Underwater Robot Control Using Neural Network Torque Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional position estimation systems for underwater robots fail to accurately reflect changes in sea level height and distance from the sea floor, leading to deteriorated position estimation performance. Additionally, existing methods for measuring the speed of underwater robots are inaccurate and prone to errors.
Innovation Solution
A control method for an underwater robot equipped with a multi-degree-of-freedom robot arm, utilizing an artificial neural network to predict the propulsive force and output torque, enabling precise control of the robot's speed and arm movement. The method also employs dead reckoning based on a Doppler velocity log and real-time Kinematic (RTK) position correction to accurately track the robot's position.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional position estimation systems (LBL, SBL, USL, GIB) are used, then position estimation is available, but position estimation performance deteriorates due to inaccurate reflection of sea level height changes and distance from sea floor
Solution Approach 1:
The patent replaces conventional acoustic-based position estimation systems (LBL, SBL, USL) with a navigation system using Doppler Velocity Log (DVL) and inertial navigation. The DVL measures speed relative to the sea floor by detecting Doppler shift of acoustic signals, while the inertial navigation system integrates speed and acceleration data to calculate position, eliminating dependence on acoustic transponders and sea level reference points.
Solution Approach 2:
The patent changes the measurement parameters from acoustic range-based estimation to velocity-based integration. The DVL provides accurate speed measurement by detecting Doppler frequency shift, and the inertial navigation system integrates these velocity measurements over time to determine position, providing continuous and accurate position estimation without relying on sea level height or distance to sea floor parameters.
2Measurement precision
If GPS-based speed measurement is used, then speed measurement is accurate, but GPS signals cannot be recognized in water
Solution Approach 1:
The patent replaces GPS satellite-based speed measurement with a Doppler Velocity Log (DVL) system that measures speed by detecting the Doppler shift of acoustic signals reflected from the sea floor. The DVL provides accurate speed measurement underwater by calculating the change in frequency of acoustic waves, enabling continuous speed monitoring in the underwater environment where GPS is ineffective.
Solution Approach 2:
The patent introduces the Doppler Velocity Log (DVL) as an intermediary device between the underwater robot and the measurement objective (speed). The DVL acts as a mediator that converts acoustic Doppler shift measurements into speed data, bridging the gap between underwater operation and accurate speed measurement without requiring direct GPS signal penetration.
3Extent of automation
If sensorless algorithm with curve fitting is used, then propulsive force can be calculated, but large measurement errors occur due to speed measurement error and propeller characteristics
Solution Approach 1:
The patent replaces the sensorless algorithm's polynomial curve fitting approach with a neural network-based propulsive force estimation system. The neural network is trained on propeller test data to learn the complex nonlinear relationship between rotational speed and propulsive force, providing more accurate predictions that account for actual propeller characteristics and operating conditions without requiring direct force sensors.
Solution Approach 2:
The patent changes the mathematical model from simple polynomial curve fitting to a neural network model with multiple layers and nonlinear activation functions. This transformation allows the system to capture complex propeller dynamics and improve propulsive force measurement accuracy by learning from training data that reflects actual propeller behavior across various operating conditions.
4Reliability
If strong friction seal is used in underwater actuator, then waterproofing is achieved, but precise control of multi-joint robot arm becomes difficult
Solution Approach 1:
The patent replaces traditional friction-based seal mechanisms with magnetic coupling technology. The actuator uses magnetic fields to transmit rotational motion and force across the waterproof barrier without physical contact, eliminating friction and enabling precise control of the robot arm joints while maintaining effective waterproofing. The magnetic coupling allows torque transmission without mechanical seals that would impede motion precision.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The method achieves precise control of the underwater robot's speed and multi-degree-of-freedom robot arm, while accurately tracking the robot's position, even in environments where GPS signals are unavailable, thereby improving the reliability and accuracy of underwater operations.
Implementation Method 1
estimating a position of the underwater robot using dead reckoning based on a Doppler velocity log (DVL) in water
Data Source
AI summary
Proposed is a control method of an underwater robot equipped with a multi-degree-of-freedom robot arm according to an exemplary embodiment, including: a) step 1-1th of obtaining a propulsive force prediction value by predicting a propulsive force of the underwater robot based on an artificial neural network and configuring a sensorless propulsion controller equipped with a propulsion system to control a speed of the underwater robot; and b) step 1-2th of configuring a underwater robot manipulator (URM) controller that obtains a torque prediction value by predicting an output torque of an actuator constituting the multi-degree-of-freedom robot arm provided in the underwater robot based on the artificial neural network.