The oil and gas industry is undergoing continuous transformation, driven by the need to enhance operational efficiency, safety, and sustainability in an increasingly complex global landscape. Robotics is emerging as a cornerstone of this evolution, offering advanced solutions to some of the industry's most pressing challenges.

By automating inspection and maintenance tasks, particularly in hazardous or hard-to-access environments, robotics is revolutionizing traditional methods and setting new benchmarks for performance and reliability. A prime example of this transformation is the deployment of robots for the inspection and maintenance of subsea pipelines. These pipelines, often spanning vast and remote subsea terrains, are critical to the transportation of oil and gas but pose significant logistical and safety challenges for human workers.

Robotic systems, equipped with advanced sensors and AI-powered navigation, can operate continuously in these environments, performing inspections with unmatched precision. They can detect potential issues such as corrosion, leaks, or structural damage early, allowing operators to address them proactively before they escalate into costly or dangerous problems. The benefits extend beyond operational efficiency. By minimizing the need for human divers to perform inspections in potentially life-threatening conditions, robotics dramatically enhances safety standards. Additionally, robotic systems reduce the downtime associated with traditional maintenance processes, optimizing productivity while lowering operational costs.

One notable innovation is the Eelume underwater robot, developed by researchers at the Norwegian University of Science and Technology in collaboration with Equinor ASA and Kongsberg Maritime AS. Designed for subsea inspection and light maintenance, Eelume is envisioned to operate continuously in underwater environments, eliminating the need for frequent recovery to the surface.

With its autonomous docking and charging capabilities, the robot can “live” permanently underwater, further reducing operational disruptions and enabling real-time response to maintenance needs. Another cutting-edge example is the Scout 137 Drone System, designed specifically for ballast tank inspections. This system allows for efficient and detailed inspection of confined spaces, replacing traditional manual methods that are labor-intensive and pose safety risks to personnel. By leveraging these robotic solutions, the oil and gas industry is advancing toward safer, more sustainable, and highly automated operations, paving the way for a future where human intervention in hazardous tasks is minimized.

Despite these advancements, the integration of cutting-edge robotic systems in the oil and gas industry continues to face significant challenges. One of the primary barriers is the complexity of programming and operating these systems. Current solutions often rely on a mission programming operator interface accessible through a centralized control room. This setup demands that operators possess extensive knowledge of both the plant’s operational parameters and the intricate mission steps required to program and control robots. Such requirements not only necessitate intensive training but also limit the scalability of these technologies across facilities with varying operational conditions.

Furthermore, the complexity is exacerbated by the need for translation layers that interface with robot-specific APIs. These layers act as intermediaries between operator commands and the robotic systems, introducing additional steps and potential points of failure. This results in an indirect and rigid interaction between human operators and robots, where reliance on pre-programmed routines and extensive manual input inhibits adaptability and responsiveness. For operations requiring real-time adjustments or dynamic decision-making, these limitations reduce the overall effectiveness of robotic solutions.

With the advent of large language models (LLMs) integrated into software stacks, these barriers are potentially diminishing. LLMs represent a transformative leap forward in this domain. By incorporating natural language processing capabilities into robotic software stacks, mission programming can become as intuitive as giving verbal instructions.

This shift simplifies the operator’s role, reducing the need for extensive training while lowering the threshold for robotics deployment across the industry. For example, a proposed system stack includes an LLM that interprets voice commands, converting them into executable mission steps for robots. This innovation can be aligned with regulatory and practical requirements, such as ensuring that voice commands are processed in the operator's native language for safety-critical applications.

The ability of robots to interact directly with operators through voice commands marks a significant evolution in human-machine interaction. It not only streamlines fleet management and mission programming but also addresses the challenges of change management, which can often impede technological adoption. In sectors like oil and gas, where safety and operational efficiency are paramount, the introduction of user-friendly robotics solutions supported by LLMs is a game-changer. Moreover, the use of LLMs in robotic systems goes beyond operational convenience. It holds the potential to enhance overall system adaptability, allowing robots to handle more dynamic and complex environments with less manual oversight. By bridging the gap between sophisticated technology and practical usability, LLMs enable a broader adoption of autonomous systems in inspection and maintenance activities, laying the groundwork for a safer, more efficient industry future.

The opportunities presented by this integration are vast. Voice-controlled robotics can significantly lower barriers to entry, making the technology accessible to a broader range of industries and applications. The reduction in complexity enables operators to manage fleets of robots more effectively, improving scalability. In the oil and gas industry, these advancements not only optimize inspection and maintenance processes but also enhance safety by minimizing human exposure to dangerous environments. The potential for LLMs to support adaptive learning further broadens the scope of their application, allowing robots to improve their performance and adapt to new tasks over time.

----
Agus Hasan is a professor in cyber-physical systems at the Department of ICT and Natural Sciences, Norwegian University of Science and Technology (NTNU)

The views expressed in this article are those of the author.