Platform Based Design of Unmanned Aerial Vehicles EE249 Class Project, Fall 2001 By Judy Liebman and Cedric Ma Abstract This project combines design philosophies from three different areas: embedded systems design, synchronous embedded control, and unmanned aerial vehicle design. The main objective of this project is the application of the principles of Platform-Based Design to the specific area of UAV system design in order to yield a greatly enhanced overall UAV control system. We demonstrate this application by implementing the final design and testing it using a hardware-in-the-loop simulation. 1. Introduction Implementing a control system on a UAV is a tremendous undertaking. One needs to combine complicated sensors such as GPS and INS, servo actuators, a wireless network, a central computer, and control laws derived from an approximate dynamic model of the vehicle. Current UAV design implementations have usually been developed by focusing on proving the feasibility of concept rather than elegance in design. This oversight often results in an overall system that is unverifiable and inflexible. For example, in many implementations, the embedded software is written exclusively for a particular mix of sensors and actuators. The substitution of a different GPS sensor in place of the original one would mean a major overhaul in the embedded software followed by an extensive verification process. We believe that by incorporating mechanisms for isolating the programmatic details of a particular sensor and actuator from the core control program, these types of changes would be much easier to handle. This modularity allows a UAV to take advantage of the latest developments in sensor technologies with minimum hassle. Therefore, in this project we strive to achieve a new design for an embedded UAV control system that maximizes modularity and enables verification. 2. Design Foundations In this section, we first present the foundations regarding the Platform-Based Design methodology, UAV components and design, as well as synchronous control methodologies, before explaining the details of our combination of the three. 2.1 Platform Based Design Approach The “Platform-Based Design” approach to embedded systems design has been developed to address several key issues in the IC development domain[1] [2]. This type of design strategy, however, is universal and can be described in a general way that will subsequently lend itself to its deployment in the area of UAV control. Figure 1 illustrates the idea of a platform. A platform can be defined as a layer of abstraction with two views. The upper view is from the application space. Here the platform allows a designer, perhaps a control engineer, to develop control applications without having to deal directly with the lower levels. The lower view is from the components and tools available. Here the platform provides specifications that the components and tools need to provide. The main benefit is that the upper view is decoupled from the lower view, and the two interact through a well-defined interface. This decoupling allows for either the applications or the components to be altered, provided the two can still meet in the middle. Figure 1: Platform Based Design A platform instance is the particular component and tool design chosen to implement the platform. It is developed by mapping the functionality needed by the upper layers onto specific components below. The choice of this platform instance may affect the final specifications of the platform. That is, there is a feedback loop connecting the platform specifications and the platform instance. Hence,the application space as well as the component space affects the platform specification, and the overall design methodology is a ‘meet-in-the-middle’ approach. The Application Programming Interface (API) is the front end of the platform that the control designer sees. Therefore, the API is the abstraction of all of the components and tools below and is the interface layer to the application space. The final main idea for general platform-based design is that this design methodology is hierarchical, or fractal in nature. That is, a platform may be comprised of several layers or sub-platforms. Similarly, a platform and its application may be used as a platform instance for a higher level platform. 2.2 Helicopter, Sensors, Actuators The helicopter group has many different models of helicopters. The Yamaha R-50 and R-Max helicopters have been decidedly easier to equip and fly due to their superior payload carrying abilities. Recently, autonomous flight has been demonstrated on one of the R-50’s and the other R-50 and R-Max’s are being prepared for similar flight experiments. We have used a reasonably accurate dynamic model of the R-50 and the control algorithm developed for the R-50. Since the model helicopters have a relatively fast physical response time, precise and detailed sensors are needed to make autonomous flight possible. While a variety of sensors can be used for landing, takeoff, and intricate maneuvers, a setup consisting of a combination of two sensors is best suited for basic autonomous flight. These two sensors are the Inertial Navigation System (INS) and the Global Positioning System (GPS). The INS uses a gyroscopic measurement system and linear accelerometers to provide acceleration and rotation data at a fairly high rate (at 100 Hz). This data can be used to recover the position of the vehicle. However, if uncorrected, the estimated position data drifts unbounded over time and is therefore insufficient if used alone. The GPS uses a triangulation scheme involving multiple satellites to recover position data that is highly accurate with bounded error at all times. However, this data is available at a lower rate (at approximately 4 Hz) than required by the controller, is occasionally unavailable, and is also subject to jamming and radio interference. Therefore, the combination of the two instruments remains the best setup. There are many choices of INS and GPS units in the market. Each may have different data formats, initialization schemes (usually requiring some bit level coding), operation rates, accuracies, data communication schemes, and even data types (velocity vs. acceleration). Needless to say, incorporating a new sensor in the control system requires significant programming. In fact, the differing communication schemes
View Full Document
Unlocking...