Primary Aircraft:
Sig Rascal 110
Wing Span: 110 inches
Length: 78 inches
Weight: 13 pounds dry, 15 pounds loaded
Engine: 25cc Gas
Wing Loading: 20oz /sq. foot
Computers Used:
Ground Station / Avionics Interface
Toshiba Satellite Pro
2.2 GHz Pentium IV
512 MB Ram
60 GB HD
NVidia Geforce4 Mobile
AI / Computer Vision Machine
Custom Workstation
Dual Pentium IV Xeon 2.2 Ghz
1 GB Ram
20 GB HD
NVidia Geforce 5200FX
System Concept:
Level One:
Although this level was completed in the past by a previous team, it was neccessary for the current team to gain the experience of completeing this on their own. Software is also needed so that when integration of levels one, two and three is needed to win the competition, a solid component is in place.
All navigation and high level decisions are handled by a custom UA ARC package called JUAV (Java Unmanned Aerial Vehicle). JUAV provides an artificial intelligence module that allows for control and collaboration of multiple aerial and ground vehicles via TCP/IP.
Our level one strategy involves two distinct phases: first, the mission is configured by adding as many waypoints are needed to complete the 3km ingress specified by the rules. After entering this, the JUAV display provides a Level One Judging viewer – which includes several subpanels. One panel contains a map showing the current position of aircraft and ground station as well as the waypoints and connecting paths. Another panel lists all the waypoints entered during the configuration phase along with two numbers: the closest the aircraft has ever come to the specified waypoint and the current distance from the waypoint. This is done not only to provide a judging viewer at competition, but also to allow for debugging and performance analysis of our system during test flights.
Level Two:
This level is the one that is having our full team’s efforts concentrated on this year. As with the first level, the first step is to configure the mission. This is done by loading a geo-referenced image into a piece of software called Joliet that analyzes the image and detects all the buildings in the image and creates a file containing this info. JUAV loads this file and uses the facts stated in the rules concerning the location of the symbol to select individual buildings from this file as candidates for containing the symbol. A file loaded by JUAV at startup containing information about the flight characteristics of the aircraft used in the mission is utilized by the AI to come up with an optimized search path.
The camera system currently being developed features a Sony VISCA compliant camera mounted in a pan-tilt turret that has several features. The first and most crucial is to steer the camera at specific angles to combat problems in aerial imaging such as the aircraft crabbing as well as to precisely aim the camera. The controller for the turret can be set to a stabilization mode, where the turret is commanded to stay aimed at a specific orientation and then uses information from gyros to hold the camera at this position, sending corrections to the actuators. The final feature is a tracking mode, this mode when enabled works closely with the computer vision software to keep a detected image centered in the frame. Specific orientations, zoom and focus settings are determined during the path planning phase described above so that the camera is always pointed at its desired target with optimum settings.
Object and portal detection is accomplished via techniques that will not be described here until after competition this year, due to the fact that it would give a competitve advantage to other teams. The results of which are reported to the AI, which can make decisions based upon it.
Level Three: (UPDATED 1/28/06!)
A team of three mechanical engineering sophmores that had originally intended on starting their own UAV project have joined our team as an autonomous subteam that will operate independant of the rest of our team, and focus solely on the level three components. They have started assembling prototypes and concept designs. When they get the mechanical aspect completed, the team’s main software will step in and make the vehicle autonomous. Details to follow soon.
Level Four:
The level four approach will consist of running the first three levels in sequence in JUAV, as each individual level has been carefully constructed as to allow seemless interfacing and connecting with the others. This will allow us to make a level four attempt as soon as we have completed the first three, due to the fact additional time was spent building a robust framework early on, rather than attempting to glue many unrelated modules near completion time.