Georgia Tech Aerial Robotics
From Aerospace Controls Group
The Georgia Tech Aerial Robotics Team (GTAR) is a team of students, faculties, and staffs of Georgia Institute of Technology who represents the Institute in the International Aerial Robotics Competition (IARC). The GTAR team has been participating in IARC since its debut in 1991, and also hold the record of winning the most missions. The GTAR team also participated in the Spring IMAV 2011 competition at which the team won first place in both outdoor events.
Latest News: GTAR team wins the 2011 outdoor Spring IMAV competition. Full news release.
The team is now working on IARC's 6th mission which is focused on GPS-denied indoor flight. Finished as the leading entry in 2010, the team was able to develop an indoor flight vehicle that can autonomously enter and explore indoor environment. The team plans to complete the mission in the 2011 competition. Volunteers are needed! Georgia Tech's students, faculties, and staffs are welcome to contribute, regardless of their majors.
Sponsored by the Association for Unmanned Vehicle Systems International (AUVSI), the International Aerial Robotics Competition (IARC) tasks university teams to develop autonomous Unmanned Aerial Vehicles (UAVs) to accomplish a specified mission. At the time that each mission concept is released, the mission is said to be beyond the capability of any existing system including
"super-power military machines". The competition is now on its 21st year. The current mission is the 6th mission focusing on developing a GPS-denied autonomous indoor flight vehicle.
Contents |
Mission 6
The 6th Mission of International Aerial Robotics Competition requires the teams to develop autnomous indoor flight vehicles capable of entering an office-like arena, search for the USB flashdrive, pick up the USB flashdrive, replace the flashdrive with its decoy, and vacate the arena. The mission was started in 2010 at University of Puerto Rico, Mayaguez, PR and will continue into 2011 at the University of North Dakota, Grand Forks, ND.
2011
Pleased with the initial performance of the GTQ in 2010, the GTAR team sought to develop additional systems to accomplish the goals of the mission. A Point Grey Firefly MV USB camera was attached facing downward on the vehicle, and a second Gumstix Overo Fire and Tobi board was added to process image data. Image processing algorithms were developed to detect whether the USB flashdrive was in view, estimate its location relative to the vehicle, and descend onto that location. Also, an undirected graph guidance system was developed for improved exploration.
GTAR entry explored furthest into the arena and received the award for best overall system.
2010
| GTQ | |
 | |
| The GTQ is a quadrotor designed for autonomous exploration in indoor environment. GTQ runs Simulaneous Localization and Mapping (SLAM) navigation and frontier-based guidance onboard. | |
| Airframe | |
|---|---|
| Basis | Ascending Technologies Pelican |
| TOW | 1.6 kg |
| Endurance | 10 min |
| Payload Capacity | 0.5 kg |
| Onboard Avionic Systems | |
| FCS | Gumstix Overo Fire with Tobi expansion board Atmel ATMega 128 Analog Devices ADIS16365 IMU MaxBotix MaxSonar-EZ1 |
| Datalinks | WiFi 2.4 GHz , RC 72 MHz |
| Payload Sensors | Hokuyo URG-04LX laser scanner |
In 2010, GTAR team determines that the GTLama used in the 2009 competition is not suitable for the increasing demands of the new mission. GTAR team seeks for a flight vehicle that is expandable, easy to maintain, and can carry more payload than the GTLama. This results in development of the GTQ.
GTQ is a quadrotor that uses Ascending Technologies Pelican as a base airframe. GTAR uses structures, motors, motor controllers, and propellers from AscTec without modification. GTAR then supplements the vehicle with processors and sensors for autonomous operations. The sensors include Hokuyo URG-04LX laser range scanner, MaxBotix MaxSonar-EZ1 sonar range finder for altitude measurement, and Analog Devices ADIS16365 inertial measurement unit. The processors include Gumstix Overo Fire with Tobi expansion board running guidance, navigation, and control software, and Atmel ATMega128 running stability augmentation software.
In lieu of Global Positioning System (GPS), GTQ uses scan data from the Hokuyo laser range scanner to do Simultaneous Localization and Mapping (SLAM) onboard. SLAM algorithm builds a map while probabilistically determine vehicle's position and heading based on the evolving map. GTQ explores indoor environments using frontier-based approach. The vehicle is commanded to fly towards unexplored area.
The 2010 competition was held on August 13th at University of Puerto Rico, Mayaguez, Puerto Rico. Participating teams include South Dakota School of Mines & Technology, University of Michigan, Embry-Riddle Aeronautical University, Oregon State University, Pima Community College, and Indian Institute of Technology, Madras. GTAR finished as the leading entry as the team was able to autonomously enter the arena, explore, and land in the arena.
Videos of that are linked to in our general list of selected movies. One of the videos is q100813c1_attempt3.wmv, showing an attempt with a planned simulated self-destruction sequence.
Mission 5
The 5th Mission requires university teams to develop indoor flight vehicles that can autonomusly enter an office-like arena, explore the area, search for a control panel and transmit a picture of the panel back to the ground station.
This indoor flight mission is considered a great paradigm shift in UAV community for several reasons:
- Radio links in indoor environments are unreliable. Global Positioning System (GPS) signals cannot be used for navigation.
- Payload choices are very critical due to weight constraints.
- Onboard computational power is very limited.
- Although many techniques from ground robots can be applied, indoor flight vehicles have a rather agile dynamics and a much more complicated six degress of freedom motion compared to ground robots.
The competition was held in 2009 at University of Puerto Rico, Mayaguez, Puerto Rico. Massachusetts Institute of Technology's team won the competition.
2009
| GTLama | |
 | |
| Build by the GTAR Team in 2009, the GTLama only uses the inherent stability of the coaxial airframe and a minimalistic sensor suit to achieve full indoor autonomous operation. | |
| Airframe | |
|---|---|
| Basis | Esky Co-Axial (Big Lama) |
| Dimensions | 60x60x30 cm |
| Endurance | 12 min |
| Onboard Avionic Systems | |
| FCS | ATMega128, Sonar, 4x IR Range Sensors |
| Datalinks | 900 Mhz Xbee, 2.4 GHz Video Link |
| Payload Sensors | Analog Camera |
Prior to the 2009 competition, there was no vehicle in the UAV Research Facility that is capable of indoor flight. GTAR took a simple and low-cost approach and developed GTLama. Based on ESky E020 “Big Lama” coaxial helicopter airframe, GTAR supplements the vehicle with Atmel ATMega128 microcontroller, sonar range finders, infrared range finders, and video camera. Coaxial configuration was chosen due to its inherent stability. GTLama uses range sensors to measure distance from walls. The vehicle is commanded to follow a wall.
The competition was held at University of Puerto Rico, Mayaguez, Puerto Rico. At the competition, GTLama was able to enter the arena and follow walls. The team from MIT was able to complete the mission. This is the first time a mission is completed on its first year.
Mission 4
The 4th Mission of International Aerial Robotics Competition required UAVs to perform several complex tasks over a large area. The mission required autonomous UAVs to fly a specified trajectory that is approximately 3km in length. At the end of the trajectory, the vehicles will arrive at an abandoned village. UAVs shall locate a specific building within the village. The building is marked by IARC sign. The UAVs or subvehicle shall enter the building and transmit video data back to ground station.
The mission was held at U.S. Army Soldier Battle Lab, McKenna Urban Operations Site, Fort Benning, Georgia. The mission started in 2001 and remained a significant challenge to comeptitiors until 2008. Although no team was able to execute the full mission in 2008, GTAR team was considered to accomplish the most, and therefore was awarded as the 4th Mission winner.
2008
| GTMax | |
| The GTMax is the workhorse of the UAVRF with the most flight hours and most research performed on it | |
| Airframe | |
|---|---|
| Basis | Yamaha R-Max |
| Dimensions | 2.0x3.6x1.2 m (WxLxH) |
| TOW | ~ 75 kg |
| Endurance | ~ 1 h |
| Payload Capacity | > 10 kg |
| Onboard Avionic Systems | |
| FCS | 2xPC104, 3 axis IMU, 3 axis Magnetometer, Sonar |
| Datalinks | 802.11 WiFi, 900 MHz Wireless Serial, L3 Video Link |
| Payload Sensors | IMU stabilized pan/tilt with regular and infrared cameras and range finder; Sick LD-MRS scanning LADAR |
Prior to the 2008 competition, the GTMax helicopter was already capable of flying a prescribed trajectory and identifying the window opening of the target building. GTAR team planned to execute the full mission.
The plan was to deploy a mobile ground robot into the building. The mobile robot shall transmit video data. The robot was attached to the end of a boom slunged from the GTMax helicopter.
At the competetion, GTMax was able to fly the prescribed trajctory and located the window as planned. The boom was successfully deployed. However, the mobile robot was instead deployed outside the building. GTAR's performance was considered the best performance overall. The team was awarded mission winner despite that the mission was not fully completed.
2007
Still competing at Level 3, the team brought two redesigned GTRovers, capable of tolerating much higher impact forces in the axial direction, the prime source of damage in the last year. Furthermore, the rover's driving logic had been improved and two additional cliff detecting sensors were added. Equipped with these new sensors, the rover was now able to detect an "upside-down" situation and could initiate a self-rightening maneuver.
Launch attempt one was successful and the rover was deployed correctly form the capsule, starting the driving logic. The rover entered the target room and initiated a 360° panorama picture taking sequence from a promising position. Unfortunately, the wireless transmission of the pictures to the control station was not operational and the pictures were not available to a human operator. The attempt was not successful. (A later recovery of the onboard storage revealed a picture that would have won Level 3.)
The team was able to fix the wireless communication issue for its second attempt. The launch and deployment were successful. The rover transmitted the winning picture shortly after having started its exploration.
The GTAR team finished Level 3 with one unused attempt left and was now ready to attempt Level 4, the complete 4th Mission, as the leading entry in the next year. No other team had yet attempted Level 3.
2006
The team designed a new 2WD rover, now called the GTRover. The utilized hardware was based around a custom build metal frame, housing a Gumstix computer and an Axis 1.3 megapixel webcam, a 802.11g data link, and IR range sensors.
The drive logic, now implemented on the gumstix, mimicked the Roomba motion pattern, utilizing random drive, wall-following, and wall-bouncing modes. Control over the motion pattern allowed for the implementation of a picture taking mode in which the rover would stop and take a 360° panorama. These pictures were then transmitted to the control station for evaluation by a human operator.
The launcher was again modified, now using a interface capsule to house the GTRover during the ballistic flight. The capsule contained a completely mechanical piston system to expel the GTRover after having impacted.
Although three successful launch were made, the GTRover accumulated more and more damage and Level 3 could not be finished.
GTAR remained the only team competing against Level 3.
2005
Having accomplished Level 2, GTAR started investigating the Level 3 problem. As the level allowed for a simulated continuation from Level 2, i.e. whoever completed Level 2 could start directly at the 10 m boundary outside the building, the team investigated solutions centered around ground robots.
Building upon previous ideas, a rubber band powered ballistic launcher was retrofitted to shoot a Roomba based rover into the building.
The utilized rover manufactured from a custom frame, fitted with the Roomba Discovery logic board, sensors, and drive train, as well as an analogue NTSC board camera and video transmitter. Due to the hard to control final attitude after impact, the custom frame was heavily padded and allowed operation in all possible attitudes. The 4:3 aspect ratio camera was mounted rotated in order to maximize vertical coverage, providing continuous live coverage to a remote video receiver. The robot's drive logic was not controllable as it followed the original Roomba's "MAX" cycle.
The lessons learned from this year were:
- analog video does not provide sufficient resolution to solve the gauge reading problem
- picture taking and motion have to be coordinated in order to minimize motion blur
- the ballistic flight path of the rover has to much scatter which renders "aiming" impossible
No other team finished Level 2, GTAR continued to be the leading entry.
2004
The Georgia Tech team took a break in this year after Level 2 had been accomplished in the last year. No other team could finish Level 2, so GTAR maintained its position as the leading entry.
Members
2010
D. Michael Sobers
Simulation, Guidance, and Control
Girish Chowdhary
Ep Pravitra
Avionics
Allen Wu
Hiroyuki Hashimoto
Chester Ong
Roshan Kalghatgi
Airframe
Claus Christmann
Faculty Advisor
Eric N. Johnson
