ADAPTIVE MODELING OF UNDERWATER ROBOT FLUID DYNAMICS BASED ON FORCE MEASUREMENT DEVICE

This work describes the development and testing of a data-driven hydrodynamic model for quadruped robots designed for adaptive and intelligent interaction in dynamic environments. To effectively manage and interpret sensor data, we employ Gaussian Process Regression (GPR) to model the underlying uncertainties in fluid-structure interactions, allowing for more precise predictions in complex and varying environments. The probabilistic nature of GPR enables quadruped robots to handle noisy data and provide robust, uncertainty-aware decision-making strategies.

We train and evaluate the model using real-time sensor data, which includes ambient environmental factors and the robot's internal states. A key focus of our study is the robot's adaptive response to different hydrodynamic conditions, such as varying speeds and fluid dynamics. The results demonstrate that the GPR-based model efficiently learns and adapts to these dynamic conditions, leading to accurate force prediction and enhanced autonomous performance in a range of real-world scenarios.