GTRI

Collaborative Unmanned Systems...

Collaborative Unmanned Systems Technology Demonstrator (CUSTD)

The Georgia Tech Research Institute has developed an unmanned systems platform for performing collaborative autonomy research. The primary research aim of this platform is to explore distributed, collaborative, autonomy algorithms for controlling multiple unmanned systems vehicles. The system consists of multiple small, Unmanned Aerial Vehicles (UAVs) and an Unmanned Ground Vehicle (UGV) that can operate autonomously and accept missions from a ground station. The system leverages off-the-shelf technologies and open standards, such as JAUS, and can be configured to carry different scientific payloads.

Often in scenarios involving multiple, heterogeneous agents, different agents may carry different sensor payloads or have different capabilities. As such, agents may be better suited for different roles and tasks. In real world scenarios, it is useful for agents to collaborate with each other to perform dynamic task allocation for a given mission. The CUSTD platform is used to perform such collaborative, heterogeneous and autonomous missions. The vehicles can send messages over the communications links to each other and negotiate task assignments in a distributed fashion, based on their capabilities and the mission.

The system architecture leverages open standards and inexpensive hardware and allows for ease of payload configuration and algorithm development. In addition, open software standards and messaging protocols (such as JAUS) are used to provide a platform that is easily extensible and can be integrated into other approaches. The vehicles carry multiple sensors and payloads and can relay data to each other or to the open source FalconView™ based ground station. Using the ground station, operators can track each vehicle simultaneously and view sensor information from these systems in real-time.

Download a printable brochure here - 600Kb

Watch Video: UAV-UGV Collaboration

Watch Video: UAV Platform

Platform Description

The UAV flight duration is up to 90 minutes, depending on mission tasks and payload. In addition to the avionics package, each airframe can carry up to 7 lbs of payload, such as the mission processor and video cameras or other sensors. The UAV runs autonomous behaviors on-board using the mission computer and processes camera images in real-time using computer vision algorithms. The UAV is also equipped with an analog video camera that streams video to the ground for use by ground station operators.

The UGV is a Porsche Cayenne SUV automobile that has been modified for autonomous operation. This vehicle was used in the 2007 DARPA Urban Challenge and is equipped with numerous sensors, including stereo vision and LIDAR sensors for obstacle detection and GPS for waypoint navigation. The vehicles all communicate with each other and the base station using JAUS standard messages.

Configurable Payload

The UAV payload area can be configured to carry different scientific sensors or other payloads (such as infrared cameras), as dictated by mission requirements. In the current configuration, each UAV carries a digital camera for processing on-board computer vision algorithms. These images can be processed and used as input to the autonomous behaviors of the system or analyzed offline. For offline processing, they can be stitched together and geo-rectified for importing into the FalconView™ based ground station.


Click for larger version

Behavior Based Architecture

The on-board mission computer runs high level autonomous behaviors and communicates with the avionics system to send low level waypoint or turn commands. The behaviors process telemetry and sensor input and can send messages to other vehicles cooperating together in the mission, using the communications link. System components communicate using a publish-subscribe message passing architecture, and new components or behaviors can easily be added to the system. Current behaviors that support missions include vision targeting behaviors and auction-based task allocation behaviors.

FalconView™ Integration

The system is integrated into a FalconView™ based ground station. FalconView™ is widely used by the US Department of Defense for its aircraft mission planning and mapping capabilities. The open-source version of FalconView™ is now freely available. The vehicles in this system communicate to a ground station server and FalconView™ plug-in to display vehicle position and telemetry information in real time, using JAUS messages. The ground station can also be used to send JAUS messages to the vehicles and to send other waypoint and mission based commands. The plug-in also displays sensor data (such as streaming video from the UAVs) in real time.


Click for larger version