9, 1, 2, 3, 4, 5, 6, Time [min] TAS_cmd [m/s] TAS [m/s] Alt_cmd [m] Alt [m] Figure 13: Altitude and airspeed performance Altitude shows a good convergence to the commanded value in response to step inputs (Figure 13), despite being expectedly slow (approximately 2sec to reach commanded value). Moreover, altitude does not overshoot, does not have relevant oscillations (+/- 4m) and the average value is close to the commanded one. Therefore, this shows no need to change altitude gains. The airspeed plot shows both small steady-state oscillations (+/-1m/s) and an average airspeed (13.32m/s) above the commanded value (13m/s), with an error of 2.4%. Nevertheless, since the airspeed and altitude controllers are related, oscillations are justifiable by the fact that the altitude was constantly changing, thus creating the observed perturbations in airspeed (slowly damped phugoid 4 Figure 15: Image taken from the onboard video 5. CONCLUSIONS This paper describes the field tests of the Apollo C4I system as well as results from the first autonomous flight field tests. It also describes future capabilities of the AsasF UAVS. In the context of the cooperation program with the Portuguese Air Force Academy (AFA), a demonstration of the multi UAV system is foreseen for March 27. In this demonstration, the main goal is to have a Hardware-in-the-loop (HWIL) simulation running with 5 different UAVs performing basic coordination tasks (Figure 16). In this demonstration there will be one human operator per vehicle. Task coordination will be achieved with the help of a human supervisor that will

6 have monitoring access to all UAVs and will be responsible for assigning aircrafts to operators as well as tasks to UAVs. With this goal in mind, a tactical communications channel will be implemented. The demonstration scenario will include 3 Piccolo equipped UAVs and 2 Micropilot equipped UAVs. Each autopilot s groundstation can communicate with more than one aircraft and the computers are connected in a local network. In order to do so, an implementation of a Vehicle Specific Module (VSM) for the Micropilot autopilot is needed (Micropilot 25). This module will accomplish the same tasks as the Piccolo module in a transparent way for Apollo. Moreover, in order to be interoperable with existing UAV systems Apollo will be STANAG 4586 (NAS 26) compliant. Consequently, our system will have increased flexibility and efficiency to meet mission objectives through the sharing of assets and common utilisation of information generated from UAV systems. Figure 16: Scenario for HWIL demonstration The AsasF C4I system is a work in progress. The results obtained from simulation show that there are no major differences in performance between Software-in-the-Loop and Hardware-in-the-loop simulations. New releases will incorporate lessons learned from operational deployments. The available functionality is being extended and improved. This includes: 1) manoeuvres for optimal trajectory tracking, for pursuit-evasion games and for reaching waypoints under disturbances; and 2) on-the-fly definition of new manoeuvres by the user/operator. ACKNOWLEDGEMENTS This research was partly funded by the School of Engineering at Porto University under the project AsasF and under the PESC program (Projectar, Empreender, Saber Concretizar) We thank LSTS and NAAM for their technical support. We gratefully acknowledge the support of the Portuguese Air Force Academy in our field deployments; special thanks go to Tenente Elói Pereira and to Tenente-Coronel José Morgado for all the logistics support, cooperation and contributions as well as to Capt. José Costa for the field support, advice and flying experience. Finally, we thank Professors António Torres Marques and Professor Fernando Lobo Pereira for their contribution to the Asasf project. REFERENCES Almeida, Pedro; Bencatel, Ricardo; Gonçalves, Gil M.; Sousa, João B.; Multi-UAV Integration for Coordinated Missions, Encontro Científico de Robótica, Guimarães, April 26 (a) Almeida, Pedro; Gonçalves, Gil M.; Sousa, João B.; Multi-UAV Platform for Integration in Mixed-Initiative Coordinated Missions, First IFAC Workshop on Multi-vehicle Systems (MVS'6), Salvador, October 26 (b) Dias, P.S.; Fraga, S.L.; Gomes, R.M.F.; Gonçalves, G.M.; Pereira, F.L.; Pinto, J.; Sousa, J.B.; Neptus - a framework to support multiple vehicle operation Oceans 25 Europe, Volume 2, 2-23 June 25 Page(s): Vol. 2 Dias, P. S., R. Gomes, J. Pinto, G. M. Gonçalves, J. B. Sousa and F. L. Pereira (26); Mission planning and specification in the Neptus framework. Humanitarian Robotics, ICRA 26 IEEE International Conference on Robotics and Automation. Girard; Anouck R.; Sousa, João; Hedrick, J. Karl; A Selection of Recent Advances in Networked Multi-Vehicle Systems; Proceedings of the Institution of Mechanical Engineers, Part I, 24 (a) Girard; Anouck R.; Sousa, João; Silva, Jorge E.; Elemental Manoeuvres for Unmanned Air Vehicles, Proceedings of Robotica 24, April 24-25, 24, pp. 63-7, Portugal (b) Girard, Anouck R.; Howell, Adam S.; Hedrick, J. Karl; Border Patrol and Surveillance Missions using Multiple Unmanned Air Vehicles, Proceedings of the IEEE Conference on Decision and Control, Paradise Island, Bahamas, 24 (c) Henzinger, Thomas A.; The theory of Hybrid Automata, Proceedings of the 11th Annual Symposium on Logic in Computer Science (LICS), IEEE Computer Society Press, 1996, pp IEEE Standard for Application and Management of the Systems Engineering Process, IEEE Std , September 25. Lee, Jusuk; Huang, Rosemary; Vaughn, Andrew; Xiao, Xiao; Hedrick, J. Karl; Zennaro, Marco; Sengupta, Raja; Strategies of Path-Planning for a UAV to Track a Ground Vehicle, In Proceedings of the 2 nd annual Autonomous Intelligent Networks and Systems Conference, Menlo Park, CA, June 23. AINS Marques, Eduardo; Gonçalves, Gil; Sousa, João; Seaware: A Publish/Subscribe Middleware for Networked Vehicle Systems, 7th Conference on Manoeuvreing and Control of Marine Craft (MCMC 26), Lisboa, Portugal, 26. Micropilot User s Manual, 25 NSA Standardization Agreement (STANAG), Standard Interfaces of UAV Control Systems (UCS) for NATO UAV Interoperability, 26. Ryan, Allison; Xiao, Xiao; Rathinam, Sivakumar; Tisdale, John; Zennaro, Marco; Caveney, Derek; Sengupta, Raja; Hedrick, J. Karl; Collaborative flight demonstration of three unmanned aerial vehicles, Submitted to the AIAA Conference on Guidance, Navigation, and Control, Keystone, CO, 26 Sousa, J. B., F. Lobo Pereira, P. F. Souto, L. Madureira and E. P. Silva (23). Distributed sensor and vehicle networked systems for environmental applications. Biologi Italiani, n. 8, pp Sousa, J. B., T. Simsek and P. Varaiya (24). Task planning and execution for UAV teams. Proceedings of the Decision and Control Conference, Bahamas. Vaglienti, B.; Hoag, R.; Niculescu M., Piccolo system user guide, July 24,