As we know much less than the oceans of the Earth, we only talk about the moon or Mars. The sea floor extends overflowing with gullies, large shores of the coast, deep tides and cliffs, most of which are not very dangerous or scarce autochthonous vehicles (AUV).
But what would happen if the reward for crossing these places would be worth the risk?
MIT engineers have developed an algorithm that has been developed by AUV to weigh the risks and potential rewards of exploring an unknown region. For example, if an underwater oil identification vehicle approached a compact and rocky trench, the degree of reward for algorithms (probability of oil lubrication near probability near this trench) and a risk level (probability obstacle) if it were to take a trench.
"If our expensive vehicles are conservative, we could say that survival is essential, then we would not find any interest," says Ayton. "But if we understand the commitment between the prize you are collecting and the risk or threat to these dangerous geographies, we can take some risks when it comes to deserving it."
Ayton says the new algorithms can calculate risk commitments in real time when the vehicle decides how to explore next. Brian Williams, a professor at the aeronautics and astronautics laboratory, and his colleagues, are using this Algorithm and others in AUV, with a view to spreading the fleet of many exploiters and intelligent explorers, including offshore oil deposits, investigating the impact of coral reefs on climate change, to explore extreme environments, the Jupiter ice moon, expecting a day's journey by the vehicle.
"If we go to Europe, and if you think it could be a billion dollar bill or breakdown, we would like to justify a European space, after the wind," Ayton says. "But algorithms that are not at risk have never encountered any change in historical observation."
Ayton and Williams, along with Richard Camilli of Woods Hole Oceanographic Institution, will present this new algorithm this week in Honolulu, the Association of Intelligence Artificial Intelligence.
The new group algorithm is the first to enable "risk-limited adjustment sampling". A sample sampling adaptation has been designed, for example, to automatically adapt an AUV path as it measures the vehicle's exploration of a particular region. Sampling missions that take into account the additional risks are commonly followed by a specific and acceptable level of risk. For example, AUVs can be programmed only with paths that do not exceed 5 percent in the graph.
But the researchers found that risk accounting only could limit potential mission rewards.
"Before entering the mission, we want to determine the risk of a certain level of reward," says Ayton. "For example, if we were to take a path to hydrothermal dryers, we would like to take this risk, but if we did not see it, we would be prepared to take less risks."
The group algorithms include information on batimetric data or ocean topography, including barriers to the environment, along with vehicle dynamics and inertional measures, to calculate the level of risk of a proposed road. The Algorithm also takes into account all the previous measurements carried out by AUV to calculate the probability, such as the awarded measurements, such as the proposed path.
If the Risk Ratio Ratio meets a certain value, as the scientists specify, the AUV continues the proposed path, evaluating the algorithm again to assess the risk and access to other paths. the vehicle progresses.
The researchers have tested the algorithm in a simulation of an AUV mission in the East Boston Harbor. They used data collected from the region during the NOAA survey, and an AUV explores at a depth of 15 meters at relatively high temperatures through regions. They have investigated how the algorithm planned the route of the vehicle in three scenarios of acceptable risk.
With the scenario with the lowest acceptable risk, the vehicle should avoid the region that has a great deal of collision, the algorithm will draw a preservative path, keeping the vehicle in a safe region, which has not been highly rewarded in this case at high temperatures. For high-risk situations, the algorithm crossed a vehicle path through a narrow cave and, ultimately, in the award-winning region.
The group also directed the algorithm through 10,000 numerical simulation, creating random environments in each simulation, by the way, planning, and the algorithm can "deactivate the risk against the prize intuitively, and dangerous actions justified by the award."
Last December, Ayton, Williams and others were given two weeks off the coast of Costa Rica, underwater diving, tested various algorithms, among others. In most cases, the proposed algorithm pathways were agreed upon by several ontologists geologists, who were looking for the best anti-oil routes.
Ayton said there was a particular moment in the risk limit algorithm that was particularly useful. An AUV was built for a preventive fall or displacement, where the vehicle could not do much harm.
"The algorithm found a method to kill it quickly, as it is worthwhile," says Ayton. "We received a source because it did not help the oil purification plants, but it helped us improve the understanding of the environment."
"What was really interesting was how to start the machine algorithms" after discovering several dives, "and we started to choose the sites that we did not choose geologists," said Lori Summa, geologist and guest researcher. Woods Hole Oceanographic Institution, took part in a cruise. "This part of the process is still evolving, but the algorithm was exciting to identify new models of large data and collect information in an efficient and secure way."
In the long-term view, researchers are expected to use algorithms for self-employed vehicles to study Earth's environments.
"If we went to Europe and did not take any risks to take care of an experiment, then the probability of finding life would be very serious," says Ayton. "You have to risk a little more to get more prizes, in general, in truth as well as in life."
The new algorithm improves the energy efficiency of underwater vehicles
Massachusetts Institute of Technology
MIT News (web.mit.edu/newsoffice/) This story is a popular site for MIT News, MIT news, research and innovation.
To calculate the risks calculated by the robot of the engineers program (2019, January 30)
They retrieved on January 30, 2019
This document is subject to copyright. The objective of private research or research for any reason
The part can be reproduced without written permission. Content information is only provided.