DARPA Looking for new Drone Flight Algorithms for Urban use

The US Defense Advanced Research Projects Agency is looking to develop algorithms that would allow small, fast unmanned air vehicles to operate in cluttered environments without the aid of GPS or external communications.

A broad agency announcement issued on 22 December says in situations such as natural disasters – when UAVs are relied on to provide surveillance in buildings because it is too dangerous for humans to enter – a UAV needs to be able to navigate a “labyrinth of rooms, stairways and corridors”.

DARPA is not looking to develop a new UAV through the Fast Lightweight Autonomy (FLA) programme, the agency says – just the applicable algorithms.

The agency is looking for algorithms and software that would allow a UAV to fly through an open window at 20m/s (45mph) into “complex indoor spaces” without having to be controlled by external operators or through using GPS waypoints.

This, DARPA says, would enhance future unmanned capabilities, reducing processing power, communications and human intervention for low level UAV tasks.

link.

A New Collaborative Machine Learning Algorithm

Machine learning, in which computers learn new skills by looking for patterns in training data, is the basis of most recent advances in artificial intelligence, from voice-recognition systems to self-parking cars. It's also the technique that autonomous robots typically use to build models of their environments.

That type of model-building gets complicated, however, in cases in which clusters of robots work as teams. The robots may have gathered information that, collectively, would produce a good model but which, individually, is almost useless. If constraints on power, communication, or computation mean that the robots can't pool their data at one location, how can they collectively build a model?

At the Uncertainty in Artificial Intelligence conference in July, researchers from MIT's Laboratory for Information and Decision Systems will answer that question. They present an algorithm in which distributed agents — such as robots exploring a building — collect data and analyze it independently. Pairs of agents, such as robots passing each other in the hall, then exchange analyses.

In experiments involving several different data sets, the researchers' distributed algorithm actually outperformed a standard algorithm that works on data aggregated at a single location.

"A single computer has a very difficult optimization problem to solve in order to learn a model from a single giant batch of data, and it can get stuck at bad solutions," says Trevor Campbell, a graduate student in aeronautics and astronautics at MIT, who wrote the new paper with his advisor, Jonathan How, the Richard Cockburn Maclaurin Professor of Aeronautics and Astronautics. "If smaller chunks of data are first processed by individual robots and then combined, the final model is less likely to get stuck at a bad solution."

Campbell says that the work was motivated by questions about robot collaboration. But it could also have implications for big data, since it would allow distributed servers to combine the results of their data analyses without aggregating the data at a central location.

"This procedure is completely robust to pretty much any network you can think of," Campbell says. "It's very much a flexible learning algorithm for decentralized networks."

link.

Robopocalypse is Already Here For Financial Traders

Recently, the global financial market experienced a series of computer glitches that abruptly brought operations to a halt. One reason for these "flash freezes" may be the sudden emergence of mobs of ultrafast robots, which trade on the global markets and operate at speeds beyond human capability, thus overwhelming the system. The appearance of this "ultrafast machine ecology" is documented in a new study published on September 11 in Nature Scientific Reports.

The findings suggest that for time scales less than one second, the financial world makes a sudden transition into a cyber jungle inhabited by packs of aggressive trading algorithms. "These algorithms can operate so fast that humans are unable to participate in real time, and instead, an ultrafast ecology of robots rises up to take control," explains Neil Johnson, professor of physics in the College of Arts and Sciences at the University of Miami (UM), and corresponding author of the study.

"Our findings show that, in this new world of ultrafast robot algorithms, the behavior of the market undergoes a fundamental and abrupt transition to another world where conventional market theories no longer apply," Johnson says.

Society's push for faster systems that outpace competitors has led to the development of algorithms capable of operating faster than the response time for humans. For instance, the quickest a person can react to potential danger is approximately one second. Even a chess grandmaster takes around 650 milliseconds to realize that he is in trouble – yet microchips for trading can operate in a fraction of a millisecond (1 millisecond is 0.001 second).

In the study, the researchers assembled and analyzed a high-throughput millisecond-resolution price stream of multiple stocks and exchanges. From January, 2006, through February, 2011, they found 18,520 extreme events lasting less than 1.5 seconds, including both crashes and spikes.

The team realized that as the duration of these ultrafast extreme events fell below human response times, the number of crashes and spikes increased dramatically. They created a model to understand the behavior and concluded that the events were the product of ultrafast computer trading and not attributable to other factors, such as regulations or mistaken trades. Johnson, who is head of the inter-disciplinary research group on complexity at UM, compares the situation to an ecological environment.

"As long as you have the normal combination of prey and predators, everything is in balance, but if you introduce predators that are too fast, they create extreme events," Johnson says. "What we see with the new ultrafast computer algorithms is predatory trading. In this case, the predator acts before the prey even knows it's there."

Johnson explains that in order to regulate these ultrafast computer algorithms, we need to understand their collective behavior. This is a daunting task, but is made easier by the fact that the algorithms that operate below human response times are relatively simple, because simplicity allows faster processing.

"There are relatively few things that an ultrafast algorithm will do," Johnson says. "This means that they are more likely to start adopting the same behavior, and hence form a cyber crowd or cyber mob which attacks a certain part of the market. This is what gives rise to the extreme events that we observe," he says. "Our math model is able to capture this collective behavior by modeling how these cyber mobs behave".

link.

Sooo...computers run in herds too.  snirk.