Bioinspired artificial intelligence improves UUV reliability

As environmental, oceanic, defense and rescue missions become increasingly complex, unmanned underwater vehicles (UUVs) have become indispensable tools. A new study led by Australian and French researchers offers an innovative solution to improve your reliability in unpredictable conditions.

The scientific team used a bioinspired artificial intelligence solution to improve the reliability of the UUV and other adaptive control systems under unpredictable conditions. This innovative approach, using the Bio-Inspired Experiment Replay (BIER) method, was published in the journal IEEE Access.

Unlike conventional methods, the method BEER aims to overcome data inefficiency and performance degradation by exploiting incomplete but valuable recent experiments, says lead author Dr. Thomas Chaffre. He explains: “The results of the study demonstrated that BIER outperformed standard Experiment Replay methods, achieving optimal performance twice as fast as the latter in the domain of putative UUVs..”

Promising tests for the BIER method

To test the effectiveness of the proposed method, the researchers performed simulated scenarios using a UUV simulator based on the robot operating system (ROS) and gradually increasing the complexity of the scenarios. These scenarios varied depending on target speed values ​​and the intensity of current disturbances.

The associate professor ofFlinders University in AI and Robotics, Paulo Santos, lead author of the study, states that the success of the BIER method holds promise for improving adaptability and performance in various fields that require dynamic adaptive control systems.

Real-world application challenges

The capabilities of UUVs in mapping, imaging and sensor control are rapidly improving, including through deep reinforcement learning (DRL), which rapidly advances adaptive control responses to underwater disturbances that UUV can find. The effectiveness of these methods is tested when faced with unforeseen variations in real-world applications.

The complexity of the dynamics of the underwater environment limits the observability of UUV maneuvering tasks, making it difficult for existing DRL methods to perform optimally. The introduction of BEER marks a significant step in improving the effectiveness of the deep reinforcement learning method in general.

Focus on UUVs

Unmanned underwater vehicles (UUVs), also known as underwater drones, are reshaping the way we explore and interact with the ocean depths. These submersible vehicles, capable of operating without human presence, are divided into two main categories: remote-controlled underwater vehicles (ROUV) and the autonomous underwater vehicles (AUV). ROUVs are piloted remotely by a human operator, while AUVs are fully automated, operating without direct human intervention.

These UUVs are specially designed to withstand the harshest ocean conditions and can operate for long periods at great depths. This opens up new possibilities for researchers and scientists, allowing them to access areas previously inaccessible with traditional manned vehicles.

Furthermore, their silent operation and the absence of polluting emissions make them ideal tools for studying marine life and habitats, without disturbing or damaging the ecosystem.

Synthetic

The BIER method, thanks to its ability to effectively navigate uncertain and dynamic environments, represents a promising advancement in the field of adaptive control systems. The researchers conclude that this approach could improve the reliability of UUVs and other adaptive control systems under unpredictable conditions, paving the way for new possibilities in various fields that require dynamic and adaptive control systems.

For better understanding

What is the BIER method?

The BIER (Biologically Inspired Experience Replay) method is a solution artificial intelligence bio-inspired project that aims to improve the reliability of unmanned underwater vehicles (UUV) and other adaptive control systems under unpredictable conditions.

The method BEER Overcomes data inefficiency and performance degradation by leveraging incomplete but valuable recent experiences. It uses two memory buffers, one focused on recent state-action pairs and the other emphasizing positive rewards.

What are the advantages of the BIER method over traditional methods?

The BIER method demonstrated optimal performance twice as fast as traditional methods in the UUV field. It showed exceptional adaptability and efficiency, demonstrating its ability to stabilize the UUV in varied and difficult conditions.

The researchers conducted simulated scenarios using a UUV simulator based on the robot’s operating system (ROS) and gradually increasing the complexity of the scenarios. These scenarios varied depending on target speed values ​​and the intensity of current disturbances.

What are the potential applications of the BIER method?

The BIER method could improve adaptability and performance in various fields that require dynamic adaptive control systems. It could also improve the reliability of UUVs and other adaptive control systems in unpredictable conditions.

Key Lessons

Teachings
The BIER method is a bioinspired artificial intelligence solution
BIER overcomes data inefficiency and performance degradation
BIER demonstrated optimal performance twice as fast as traditional methods
BIER was tested using a UUV simulator based on the Robot Operating System (ROS)
BIER could improve adaptability and performance in various fields that require dynamic adaptive control systems
BIER could also improve the reliability of UUVs and other adaptive control systems in unpredictable conditions

References

Flinders University, Dr. Thomas Chaffre, Professor Paulo Santos, IEEE Access

The BIER (Biologically Inspired Experience Replay) method was published in IEEE Access Journal.

(Writing)

Leave a Comment