MOSAIC-SV: Real-Time Adaptive Control for Aquatic Robotics
A new system enhances control of aquatic robots by adapting dynamics models in real-time.
Researchers have developed MOSAIC-SV, an innovative system for real-time adaptive dynamics identification and control of aquatic robotic platforms. This system eliminates the need for costly identification trials by utilizing a prior engineering model to identify a control-sufficient dynamics model.
MOSAIC-SV employs a physically admissible unscented Kalman filter to re-estimate hydrodynamic and actuator parameters at each control step during a closed-loop mission. It also incorporates a command-dependent projection to manage corrections between actuator effectiveness and external forces.
In simulations with the CyberShip II plant, MOSAIC-SV demonstrated robust performance, maintaining transit times within 20% of calibrated models even under significant parameter mismatches. Field trials on the Blue Robotics BlueBoat showed a transit speed increase of at least 25% and a reduction in motion prediction error by over 56% compared to traditional models.