Delhi College of Engineering 2009 AUVSI STUDENT UAS COMPETITION. Team UAS DCE Journal Paper

Size: px

Start display at page:

Download "Delhi College of Engineering 2009 AUVSI STUDENT UAS COMPETITION. Team UAS DCE Journal Paper"

Madison Shepherd
5 years ago
Views:

Delhi College of Engineering 2009 AUVSI STUDENT UAS COMPETITION Team UAS DCE Journal Paper ABSTRACT The following paper discusses the design and implementation of an Unmanned Aircraft System (UAS)

While in flight, the airplane needs to find and identify targets, as fast as possible. This is the first participation of the college at this prestigious competition.

1 Delhi College of Engineering 2009 AUVSI STUDENT UAS COMPETITION Team UAS DCE Journal Paper ABSTRACT The following paper discusses the design and implementation of an Unmanned Aircraft System (UAS) for the Association for Unmanned Vehicle Systems International (AUVSI) UAS 2009 competition. At this event, the UAS has to perform an autonomous reconnaissance on a predefined path. While in flight, the airplane needs to find and identify targets, as fast as possible. This is the first participation of the college at this prestigious competition. Our aim is to develop a Fixed wing Unmanned Aircraft System consisting of An Indigenously developed Low Cost, 6 Degrees of Freedom Autopilot Real Time Ground based Target Identification using Machine Vision The vehicle used as the platform of this year s competition is the Sig Rascal 110 ARF. It is powered by a 2 stroke glow fuel engine. It is equipped with a CCD camera, a single board computer (SBC), wireless routers and an Attitude Heading and Reference Positioning System.

2 CONTENTS 1. Introduction 3 2. Mission Requirements 3 3. System Design 4 4. Aerial Vehicle 5 5. Autopilot Development 7 6. Software Architecture Ground Station Machine Vision Power System Safety Features Conclusion Acknowledgements References 20 2 P age

3 1. Introduction The importance of Unmanned Aircraft Systems is evident in their vast applications in defence, agriculture, traffic management, weather monitoring, disaster management etc. The team intends to develop a UAS keeping in mind the various basic requirements image acquisition, stable autonomous flight and waypoint navigation. Another key goal of this team is to develop the UAS with least dependence on Commercial off the shelf (COTS) products, primarily the autopilot. The development of a UAS is a very challenging task and it requires experience and knowledge of several fields. The team is comprised of 10 undergraduate students pursuing various disciplines of engineering. This paper describes DCE s entry for the 2009 competition and how its team of undergraduate students approached this year s challenge. 2. Mission Requirements The mission requirements, as specified in the competition rules are summarized in the table below: Table 1: Key Performance Parameters Parameter Threshold Objective Autonomy Imagery Target Location Mission time During waypoint navigation and area search Identify any two target characteristics: Shape Background color Orientation Alphanumeric Alphanumeric color Determine target location within 250 ft Less than 40 minutes total Imagery/location/identification provided at mission conclusion All phases of flight, including takeoff and landing Identify all five target characteristics Determine target location within 50 ft 20 minutes Imagery/location/identification provided in real time In-flight updates Add waypoints Adjust search area 3 P age

4 3. System Design 3.1 Design Approach In order to successfully complete the project, it was decided that a systems engineering approach to the problem would be highly beneficial. The key areas of work were identified in the initial planning stages on the basis of a thorough literature review. Aerial Vehicle RC Flight Training Assembly of UAS Mounting of camera and electronics AHRPS Data Acquisition Sensor Fusion Autopilot Design Navigation Development of Algorithm Simulation and Calibration Servo Control Control System Design of 6 DOF model Design of Control Loops Software Architecture Development of various modules Operating System modifications System Integration Machine Vision Development of unsupervised clustering algorithm Design of a neural network/identification technique Development of interface Power System Design of Power Distribution Board Ground Station Development of GUI Wireless Communications Fig 1: Task Division 4 P age

5 3.2 System Overview The aircraft will be flown manually for take off and then the control will be transferred to the autopilot. The maximum flight time estimated is 40 minutes with a cruise speed of 45 ft/s.the self developed autopilot system guides the aircraft through the GPS waypoints and then enters the search area. The communication between the aircraft and ground station is achieved via a wireless serial link. The gimballed camera is used for image acquisition. The gimbal allows the camera to move along 2 axes and increases the field of view. The gimbal is self stabilizing and manually operated from the ground by the operator if he spots a potential target. The image is sent to the ground station via a separate 2.4 GHz 1000mW transmitter and receiver set. The images are then saved and further processed. Inertial Measurement Unit Rs 232 Link RC Receiver GPS Rs 232 Link Linux SBC Rs 232 Link ATMEGA 128 Microcontroller Manual Over ride switch 2.4 GHz wireless Link GROUND STATION Pressure sensors 12 bit ADC Servo Glitch Remover Servos Fig 2 : UAS System Architecture 4. Aerial Vehicle 4.1 Airframe Selection While selecting the airframe of the vehicle, the following factors were kept in mind: Basic aircraft type: fixed wing rotor. Since the competition does not require the UAS to have hovering capabilities, a fixed wing platform was chosen as its control system can be designed more effectively. 5 P age

Type of fixed wing: The chosen vehicle was to have a low cruise and stall speed for a more effective area search during image acquisition. Thus, a high winger was selected.

Material of the airframe: Though a carbon fiber composite airframe would have been structurally stronger and lighter, it wouldn t have allowed any further modifications to be made onto it.

6 Type of fixed wing: The chosen vehicle was to have a low cruise and stall speed for a more effective area search during image acquisition. Thus, a high winger was selected. A highwinger also demonstrates more stability than other wing arrangements. This too aided in the development of the control system. Material of the airframe: Though a carbon fiber composite airframe would have been structurally stronger and lighter, it wouldn t have allowed any further modifications to be made onto it. Also, carbon fiber limits the wireless transmission through it. Therefore, an airframe made of balsa and ply was chosen. Payload: The aircraft was then selected on the basis of the lift it generated so as to easily carry the required payload. Based on all the above parameters, the team decided to use Sig Rascal 110 ARF, manufactured by Sig Mfg[3]. The Sig rascal is a proven, very stable RC aircraft. 4.2 Engine selection Fig 3 : Pre Flight Preparations Based on the power requirements for the payload and the flight time of the aircraft, a 2 stroke glow fuel engine was preferred over the other alternatives. With the payload then finalized, the value of thrust required was calculated. The propeller selection was done using thrust calculations in JavaProp [1]. The compatible engine was then selected based on the power output. The engine that is being used is the OS 1.60 FX 2 stroke glow fuel engine [2] along with a 16x8 propeller, delivering ample thrust for the UAS. To achieve a flight time of 40 minutes, a 36 oz fuel tank is being used. The fuel usage estimation was done on the basis of the earlier flight tests data where the fuel usage was monitored carefully with the position of the throttle. 6 P age

7 5. Autopilot Development 5.1 Six Degree of Freedom Model A 6 DOF model was developed for the UAS using Matlab/Simulink. The model describes the rate of change of 9 state variables defining the state of the aircraft at any instant. For the model, various Stability and Control coefficients were needed which have been determined using USAF Digital Datcom Software and Design Foil [4]. Then using the coefficients determined, 3 fundamental forces and moments acting on the aircraft are evaluated in terms of the state variables and control inputs. Finally, 9 differential equations are formed which are in terms of the state variables and control inputs. To simulate the aircraft, an initial state is given as input to the mathematical model and control surfaces deflections are described for the period of simulation. The simulation then gives the aircraft s behaviour for the period of simulation. To evaluate the model for its accuracy, we performed a phugoid test [4] on the model. This was performed in Simulink where elevators were given a particular input command and the response of the aircraft was studied. Fig 4 : 6 DOF Mathematical Model of the UAS (Open loop test) 7 P age

Pitch Response Pitch rate Response Height Response Velocity Response Fig 5 : Phugoid test on the UAS 5.

Position system is used to determine the following flight parameters: Roll, Pitch, Heading, Altitude, Air

The above parameters are determined using inertial, magnetic and pressure sensors, as well as GPS.

8 Pitch Response Pitch rate Response Height Response Velocity Response Fig 5 : Phugoid test on the UAS 5.2 Attitude Heading and Reference Position System Development (AHRPS) The Attitude Heading and Reference Position system is used to determine the following flight parameters: Roll, Pitch, Heading, Altitude, Air speed, Ground Speed, Ground Heading, Latitude, and Longitude. The above parameters are determined using inertial, magnetic and pressure sensors, as well as GPS. We used Crossbow s MNAV Inertial Measurement Unit to determine the aircraft s acceleration in the 3 directions as well as the gyro readings about 3 axes. MNAV also gives us GPS data such 8 P age

as Latitude, Longitude, Ground Speed and Altitude. However, its GPS receiver is prone to interference from its own IMU as well as other avionics in the aircraft.

The choice was made as this GPS has proven performances in systems where GPS signals are prone to interference.

9 as Latitude, Longitude, Ground Speed and Altitude. However, its GPS receiver is prone to interference from its own IMU as well as other avionics in the aircraft. Therefore, it was decided to use Navman s Jupiter 130 GPS [5] module for computing the above parameters. The choice was made as this GPS has proven performances in systems where GPS signals are prone to interference. To determine airspeed and barometric altitude, the team used separate Dynamic and static pressure sensors with self designed amplifier circuits to measure airspeed and altitude. MNAV s X Bow IMU Navman Jupiter 130 GPS Fig 6 : AHRPS Components For determination of the attitude of the aircraft from the raw accelerometer and gyro readings a 6 state Extended Kalman Filter (EKF) was implemented. The EKF was implemented in the following fashion: All the sensors were first calibrated by the procedure suggested by Crossbow. [6] The accelerometers were calibrated against earth s gravity The magnetometer was calibrated against earth s magnetic field using the values of horizontal and vertical component of earth s magnetic field. [6] [7] The algorithm for the EKF was implemented at a frequency of 50Hz on the SBC. Testing of the EKF Ground test: Luckily, DCE s Autonomous Underwater Vehicle (AUV) team recently purchased a commercial AHRS, the Xsens MTi [16]. The AUV team generously allowed us to use their AHRS for testing purposes. This enabled rapid development of the EKF and dramatically reduced the design time, while at the same time providing a reliable source for comparison. The results of final tests are below: 9 P age

10 Fig 7 : Results of Ground Testing of AHRS The results clearly demonstrate that EKF implemented is as good as a commercially available AHRS. 5.3 Control System The control system of the UAS is PID based. For implementation of a PID based system a linear model of the aircraft was required. The PID loops are then implemented about certain stable points in the system. For this purpose, the 9 non linear simultaneous equations of the model were solved in Matlab. The system was then decoupled into Lateral and Longitudinal systems. [15] The Lateral System has the following control loops Yaw Damper Bank to Aileron The Longitudinal System had the following control loops 10 P age

Pitch to elevator Altitude to elevator Throttle to Airspeed The gains were adjusted during simulations of the UAS in Simulink and finally during the flight tests of the UAS. 5.

11 Pitch to elevator Altitude to elevator Throttle to Airspeed The gains were adjusted during simulations of the UAS in Simulink and finally during the flight tests of the UAS. 5.4 Servo Control For autonomously addressing the control surfaces (servo motors), Atmega 128 microcontroller by Atmel [8] is used. The microcontroller receives angles desired for each servo by the control system from the SBC over a serial (RS232) link. The microcontroller then generates required PWM signals to command each servo to the requested angle. 5.5 Navigation Fig 8 : Atmega 128 Microcontroller Board The navigation algorithm is based on the vector field approach [9], which has been successfully used in Guidance Control of miniature UAV s. The main advantages of using this approach are: 1. Most of the work done in the field of UAS navigation has been done by treating it as a trajectory tracking problem. While this approach has been successful in many instances, it suffers from serious drawbacks. The primary difficulty comes due to the discrete time nature of the system, due to which errors in measurement affect performance significantly. Moreover, this approach also does not take into account the effects of wind, which becomes a huge factor at high altitudes. 2. The vector field approach eliminates both of the above problems. It is simple and intuitive and the algorithm was easily developed in a few days. For testing and simulation of the algorithm, three stages were planned. 1. First, the algorithm was tested on a simplified 2 D mathematical model of the UAS. 2. Second, the algorithm was tested along with the actual autopilot model of the UAS. 3. Finally, verification and fine tuning of the algorithm was done in actual flight tests. 11 P age

12 6. Software Architecture The onboard processing of the UAS shall be carried out on the TS 7800 [10 ], a Linux OS based Single Board Computer (SBC). The reasons for selecting Linux as a development environment were due to the design constraints imposed by the nature of the UAS. A number of tasks such as the data acquisition, AHRS, control system, navigation, communications etc. Are to be carried out concurrently and in real time. Reliable mechanisms for inter process communications are needed. The C programming language is the most suitable for our application due to its extremely fast and efficient nature. Keeping these factors in mind, Linux turned out to be the most optimum choice. 7. Ground Station Fig 9 : TS 7800 SBC All communication between the ground station and aircraft except for video is handled by the ground station software. JAVA was chosen as the programming language for the UAS Ground Station design. 7.1 Development of Front End The interface was developed which indicates the following: Text boxes for various parameters such as yaw, pitch, roll. Virtual Horizon developed using circles, lines and other basic JAVA tools. A map showing the current UAS position, waypoints and no fly zones. Text boxes for setting different flight parameters like new waypoints. The front end has to ensure that all the values are within their geometrical and structural limitations, and errors are indicated by highlighting the respective field. 12 P age

13 Fig 10 : Ground Station Interface 7.2 Development of Back End The back end code is the back bone of the Ground Station. It does the following: Receives data serially from RS 232 link with the help of RX TX Communications. Directs the incoming serial packet to various text boxes like yaw, pitch and roll. Depending on the incoming serial packet, current UAS position is plotted continually on the map. The incoming serial packet is used to plot the Virtual Cockpit. A log of the flight telemetry is created and is updated continually in real time. The back end code was integrated with the front end code. 7.3 Communication Link A wireless serial (RS 232) link was established between the ground station and the SBC. For this purpose, Xstream s 2.4 Ghz OEM RF module was used. 13 P age

14 Fig 11 : Wireless Module 8. Machine Vision The imaging system is one of the critical elements needed to achieve a successful mission. It had to fulfil essential properties like reliability, energy and cost efficiency. And while achieving all these requirements, it must provide the best resolution possible. The weight and placement constraints fixed by our design also governed the choice in deciding the camera and gimbal system. A thorough research produced quite a number of options ranging from surveillance network cameras to digital cameras but finally, Range Video s KX191 camera along with Pandora pan/tilt kit was thought of as the most optimum answer to the above mentioned problems. Specifications of KX 191: 520 TVL horizontal resolution 1/3 Sony Super HAD CCD Back light compensation Automatically switches to black and white night mode. f4mm IR sensitive lens Auto White Balance 35x35mm, 46 g with lens 12V±10% DC, 210mA draw 14 P age

Fig 12 : Range Video KX191 camera with pan/tilt kit The analog video output of the camera was connected to Black Widow AV 2.4 Ghz 1000mW tx/rx set.

The S video RCA to USB converter was chosen to convert the transmitted video in composite format to digital format so that it can be processed by the imaging ground station Fig 13 :

Though some compromise was made on the speed but the ease and reliability were some good advantages.

15 Fig 12 : Range Video KX191 camera with pan/tilt kit The analog video output of the camera was connected to Black Widow AV 2.4 Ghz 1000mW tx/rx set. To have an uninterrupted video signal, a circularly polarised patch antenna was used on the receiver side. The S video RCA to USB converter was chosen to convert the transmitted video in composite format to digital format so that it can be processed by the imaging ground station Fig 13 : BlackWidow AV TX/RX and RCA to USB adapter The Imaging ground station works completely in Mathworks MATLAB R2008a. Though some compromise was made on the speed but the ease and reliability were some good advantages. Moreover the compatibility of the frame grabber was also a key factor in making this choice. The interface displays the current video along with options to capture the target frames. Then the frames are processed to identify the shape, colour and alphanumeric. Segmentation on the basis of colour and topology was used to extract the different objects present in the image [13][14]. The respective objects were then identified into a character or a 15 P age

shape using a neural network [11][12]. Once the target is identified, the GPS coordinates are obtained using the data logs available on the main ground station.

16 shape using a neural network [11][12]. Once the target is identified, the GPS coordinates are obtained using the data logs available on the main ground station. Fig 14 : Character Recognition using Neural Network 9. Power system The main driving force of the vehicle is a glow fuel powered 2 stroke engine. Hence, batteries are used only to power the avionics of the vehicle. To determine the batteries and DC to DC converters required for the UAS avionics, a power distribution system was designed keeping in mind different power requirements of each sensor, computer and actuator. The battery was then chosen by computing the peak power drawn by each component and keeping in mind the estimated flight time 16 P age

Table 2 : Power Distribution Voltage (V) Current (ma) Components SBC TS7800 5 1000 IMU MNAV100C 5 150 GPS Jupiter 30xLP 5 100 Wireless Maxstream Xtend RF modems 5 750 PWM Receiver 5 50 RxMux Safety

17 Table 2 : Power Distribution Voltage (V) Current (ma) Components SBC TS IMU MNAV100C GPS Jupiter 30xLP Wireless Maxstream Xtend RF modems PWM Receiver 5 50 RxMux Safety Switch Magnetometer 5 25 AtMega 128 Servo Controller Board Atmega 128 Servo Controller Board Video Transmitter Looking at the above table, a Lithium Polymer battery of 14.8V, 3300mAh was chosen as the suitable battery for a flight of around 1 hour. Astrodyne s DC DC Convertors (14.8V to 12V and 14.8V to 5V) were used to convert the DC Voltages to required levels for the various components. Fig 15 : Power Distribution Board 17 P age

18 A PCB has been designed to serve as the common distribution board for powering all the components on the UAS. 10. Safety Features Various features were incorporated into the UAS to provide increased reliability and safety Modifications made to the airframe Based on the payload the airframe was supposed to carry, several modifications were made to the airframe while constructing it. These include The usage of heavy duty hinges and stronger control horns for all the control surfaces. The control mechanism for all the control surfaces was also changed by using stronger and more reliable push rods with rod end bearings at the ends. High torque metal geared servo motors were used as actuators for controlling these surfaces. The servo motors for the rudder and elevator were relocated to the rear of the vehicle so as to free the space in the fuselage for other components. To reduce the effect of engine vibrations on the components inside, special engine mounts were used. Also, mounts were made for the Single Board Computer (SBC), the IMU (Inertial Measurement Unit) and the Magnetometer for their installations in the aircrafts during the test flights. The placement was done so as to ensure that the CG of the aircraft wasn t disturbed from its original position. The fuselage of the aircraft was strengthened using hardwood to provide additional support during landing, due to the heavy payload Onboard Safety Features In case of autopilot failure, the control of the servos shifts to the safety pilot using a manual override switch. The switch is based on RxMux by Reactive Technologies. In case of primary battery discharging or failure, an additional backup battery was provided. The EMS Jomar Ultimate Battery Backer detects the drop in voltage of the primary battery and automatically switches to the secondary battery when needed. To make sure that the servos were not triggered by false signals, an EMS Jomar Glitch Buster was used. An onboard glow igniter was used to ensure that the glow plug always remains heated and combustion takes place during flight. 18 P age

19 11. Conclusion For this year s competition, a low cost autopilot system was created, along with supporting software and hardware. In doing so, team members built up valuable systems engineering experience, while at the same time, solving many technical challenges through team work and innovative ideas. During the course of this project, procurement of resources was a major bottleneck as almost all the components had to be imported thereby constraining the development time. Since almost all the systems have been custom built, a great deal of modifications could be made, which would not have been possible with the use of Commercial Of The Shelf (COTS) components. 12. Acknowledgements We would like to thank the following members of Delhi College of Engineering for all of the Invaluable guidance and support that they have given our team: Prof. P.B Sharma (Director, DCE) Prof. R.K. Sinha (Dean, Industrial Research and Development) Dr. D.S. Nagesh (Faculty Advisor) Mr. N.S. Raghava (Faculty Advisor) At the same time we would also like to state our appreciation for the following Graduate students of our College who took part in the 2006 International Aerial Robotics Challenge (IARC) at Georgia. We benefitted greatly from their experiences and technical expertise: Arvind Jayaraman (M.S Control Systems, University of Michigan, Ann Arbor) Sahil Malhotra (Engineer, Hitech Robotic Systemz, New Delhi) Pranav Kedia 19 P age

20 13. References 1 aerotools.de/airfoils/javaprop.htm Airplane Stability and Automatic Control by Robert C Nelson Derek R. Nelson, D. Blake Barber, Timothy W. McLain, Randal W. Beard, "Vector Field Path Following for Small Unmanned Air Vehicles," American Control Conference, Minneapolis, Minnesota, June detail.php?product=ts Alexander J. Faaborg, Using Neural Networks to Create an Adaptive Character Recognition System, March Fundamentals of Neural Networks Architectures, Algorithms and Applications by Laurene Fausett. 13 Computer Vision: A Modern Approach by David A. Forsyth and Jean Ponce. 14 Digital image processing using Matlab Rafael C. Gonzalez, Richard E. Woods, Steven L. Eddins. 15Hierarchical Path Planning and Control of a small fixed wing UAV: Theory and Experimental Validation by Dongwon Jung (Georgia Institute of Technology, December 2007) P age

FLCS V2.1. AHRS, Autopilot, Gyro Stabilized Gimbals Control, Ground Control Station

FLCS V2.1. AHRS, Autopilot, Gyro Stabilized Gimbals Control, Ground Control Station AHRS, Autopilot, Gyro Stabilized Gimbals Control, Ground Control Station The platform provides a high performance basis for electromechanical system control. Originally designed for autonomous aerial vehicle