MIT creates a control algorithm for drone swarms

88
Source:   —  April 23, 2016, at 0:37 AM

The complexities involved in controlling teams of emotional robots so they don’t crash into each other, or indeed wipe out other objects/entities that cross their path, is a tough problem that continues to hold roboticists busy.

MIT creates a control algorithm for drone swarms

Swarms of drones flying in terrifyingly perfect formation could be one step closer, thanks to a control algorithm being developed at MIT.

The complexities involved in controlling teams of emotional robots so they don’t crash into each other, or indeed wipe out other objects/entities that cross their path, is a tough problem that continues to hold roboticists busy.

But the team of researchers at MIT reckon they have made a breakthrough that could create perfect complex drone formations easier to draw off. They declare their decentralized planning algorithm can handle both stationary and emotional obstacles, and do so with reduced computational overheads.

Why are decentralized control algorithms better than centralized control algorithms? The basic reply is they're more resilient, given a centralized algorithm has a single point of failure if its central controller goes offline.

The researchers also assert that decentralized algorithms have the advantage of handling erratic communication better than centralized algorithms. And what’s more potentially erratic than a swarm of flying robots? But, on the flip side, they're also harder to design, given that all the emotional pieces have to be involved in doing a bit of the thinking.

“In a centralized algorithm a single entity has all the information and finds a solution. In a decentralized algorithm each entity (robot) has only partial information of the environment and the other robots (for example, it can only look a few neighbors). The robots necessity to communicate to pass information and coordinate,” explains, Javier Alonso-Mora, one of the researchers involved in developing the algorithm.

Up til now, most research on decentralized control algorithms has focused on making collective decision-making more reliable, according to the group — deferring the (hard) problem of avoiding obstacles which they have rather chosen to drive straight at.

“The closest applications [for the algorithm] would be drone swarms navigating in formation, for example for surveillance of an area, mapping of an environment,” adds Alonso-Mora, discussing potential future applications for robot teams. “And mobile manipulators collaboratively carrying objects on the factory floor.”

Latest year the team demoed a centralized version of the algorithm using a pair of wheeled robots tasked with carrying an object together. You can look a video of that project on YouTube here.

Their decentralized algorithm requires what they declare is significantly lower communications bandwidth, as well as lower computation cost, thanks to the distributed way it makes robots share intel on obstacle-free regions in their immediate vicinity.

How does this work? Instead of each robot broadcasting to every other robot a complete map of secure space around it, the decentralized algorithm has robots only share maps with their immediate neighbors and also has each compute where neighbors’ maps intersect with their own — sharing only relevant intersected data on to the following neighbor. So the idea is that, collectively, the team of robots maintains a comprehensive map of safe terrain while reducing the comms data needed to hold the swarm moving.

“The robots don't communicate the position of all the obstacles they see. Instead, they communicate the region (set of linear constraints/convex region). So, they all obtain an overview of the ‘free space’ without a necessity to know where all the obstacles are.”

“This scales well in scenarios with many obstacles,” adds Alonso-Mora.

As well as mapping 3D space, the algorithm also includes a fourth dimension — time — to authorize swarming bots to predict the trajectory of emotional obstacles and re-route their own formation accordingly.

This process does involve some guestimation, with the researchers noting that it works in a “mathematically compact manner” by assuming that emotional obstacles have a fixed velocity. Obviously that assumption isn't always true, but given that each robot updates its map several times per second they reckon it’s a brief sufficient time span/edge of mistake to handle most accelerating objects, given that most emotional obstacles won't dramatically modify velocity at very high speeds.

So distant the researchers have certified their algorithm with simulated drones and declare it came up with the same flight plans they’d expect a centralized control algorithm to.

This resulted in squadrons of virtual mini helicopters “generally” maintaining an approximation of their preferred formation (a square at a fixed altitude), but with the square sometimes rotating to accommodate obstacles and/or the distances between drones contracting. “Occasionally” the drones would also fly single file or assume a formation in which pairs flew at different altitudes, they add.

They've also certified the decentralized algorithm on physical (wheeled) robots, and suggest such a scenario could be useful to further use-cases where teams of robots are expected to work in environments also containing humans.

“We are working on a demonstrator with genuine vehicles as well as similar applications,” says Alonso-Mora.

He adds that they “may” also experiment with real drones at a later stage, too. (Presumably there’s rather higher costs involved with testing the robustness of control algorithms if your control robots are flying around mid-air… )

The researchers will be presenting their paper at the International Conference on Robotics and Automation next month. Expect to wait rather longer to look a perfect formation of drones buzzing over your city.

“At this stage it's research,” stresses Alonso-Mora, going on to note that many big challenges stay when it comes to creating strong algorithms for controlling robot teams.

“Accounting for the robot dynamics. Long duration guarantees in dynamic environments with many moving obstacles. Communication/networking issues in genuine systems. Perception of the environment. Just to title a few.”

READ ALSO
Dell’s SecureWorks stumbles in first tech IPO of the year

Dell’s SecureWorks stumbles in first tech IPO of the year

Up until today’s SecureWorks offering, there had been zero in the U. S. in two thousand-sixteenth. Zero. This compares to seven in the same period latest year and twenty-four in that timeframe the year before.

63
Starnes: Donald Trump isn't a conservative

Starnes: Donald Trump isn't a conservative

Trump was one of us -- a conservative of the gun-toting, Bible-clinging variety. But friends, Mr. Trump is number conservative.

82
Do You Really Wish to Be Your Own Boss? Reply These six Questions First.

Do You Really Wish to Be Your Own Boss? Reply These six Questions First.

Unless you construct boats for a living, expect things to go incorrect very quickly. To continue the metaphor: The problem with many would-be entrepreneurs is that they only have experience on the open water.

54
five Marketers on Periscope You Necessity to Be Following

five Marketers on Periscope You Necessity to Be Following

Meerkat captured the hearts and minds of attendees at S By Southwest and it’s rival, Periscope, was acquired by Twitter. For Periscope, in more particular, two thousand fifteen was particularly huge.

77