EXPERIMENTAL FLYING AUTONOMOUS VEHICLE

Transcription

1 EXPERIMENTAL FLYING AUTONOMOUS VEHICLE Bharamee Pongpaibul MEng Cybernetics, ABSTRACT Flying robots have had rapid advances in the last few decades; this is due to the miniaturisation of electronic components and the emergence of new applications such as search and rescue, surveillance and remote inspection. This paper describes the design and construction of a computer controlled small scale flying vehicle and also the novel aspects of sensors used to feedback to the controller. As this flying vehicle will be the base of a new platform used to demonstrate to control engineering students how different control techniques can be used to improve the performance of the vehicle in the laboratory, therefore safety is paramount. The design and construction of the vehicle shows that safety has been a primary objective. 1. INTRODUCTION Flying vehicles of the past has always been large and expensive [1], and therefore unsuitable to be a test bed for control strategies in a laboratory environment. By reducing the size it should be possible to use it as a test bed for control strategies in the laboratory, so therefore giving control engineering students a more interesting system to implement different controllers and observe its affects. A small scale control test bed has many advantages over a large scale vehicle. The most important advantage is the cost of the vehicle, as a small scale vehicle can be constructed with a few hundred pounds. The complexity of the vehicle is quite low for a small scale system therefore it is suitable for control engineering undergraduates. In past projects remote controlled (R/C) helicopters has been the base of the designs, this is because these helicopters are readily available, have known flight dynamics and being reasonably easy to control even manually by an operator. But it is not suitable to work in a laboratory environment because of the big rotor blade size. They also rely on the known dynamics of the system rather than applying a large magnitude of automation as in [2] [3]. There has been no attempt to design a cheap small scale computer controlled vehicle that can be position controlled so this provides a challenge to use novel sensors to measure the position of the vehicle. Accelerometers are used as the main sensor being fed back to the controller; by double integrating the output from the accelerometers the position of the vehicle can be found. The interface to the computer is provided by the PCI- DAS1200 computer board, and the signals from the optical encoders are fed into the PCI-QUAD04 board. MATLAB is the main software that is used to control the vehicle. There are many design decision that affects the basic form of the vehicle, the most important being the number of propellers. After making many design decisions the final design consists of two propellers mounted horizontally, and the direction of thrust can also be redirected by using servos connected to the motors that are powering the propellers. The vehicle must be able to operate in close range with people within the laboratory, therefore a protective cage must be installed to enclose the propellers Design 2. DESIGN AND CONSTRUCTION There most important design decision that has to be made was the number of propellers on the vehicle, by looking at the advantages and disadvantages of each design a decision was made, the finalised design has two propellers. The advantage of this design is that the reaction torque of the motors is cancelled as the propellers rotate in different directions. This is also the case for the four propeller design but with that design the weight and cost of the vehicle would increase because two more motors are added. With two propellers the vehicle can change its pitch by changing the speed of one motor and keeping the other one the same. A problem emerged with this design, as the vehicle cannot exert forces that will stabilise the vehicle in the axis perpendicular to the frame. The solution to this problem is to attach servos to the propellers so that the direction of thrusts of the propellers can be changed; this redirection of thrust is restricted to one plane, because the vehicle can be controlled in the other plane using the same method as for changing the pitch. Another design decision was whether to used small ready made ducted fan as the thrust, and manufacture all parts from very light material so that the vehicle would be classified as a micro air vehicle [4]. This has the

2 advantage of the propellers being already protected and so no more time has to be wasted on building the protection for the propellers and also the data sheet of the ducted fans gave the maximum thrust of the fan. It was decided that this design was not suitable because it is not possible to get two ducted fans of the same make and model spinning in opposite direction. Therefore electric helicopter motors were used for this project. The motors are manufactured for used on electric R/C helicopter so they are made to be powered by Nickel-Cadmium battery packs which provide a large amount of current in a short period of time, a ten cell battery pack can supply about 30 Amp. For this reason high current power supplies are needed to power the motors. Preliminary experiments were carried out to discover if the motor has enough torque to overcome the air resistance of the propellers to rotate at the right velocity. After the initial results it was found that a direct drive coupling between the motor and the propellers were unsuitable as the motor did not rotate up to its rated maximum speed of 25000RPM. A gearbox was introduced with a gear ratio of 1:2.5, with this gear ratio the propellers rotates at a maximum of 10000RPM and with more torque to turn the propellers. The gearbox used was a gearbox made to be used on R/C airplanes and was widely available. bending a thin strip of aluminium into a circle bigger than the diameter of the propellers, to support this struts are introduced to attach to the side to give the cage horizontal strength. The aluminium used was very thin so on its own it cannot give a rigid and strong structure. To combat this, the aluminium was bent into box shape so that the strength is increased but the weight does not increase significantly. The frame was made from pieces of U shaped aluminium screwed together so that the box shape can be achieved. As the whole motor unit is quite heavy and therefore it cannot be supported by the servo alone so the frame is extended to support the weight of the motor unit from the opposite side. Motor Figure 1. Simple model of the vehicle Figure 1 shows the finalised design composes of three parts, the first part is the main frame of the vehicle, the will support all the position sensors and most of the electronics needed. The servos that tilt the motors are attached to each ends of the frame. The second part is the motor casing, this supports the motor and is the part the servos can attach to and also where the optical encoders are fixed to. The third part is the protective cage for the propeller, and this is the compulsory objective of the project because it must be safe to operate in laboratory environment. It must protect the obstructing object so that if the controller implementation is unsuccessful then there will be no damage caused by the vehicle Construction Servo Frame Propeller casing The material used to construct the vehicle was aluminium, this is because aluminium is light and easy to manufacture [5]. The side of the cage was made by Figure 2. Motor casing and servo mount construction Gimbals must be made for the accelerometers, reasons for this is explained in the sensors section. The gimbals is made up of three sections, the first is the outer casing, this is made out of a thin strip of aluminium bent into box shape, it supports the axel coming from the middle section. The middle section is made out of high density foam, and it accommodates the bearings for the two axes. The inner section is where the circuit board of the sensors sits. The gimbals allow the circuit board to stay horizontal when the vehicle tilts. Figure 3. The gimbals construction

3 The propellers used were ordinary propellers for R/C airplanes. The choice of propellers was rather limited because it is difficult to acquire two propellers of the same make and model that rotates in opposite directions. Figure 4. The final layout of the vehicle 3. SENSORS On a big scale project a GPS system would have been very suitable method to measure the vehicle s position, but on a small scale GPS is not accurate enough to control the vehicle to within a few centimetres, and also these GPS systems are very expensive. At the planning stage of the project many ideas were put forward for a method of measuring the position of the vehicle. Accelerometers were not one of the options because they were believed to be expensive. Ultrasonic sensors were thought to be the most suitable method because the sensors are relatively cheap. This method would involve using triangulation technique to calculate the position and height of the vehicle. Three ultrasonic transmitters would be set up on the ground, the receivers on the vehicle would receive the pulse, and the time of flight of the pulse would determine the distance, once three distances from the base stations are known then the position can be found. After some more research into accelerometers on the internet a very cheap dual axis accelerometer were found to be more suitable than this system. The vehicle incorporates many sensors to measure its position the most important being the accelerometers. These accelerometers can also measure static acceleration (the Earth s gravity) therefore it can also be used as a tilt sensor, this is an advantage over using solid state gyroscopes to measure tilt because the accelerometers output absolute tilt instead of tilt velocity, and also because the gyroscopes gives out a very noisy signal. Since they can measure tilt and acceleration their output can be a mix of both and there is no method of separating the signal. To separate the signals there must be two accelerometers, one measuring the tilt and acceleration and the other just pure acceleration. And then the mixed signal can be subtracted from the acceleration alone to get the tilt. To be able to have the accelerometer giving the acceleration signal only, gimbals is needed. This will allow the accelerometer to be horizontal at all times and therefore measuring just the acceleration only. The accelerometer measure acceleration on two axes therefore the vehicle has three accelerometers, one to measure the tilt, and one to measure acceleration in the x and y axis and the other in z axis. By integrating the signals from the sensors the position from the start can be calculated. To measure the heading of the vehicle a solid state gyroscope is used. This cheap gyroscope does not give a clean signal and is not very accurate but because the heading of the vehicle is not as damaging to the stability of the vehicle as the pitch and roll, it can be tolerated. The servos used to control the position of the motor has on-board control so when the computer sends a signal to the servos to move to a certain position it can be assumed that the servos move to that exact position. Lastly the angular velocities of each of the motors are needed, this is achieved by using optical encoders, and they consist of a code wheel and reader. The code wheel is attached to the bottom end of the motor shaft. The signals from the encoders do not go to the same interface card as the rest of the sensors because the specialised card for taking signals form optical encoder are supplied. To summarise, the controller will have the following measurements available: the angular velocity of each motors, the heading of the vehicle, the acceleration in x, y and z axis, and the tilt of the vehicle Control Hardware 4. CONTROL The vehicle is computer controlled so all the signals from the sensors are fed to the computer. There are two interface boards that are used to interface between the computer and the electronics. The hardware used to control the vehicle is shown. The arrows indicate the direction the data is transferred. PC PCI- DAS120 0 Motors and Servos PCI- QUAD04 Optical Encoders Accelerometers and Gyroscope Figure 4. Hardware setup

4 The two analogue output channels on the PCI- DAS1200 board are connected to the motors. The software sends out a pulse width modulated signal to the motor to control its angular velocity. The digital and analogue inputs on the board are used to acquire the signals from the accelerometers and gyroscope. To control the servos, a PIC microcontroller is used to translate the signals from the digital outputs of the computer board to the servos. This is needed because there is no way of controlling the servos using MATLAB /SIMULINK. All the PIC does is it waits for a digital signal from the software and then converting it to a timed pulse width so the servo will know the position to relocate to. The PCI-QUAD04 computer board is a specialised board for receiving signals from optical encoders. All the protocols of the way optical encoders send information is built into the board so it is easy to interpret information from the encoders in the software Control software The computer interface boards are represented on SIMULINK and therefore it is straightforward to control a specific port on the board without any extra effort. The inputs for the controller are the angular velocity of each motor, and the position of each servo. For the operator to control the vehicle the software present the operator with a two dimensional environment and a representation of the vehicle, the operator can specify a point within the environment and a height, the program calculate the inputs needed to manoeuvre the vehicle to the desired position, the inputs are then passed to the controller. The controller always monitors the actual position of the vehicle so motion of the vehicle is always stable and controllable. There are three steps to design the software to control the vehicle. The first is to derive the equations that define the vehicle dynamics. These equations are derived from simple Newton s law of motion. Once the force equations have been derived it can then be applied to state space techniques so a model of the system can be constructed. The model consists of 4 matrices, where two matrices are the input matrices and the other two is the output matrices. It is not possible for the model to be an accurate representation of the system because there are a lot of unknown forces involved that cannot be predicted or measured, for example, the impact the weight of the cables on the system dynamics, therefore these unknowns will be estimated and then the controller can update its parameters from the feedback data received in real time. The system is a non linear system so there will be some linearization around a suitable point. The second step is controller design. Different controllers have to be implemented so the performances can be compared. The two control methods that will be implemented are Proportional/Integral/Differential (PID) control and optimal control. The final step is to simulate the system using the model. The desired input is entered into the software and the output is observed to see if the controller has achieved the desired affects. Once the desired affects has been achieved the output of the controller can be disconnected from the model and connected to the real system. The vehicle can now be controlled by the operator. 5. TESTING There will be two tests for the system, the first is a stability test. To carry out the test the desired input to the system has a height change, so therefore the effect of this is to let the vehicle hover in midair without changing the horizontal position. By superimposing the desired input plot on top of the actual output the effectiveness of the controller can be observed. Once stability has been achieved using the PID controller, system identification techniques can be used to identify a more accurate model than the estimate used when the PID controller was designed. This will result in more accurate and better performance for implementation of future controllers. When the simple hover has been performed and verified that the system is stable, disturbances can be introduced, by wind or hovering to a greater height so that the cables will have a larger effect on the system. The second test will be to test the position control of the vehicle. The input will be the desired position of the vehicle, which should be different to the present position. The vehicle will move to the desired position and a similar graph to above can be plotted and again the controller can be changed and the test repeated. The graphs shows the strengths and weaknesses of each controllers so it can be decided which controller is most suitable for which situation or environment. 6. FUTURE WORK The goal of this project was to demonstrate the concept of using flying vehicle to teach control engineering graduates the importance of controller design. Given the time and budget constraints there are many areas that have not been covered. These areas include: On-board control On-board power Wireless link between the computer and the vehicle Adding extra sensors to sense the environment Decrease noise level The cables connecting the vehicle to the computer is needed to provide power and control signals but it does influence the vehicle dynamics greatly therefore if it can

5 be removed then the complexity of the control problem will be reduced. The noise level of the project was not an issue at the beginning but after the project has been finished it was realised that the propellers produce excess noise for a comfortable environment, this could have been improved by changing from airplane propellers to helicopter rotor blades. The blades will have less air resistance and therefore produce less noise, to combat the decrease in lift of this solution the rotor blades could rotate faster. This will be possible because as the rotor blades has less air resistance the same amount of torque from the motor can turn the rotor blades at higher velocity. By adding extra sensors that can sense its environment, for example, ultrasonic sensors to detect obstacles, it will allow the controller to plan the routes of the vehicle in a dynamic environment. The weight of the vehicle can also be reduced by changing the construction material from aluminium to carbon fibre. For future work building on this project the above areas should be covered. 7. CONCLUSION It must be stated that at the time of preparing this paper, the vehicle has not been tested fully so all the details for testing and controller design is the planned course of action. There is a slight concern with the amount of thrust the propellers will produce, if it turned out that vehicle cannot achieve lift-off then the first solution is the reduce the weight of the vehicle. This could be achieved by removing the propeller protection which is a third of the weight of the vehicle, but this would negate the primary objective. The vehicle might then have to be redesigned to use as little material as possible. This paper reveals the concept of low budget education equipment for control engineers. This solution is more impressive and cheaper than the commercially available educational tools [6]. This free flying design gives the experiment an interesting challenge to the controller designer which will enhance the control theory education. The development of hardware and software necessary to enable this experimental test bed to be stable has been demonstrated. It has also shown that accelerometers can be used to sense the position of the vehicle in real time so the data can be feedback to the controller. This approach would have been better and more cost effective than using gyroscopes on three axes to measure the pitch, yaw and roll velocity, or the ultrasonic method. The testing of this vehicle should reveal that the vehicle will be suitable for at least two types of controllers. This project has more application than just to educate control engineering undergraduates, it can be used to carry out further experiments on the use of different position sensing methods. Many vehicles can be built so collaborative behaviours on robots in dynamic environments can be implemented and studied. Acknowledgements: The work discussed in this paper was the result of collaboration between the author and Ryan Corkery. The support and advice from the project supervisors, Dr. Victor Becerra and Dr. William Browne, is greatly appreciated. Many staffs from the Department of Cybernetics have lent their assistance especially Mr. Phil Taylor. 8. REFERENCES [1] S.Wegener Et Al, UAV Over-the-Horizon Disaster Management Demonstration Projects, White Paper, NASA Ames Research Centre, February 2000 [2] E. Altug, J.P. Ostrowski, R. Mahony, Control of a Quadrotor Helicopter Using Visual Feedback, Proceedings of the 2002 IEEE international Conference on Robotics & Automation, Washington DC, May 2002 [3] G.R. Gress, Using Dual Propellers as Gyroscopes for Tilt-Prop Hover Control, American Institute of Aeronautics and Astronautics Conference Paper, November 2002, Available: [4] R.C. Michelson, S.Reece, Update on Flapping Wing Micro Air Vehicle Research, 13th Bristol International RPV Conference, Bristol England, 30 March 1 April 1998 [5] J.M.Gere, S.P.Timoshenko, Mechanics of Materials (4 th edition), Stanley Thornes (Publishing) Ltd, 1999, pp [6] 3 Degree of Freedom Helicopter, Specialty Control Challenge, Quanser Consulting, Available: