INESC TEC
INESC TEC & Faculty of Engineering of the University of Porto
A very significant part of the deep sea is still unmapped, particularly in areas that are difficult to access. Earlier this year, new previously unknown “underwater mountains” were discovered in Chilean waters, teeming with deep-sea life . Exploring the deep sea reveals new species and ecosystems unknown to science, and this expedition brought to our knowledge about 100 new species. These discoveries highlight the potential for finding many more unique organisms, providing insights into how life thrives in extreme environments, which can lead to advancements in biology and ecology, and even provide clues about life on other planets.
Until now, the resolution of available maps was so coarse that these mountains were “hidden.” Besides the morphology, scientists find dozens of new species on each expedition. These discoveries raise many questions that remain unanswered: how do these species live and interact with each other or with their surrounding environment? How have these species evolved in these extreme environments, in the complete absence of sunlight and under immense pressure?
The deep sea holds geological information crucial for understanding Earth’s history and processes. Discoveries like underwater mountain chains provide significant data on tectonic activity, volcanic processes, and crust formation. Mapping these landscapes helps scientists predict natural disasters, such as earthquakes and tsunamis, and develop strategies to mitigate their impact. Moreover, it is undoubtedly an unknown source of raw materials as the green transition is pushing the limits of inland mines for new and high concentrations of specific minerals.
The unique conditions of the deep sea lead to organisms with remarkable adaptations. Studying these can result in breakthroughs in medicine and technology, such as enzymes functioning under extreme conditions for industrial or pharmaceutical use. The deep sea is also a source of novel compounds and materials, driving innovation and contributing to various fields.
Deep-sea exploration is essential for environmental conservation: understanding the biodiversity and dynamics of these areas helps protect them from human activities like deep-sea mining and fishing. Detailed knowledge allows for better conservation strategies, vital for maintaining ocean health, supporting global biodiversity, and ensuring the well-being of human populations relying on marine resources.
Deep Sea is posing huge challenges due to the high pressure, the harshness of the ocean, the lack of light and the impossibility of humans going there in safe conditions. We must improve our robotics and autonomous systems to go deeper and stay longer underwater.
The present technological developments available for deep-sea deployment, in addition to the lack of reliability and robustness, are limited by the depth requirement, the level of autonomy, the computer power processing, the endurance, and the lack of reliable underwater communications capacity.
It is important to tackle the technological limitations and gaps and establish advanced methodologies for effective exploration and real-time monitoring of environmental impacts.
The deep sea is characterised by immense pressures that increase with depth, requiring specialised equipment capable of withstanding such conditions without being crushed. In addition, near-freezing temperatures also affect electronic equipment, requiring materials and systems designed to operate reliably in cold environments. These harsh conditions increase the likelihood of technical failures; therefore, redundancy, rigorous testing, and robust design are essential to mitigate risks.
By analysing historical data and failure patterns, AI can recommend optimal maintenance schedules. AI tools can analyse data from sensors embedded in equipment to predict potential failures before they occur, reducing the risk of catastrophic failures. For instance, it can monitor pressure sensors, temperature gauges, motion patterns, and power consumption, to identify subtle signs of wear or malfunction.
Sunlight does not penetrate the deep ocean, so video systems require advanced lighting systems and proximity to get meaningful data. AI can process and analyse images and video footage in real-time. Techniques such as machine learning and computer vision can enhance visibility in low-light conditions, identify key features, and even classify objects or species.
Maintaining communication with underwater equipment is difficult due to the limitations of radio waves and reliance on acoustic signals, causing delays and reduced control capabilities. A simple message takes 10 seconds to receive a reply from a device that is 7.5km away using acoustic signals. Typical autonomous underwater vehicles are programmed to follow pre-determined routes, usually defined as a sequence of waypoints, and all possible scenarios, like facing hypothetical obstacles, must be considered. AI tools can make real-time adjustments to the vehicle’s course and operational parameters based on sensor data, ensuring the vehicle can handle unexpected changes in the environment. Furthermore, AI tools can optimise navigation paths for underwater vehicles, avoiding obstacles and adapting to changing conditions, including the state of performance of thrusters or other onboard equipment. This helps in optimising the exploration of uncharted areas with minimal human intervention.
Another technical challenge is to obtain accurate location of underwater devices and provide an accurate estimate of related measurements. Main techniques rely on underwater acoustics, prone to their own sources of errors, e.g., due to the varying conditions of sound propagation. AI tools can estimate a model of the acoustic propagation channel, based on local measurements, and reduce the location errors.
Finally, all existing autonomous robotic systems, relying on batteries, have limited operational durations and spend a significant part of their energy in descending/ascending to/from the areas of interest. To this end, AI tools can optimise power consumption by adjusting operational parameters (e.g., velocity, active sensors, thruster allocation) based on real-time performance data.
Exploration activities can disturb fragile ecosystems such as most deep-water environments; therefore, developing sustainable exploration practices and careful planning of deep-sea operations is crucial to ensure that these activities do not harm unique and delicate deep-sea environments. Reaching most deep-sea areas requires significant logistical planning and resources, including launching and recovering exploration vehicles from ships in remote ocean regions. Hence, deep-sea exploration is expensive, involving high-tech equipment, specialised vessels, and extensive support teams. Funding such missions often requires substantial investment from governments, private companies, or research institutions.
AI tools can optimise mission planning, not only by designing routes for complementary robots but also by selecting ideal mission parameters (e.g., velocities and depth profiles) for specific objectives. Moreover, AI tools can be used to analyse global data, find unexpected correlation patterns, and help identify preferred locations for specific goals. With the new developments of AI tools based on large language models, it is possible to anticipate new capabilities for autonomous vehicles where the operator only sets the main objective or research questions, and the robot is fully “driven” by a virtual operator.
Managing the vast amounts of data collected, including video, images, and sensor readings, is complex and resource-intensive. Effective storage, processing, and analysis are essential for yielding meaningful insights. AI can identify unusual patterns or changes in environmental conditions that could indicate potential issues, such as sudden temperature drops or pressure anomalies, and isolate particular sets of data relevant to a specific study. In other cases, for example in deep-sea observatories, AI can analyse the behaviour of marine life and environmental changes over time, providing valuable insights into ecosystem health and dynamics.
Remote-operated vehicles (ROVs) typically operate with a tether that provides immediate situation awareness (mostly with image and sonar feedback), facilitating underwater manipulation. In deep waters, where the cable is unwanted, autonomous manipulation is rarely used, in a very limited scope. AI-driven robotic arms can perform maintenance and repair tasks on underwater equipment. They can operate with precision and adapt to complex scenarios, potentially fixing issues without needing human intervention.
The challenges of Deep-sea Exploration are huge, as well as the challenges for AI tools with a great potential to revolutionise this field. By leveraging AI, we can enhance efficiency, safety, and the overall understanding of deep-sea environments. Advanced data processing, autonomous operations, predictive modelling, real-time monitoring, and improved decision-making, are some of the possible tools that will improve the capacity to increase the knowledge and explore the Ocean, enhancing the efficiency and safety of deep-sea activities also contributing to the sustainable management of ocean resources.