New Algorithm Enables Marine Robots to Weigh Risks and Probable Rewards of Exploring Unknown Region

The knowledge of humans about the oceans of Earth is far less than that about the surface of Mars or the moon. The sea floor is carved with towering seamounts, expansive canyons, sheer cliffs, and deep trenches, a majority of which are regarded to be inaccessible or very dangerous for autonomous underwater vehicles (AUV) to navigate.

But what if the reward for traversing such regions was worth the risk?

Currently, engineers at MIT have created an algorithm that allows AUVs to weigh the risks and possible rewards of exploring an obscure region. For example, in case a vehicle assigned the job of recognizing underwater oil seeps reached a steep, rocky trench, the algorithm could evaluate the reward level (the probability of the existence of an oil seep close to this trench), as well as the risk level (the probability of colliding with an obstacle), if it were to traverse through the trench.

If we were very conservative with our expensive vehicle, saying its survivability was paramount above all, then we wouldn’t find anything of interest. But if we understand there’s a tradeoff between the reward of what you gather, and the risk or threat of going toward these dangerous geographies, we can take certain risks when it’s worthwhile.

Benjamin Ayton, Graduate Student, MIT’s Department of Aeronautics and Astronautics.

According to Ayton, the new algorithm has the ability to compute tradeoffs of reward versus risk in real time, as a vehicle makes the decision of which place to explore next. He and his team in the lab of Brian Williams, professor of aeronautics and astronautics, are using this algorithm as well as others on AUVs, with the idea of deploying fleets of intelligent, bold robotic explorers for several missions, such as examining the effect of climate change on coral reefs, searching for offshore oil deposits, and investigating extreme environments like Europa, one of Jupiter’s moon that is ice-covered and which the researchers believe vehicles will one day traverse.

If we went to Europa and had a very strong reason to believe that there might be a billion-dollar observation in a cave or crevasse, which would justify sending a spacecraft to Europa, then we would absolutely want to risk going in that cave. But algorithms that don’t consider risk are never going to find that potentially history-changing observation.

Benjamin Ayton, Graduate Student, MIT’s Department of Aeronautics and Astronautics.

Ayton and Williams, together with Richard Camilli of the Woods Hole Oceanographic Institution, will present their innovative algorithm at the Association for the Advancement of Artificial Intelligence conference this week in Honolulu.

A Bold Path

The new algorithm developed by the team is the first to allow “risk-bounded adaptive sampling.” For example, an adaptive sampling mission has been designed to automatically modify an AUV’s path in accordance with new measurements taken by the vehicle as it explores a specified region. A majority of the adaptive sampling missions that take typically take the risk into account do so by identifying paths with a concrete but acceptable risk level. For example, it is possible to program AUVs to only traverse paths with a likelihood of collision that does not exceed 5%.

However, the team identified that considering just the risk alone could acutely restrict the potential rewards of a mission.

“Before we go into a mission, we want to specify the risk we’re willing to take for a certain level of reward,” stated Ayton. “For instance, if a path were to take us to more hydrothermal vents, we would be willing to take this amount of risk, but if we’re not going to see anything, we would be willing to take less risk.”

The algorithm developed by the team considers bathymetric data (i.e. information related to the ocean topography, including any obstacles around) together with the dynamics and inertial measurements of the vehicle to determine the risk level for a particular proposed path. The algorithm also considers all measurements taken by the AUV earlier to determine the probability that such high-reward measurements may occur along the proposed path.

In case the risk-to-reward ratio satisfies a specific value, determined by the researchers in advance, the AUV goes forward with the proposed path and takes more measurements that are fed back into the algorithm to enable it to assess the reward and risk of other paths as the vehicle moves ahead.

The team tested its algorithm through the simulation of an AUV mission east of Boston Harbor. Bathymetric data gathered from the area at the time of an earlier NOAA survey was used and an AUV was simulated to explore at a depth of 15 m through regions at comparatively high temperatures. They observed the ways in which the algorithm outlined the route of the vehicle under three different scenarios of acceptable risk.

In the lowest acceptable risk scenario, that is, when the vehicle must avoid any regions with a very high likelihood of collision, the algorithm charts out a conservative path, holding the vehicle in a safe region that also lacked any high rewards—in this instance, high temperatures. For higher acceptable risk scenarios, the algorithm mapped out bolder paths that made a vehicle to move through a narrow chasm, and eventually to a high-reward region.

The algorithm was run by the researchers through 10,000 numerical simulations, producing in each simulation random environments through which a path can be planned, and it was found that the algorithm “trades off risk against reward intuitively, taking dangerous actions only when justified by the reward.”

A Risky Slope

In December 2019, Ayton, Williams, and colleagues deployed underwater gliders by spending two weeks on a cruise off the coast of Costa Rica. They tested a number of algorithms, including the latest one. In most cases, the path planning of the algorithm was in agreement with those put forward by various onboard geologists who were searching for the best routes to identify oil seeps.

According to Ayton, there was a specific instance when the risk-bounded algorithm was found to be particularly handy. An AUV was traversing up through a precarious slump, or landslide, where the vehicle could not afford to take too many risks.

“The algorithm found a method to get us up the slump quickly, while being the most worthwhile,” stated Ayton. “It took us up a path that, while it didn’t help us discover oil seeps, it did help us refine our understanding of the environment.”

What was really interesting was to watch how the machine algorithms began to ‘learn’ after the findings of several dives, and began to choose sites that we geologists might not have chosen initially. This part of the process is still evolving, but it was exciting to watch the algorithms begin to identify the new patterns from large amounts of data, and couple that information to an efficient, ‘safe’ search strategy.

Lori Summa, Geologist and Guest Investigator, Woods Hole Oceanographic Institution.

Lori Summa was also part of the crew that took part in the cruise.

The long-term vision of the researchers is to employ such algorithms to assist autonomous vehicles in exploring environments beyond Earth.

“If we went to Europa and weren’t willing to take any risks in order to preserve a probe, then the probability of finding life would be very, very low,” stated Ayton. “You have to risk a little to get more reward, which is generally true in life as well.”

This study was supported, in part, by Exxon Mobile, as part of the MIT Energy Initiative, and by NASA.