Intelligent Control of Multiple UAVs

Overview

This research project is a collaboration of the Penn State Aerospace Department and the Applied Research Lab. The current objective is to construct and flight test a fleet of unmanned aircraft in order to provide a platform for intelligent and collaborative control research. Large R/C trainer aircraft ("Almost Ready to Fly" kits) are outfitted with a Cloud Cap Technology Piccolo Plus autopilot and a single board computer to run the intelligent control (IC) program. Currently, only one airplane is equipped with the single board computer required to run the intelligent control software. Once it is flight tested, more aircraft will be equipped. Only two airframes are equipped with autopilots, however a third is being built. We are currently just beginning the intelligent control flight test phase. After successfully demonstrating autonomous flight under the direction of the IC, we will transition to flying multiple aircraft simultaneously. The following figure shows the objective: multiple aircraft communicating with each other and the ground, autonomously making decisions to accomplish the mission and accommodate deviations from the plan.

(Top)

Hardware

Aircraft The Sig Kadet Senior is a large trainer aircraft which is highly stable. It has an 80 inch wingspan and weighs around 13 lbs fully loaded with the autopilot and intelligent control computer. It is powered by an O.S. .91 cubic inch glow fuel engine and can carry a 32 oz fuel tank. The pictures below show our existing aircraft: a trainer aircraft (a stock kit), a camera equipped trainer for testing the camera, UAV-1 (equipped with only the Piccolo autopilot), and UAV-2 (equipped with both autopilot and IC computer). A third airframe capable of carrying both the piccolo and the computer is currently being built. Visible in the pictures are the pitot-static probe on the wing and the communication and GPS antennae on the fuselage. UAV-2 is also equipped with a 2.4 GHz antenna for wireless networking. Autopilot The Piccolo autopilot from Cloud Cap Technology is based on the Motorola MPC555, a 40 MHz Power PC processor. It is capable of attitude estimation using 3-axis gyroscopes and accelerometers and contains integrated GPS. It uses proportional-integral-derivative feedback to control the aircraft. Multiple autopilots can be controlled though a single laptop connected to one groundstation. The Piccolo is capable of hardware in the loop simulation, which greatly reduces flight testing risk. The following photos show the Piccolo itself and the installation in UAV-2. In the second picture behind the Piccolo, the IC computer is visible. Single Board Computer The single board computer used is the Ampro ReadyBoard 800. The SBC uses a 1.4 GHz Intel Pentium M processor and consumes a maximum of 12 watts continuous power at 5 VDC. It includes a flash based system disk, wireless networking, and a serial port for communication with the Piccolo. The SBC will run the Intelligent Controller program and any other data processing programs. Cameras We have successfully flown the Nikon A510 digital camera.

(Top)

Software

Intelligent Controller (IC) The IC software is designed to process sensor data and issue commands to the autopilot. Two main modules handle these tasks: Perception and Response. The perception module creates an internal representation of the external world. The response module plans and executes commands in real time based on the situation awareness of the perception module. The IC communicates commands to the Piccolo though a serial connection. It also sends data back to its groundstation via the wireless network. This software was initially created by the Penn State Applied Research Lab. Piccolo Simulator to Simulink Interface Normally, the autopilot gains are adjusted in a trial and error fashion using hardware in the loop simulation. In order to develop more accurate gains in a quicker fashion, the ability to create a linear model from the Piccolo simulation was required. Since the simulation model for out UAV was already developed for the Piccolo simulation, an interface from the simulation to MATLAB/Simulink was developed. This allowed linear models to be created and feedback gains to be created from the resulting transfer functions. It also enable the Piccolo simulator to be run as a block in Simulink. The use of the linear models did not seem to speed up autopilot gain development. While linear model generation is useful, the trial and error method works well and quickly for a stable aircraft like the Sig Kadet Senior. Open Glass Cockpit Interface with Groundstation Telemetry In order to gain familiarity with the Piccolo communications protocol, and interface between the Operator Interface program (the ground control program for the Piccolo) and Open Glass Cockpit was developed. OpenGC is a program that displays an artificial horizon given a telemetry stream of roll, pitch, yaw, speed, altitude, etc. This allows a real time artificial horizon to be displayed on the operator's computer. A sample is shown below.

(Top)

Flight Test

We have completed multiple flights with UAV-1 (Piccolo only), while only suffering one crash. The fuselage was relatively undamaged, while the wing had to be replaced. UAV-2 will fly shortly. What follows are some aerial photos from the camera test plane, and a GPS track from UAV-1 superimposed on a satellite image of the flying field. The landing strip is outlined in red. These last pictures show a test of a simple image processing algorithm coded in MATLAB. The camera plane flew over a grouping of four posterboards of different color. The algorithm finds the red posterboard and outputs its location in the picture.

(Top)

Related Literature

Miller, Jodi A., Minear, Paul D., Niessner., Albert F., Jr., DeLullo, Anthony M., Geiger, Brian R., Long, Lyle L., Horn, Joseph F., "Intelligent Unmanned Air Vehicle Flight Systems", Infotec@AIAA Paper, Sept 2005

(Top)