Military

DARPA puts out call for super-agile UAVs

DARPA puts out call for super-...
DARPA is seeking a UAV that can operate in debris-strewn close quarters (Image: DARPA)
DARPA is seeking a UAV that can operate in debris-strewn close quarters (Image: DARPA)
View 1 Image
DARPA is seeking a UAV that can operate in debris-strewn close quarters (Image: DARPA)
1/1
DARPA is seeking a UAV that can operate in debris-strewn close quarters (Image: DARPA)

Unmanned Aerial Vehicles (UAV) are the eyes in the skies for soldiers and disaster relief crews, but despite over a century of aviation progress, they still leave a lot to be desired and close quarters are very difficult for them to navigate on their own. To make UAVs more practical in debris-strewn areas, DARPA's Fast Lightweight Autonomy (FLA) program aims to develop algorithms that will allow autonomous fliers to negotiate obstacles as easily as a bird of prey.

Goshawks and other raptors put aviation engineers to shame. These nimble birds can fly with alarming speed through thickly wooded forests, dive through thickets, and come to a perfect landing on a branch in a maneuver that would leave a UAV a twisted pile of plastic and electronics at the first tree trunk. And the bird does it as a matter of routine.

This is the ideal that DARPA wishes to emulate as it makes a Broad Agency Announcement soliciting proposals for the FLA program. The idea is that the participants will be given a standardized small UAV that they will have to develop algorithms that will enable them to fly through an urban obstacle course of rooms, stairways, corridors, and other obstructions at 20 m/s (45 mph) without a human piloting it, or using GPS, remote sensors, or other outside navigation aids. In other words, if the bird doesn't have it, neither does the airborne robot.

The goal is to develop an autonomous system for the UAV to make it easier to use by requiring much less human intervention, so the operators can concentrate on the mission at hand rather than keeping the robot from banging into things. The plan is for the UAV to develop a sort of "muscle memory" – a sort of electronic reflex to help it perceive and quickly avoid objects without supervision.

DARPA says that if FLA is successful, it could lead to the system finding applications not only in UAVs, but in marine and land-based systems as well. The agency will hold a Proposers Day on January 6, with proposals accepted until January 2.

The DARPA/BBC video below shows a Goshawk's view of flying through a woodland.

Source: DARPA

2 comments
Nairda
First of all, a Goshawk has nothing on a bat when it comes to high speed dodging. And that is the key to how this will work. A bat uses its sonar as an active mapping tool in real time to avoid objects and track others. The developers of Hector had the right idea: (http://www.gizmag.com/hector-stick-insect-robot-walking/35296/)
If the body knows it needs to go from A to B and has an idea on the path, its various sub systems (legs, wings, flaps, impellers) can then instinctively react to make it so.
For the FLA what this means either an active LIDAR or ultrasonic avoidance system. It will be slow at first, but if programmers can get away from the idea of micromanaging movement and instead instilling a set of basic avoidance rules/behaviour they will be better off for it.
Donald Vitez
Giving the participants a standard small UAV limits the outcome of the program. The solution that DARPA is seeking will be a result of developments in not only software but hardware. Having the algorithms to avoid obstacles is meaningless without hardware to enable enhanced agility. Perhaps equipping a multi-copter with conventional control surfaces such as a rudder and ailerons will permit enhanced yaw and pitch control, allowing the UAV to avoid the obstacle once the combined sensory and software have determined the existence and position of the obstacle relative to the UAV