Sometimes the quickest path is not a straight line

Mar 08, 2012 by David L. Chandler
Graphic: Christine Daniloff

Sometimes the fastest pathway from point A to point B is not a straight line: for example, if you’re underwater and contending with strong and shifting currents. But figuring out the best route in such settings is a monumentally complex problem — especially if you’re trying to do it not just for one underwater vehicle, but for a swarm of them moving all at once toward separate destinations.

But that’s just what a team of engineers at MIT has figured out how to do, in research results to be presented in May at the annual IEEE International Conference on Robotics and Automation. The team, led by Pierre Lermusiaux, the Doherty Associate Professor in Ocean Utilization, developed a mathematical procedure that can optimize path planning for automated underwater vehicles (AUVs), even in regions with complex shorelines and strong shifting currents. The system can provide paths optimized either for the shortest travel time or for the minimum use of energy, or to maximize the collection of data that is considered most important.

This video is not supported by your browser at this time.
Video: Melanie Gonick

Collections of propelled AUVs and gliding AUVs (also called gliders) are now often used for mapping and oceanographic research, for military reconnaissance and harbor protection, or for deep-sea oil-well maintenance and emergency response. So far, fleets of up to 20 such AUVs have been deployed, but in the coming years far larger fleets could come into service, Lermusiaux says, making the computational task of planning optimal paths much more complex.

He adds that earlier attempts to find optimal paths for underwater vehicles were either imprecise, unable to cope with changing currents and complex topography, or required so much computational power that they couldn’t be applied to real-time control of swarms of robotic vehicles.

While researchers have studied such systems for many years, “what was missing were the methodology and algorithm,” he says — the mathematics allowing a computer to solve such path-planning riddles rigorously but quickly enough to be useful in real-world deployments. “Because ocean environments are so complex,” he says, “what was missing was the integration of ocean prediction, ocean estimation, control and optimization” for planning paths for multiple vehicles in a constantly changing situation. That’s what MIT’s Multidisciplinary Simulation, Estimation, and Assimilation Systems (MSEAS) group, led by Lermusiaux, has now developed.

The team’s simulations have successfully tested the new algorithms in models of very complex environments — including an area of the Philippines amid thousands of islands with convoluted shorelines, shallows and multiple shifting currents. They simulated a virtual fleet of 1,000 AUVs, deployed from one or more ships and seeking different targets. Adding to the complication, the system they devised can even account for “forbidden” zones that the craft must avoid and fixed obstacles that affect both the underwater craft and the flow of the currents, and even moving obstacles, such as passing ships.

Taking advantage of the “free ride” offered by the currents, the craft often follow startlingly indirect pathways, meandering around in loops and whorls that sometimes resemble a random walk. That’s because it can be much quicker to drift with a current and then double back than to try to cut straight across, fighting the flow the whole time. In other cases, the AUV may find a quicker or more energy-efficient path by rising over, or diving under, jets, currents, eddies or other ocean features. Uncertainties in ocean predictions — and how they affect the optimal paths — can also be accounted for.

In addition to finding paths that are quickest or most efficient, the system allows swarms of data-collection vehicles to collect the most useful data in the fastest time, Lermusiaux says. These data-optimizing approaches could be useful for monitoring fisheries or for biological or environmental studies — such as a new National Science Foundation effort to characterize the New England Shelf Break, an area important to the region’s fisheries as well as for climate research.

While the methodology and algorithms were developed for an underwater environment, Lermusiaux explains that similar computational systems could be used to guide automated vehicles through any kind of obstacles and flows — such as aerial vehicles coping with winds and mountains. Such systems could even potentially help miniature medical robots navigate through the circulatory system, he says.

The algorithm allows for real-time control and adjustments — such as to track a plume of pollution to its source, or to determine how it is spreading. The system can also incorporate obstacle-avoidance functions to protect the AUVs.

The team included mechanical engineering graduate students Tapovan Lolla and Mattheus Ueckermann SM ’09, Konuralp Yigit SM ’11, and research scientists Patrick Haley and Wayne Leslie. The work was funded by the Office of Naval Research and by the MIT Sea Grant College Program.

Glen Gawarkiewicz, a senior scientist at Woods Hole Oceanographic Institution who was not involved in thus research, says, “This work is significant. It brings rigor to the difficult problem of designing sampling patterns for autonomous vehicles. As the capabilities and the numbers of autonomous vehicles increase, this methodology will be an important tool in oceanography and other fields.”

Explore further: Computer scientists can predict the price of Bitcoin

Related Stories

Predicting motions and sounds of the ocean

Dec 01, 2010

Ocean variability -- the perpetual changing of currents, temperatures, salinity and the contours of the seafloor -- alters the way sound travels through the water. A new analysis of how this variability affects ...

Smarter robot arms (w/ video)

Sep 22, 2011

(PhysOrg.com) -- A combination of two algorithms developed at MIT allows autonomous robots to execute tasks much more efficiently — and move more predictably.

AUVs: From idea to implementation

Nov 07, 2011

Since the 1970s, when early autonomous underwater vehicles (AUVs) were developed at MIT, Institute scientists have tackled various barriers to robots that can travel autonomously in the deep ocean. This fo ...

Speeding swarms of sensor robots

May 03, 2011

Concerns about the spread of radiation from damaged Japanese nuclear reactors — even as scientists are still trying to assess the consequences of the year-old Deepwater Horizon oil spill — have provided ...

The deep sea, from a robot's perspective

Nov 02, 2010

Robots do not have to breathe. For this reason they can dive longer than any human. Equipped with the necessary sensor technology they inspect docks or venture down to the ocean fl oor to search for raw materials. ...

Recommended for you

How to find a submarine

1 hour ago

Das Boot, The Hunt for Red October, The Bedford Incident, We Dive At Dawn: films based on submariners' experience reflect the tense and unusual nature of undersea warfare – where it is often not how well ...

Google offers peek into Bhutan with Street View launch

2 hours ago

Google provided a sneak peek into Bhutan Thursday by unveiling a Street View project for the remote Himalayan kingdom, featuring panoramic views of its majestic mountains, monasteries and crystal-clear rivers.

Nokia turnaround since handset unit sale continues

5 hours ago

Nokia appears to have turned around its fortunes after the sale of its ailing cellphone unit to Microsoft, reporting a third-quarter net profit of 747 million euros ($950 million), from a loss of 91 million euros a year earlier. ...

User comments : 0