Deep Neural Data-Driven Koopman Fractional Control for Worm Robots

Worms are a fascinating model for bio-inspired robotics as they exhibit remarkable adaptability, agility and efficiency in navigating environments. Studying their biomechanics and behaviors provides insights into locomotion principles which can translate into novel robot systems. Bio-inspired worm robots can operate in environments that are otherwise inaccessible or hazardous to humans, however, traditional control methods often struggle with the nonlinear motion dynamics, making precise control difficult.

 

Researchers at Arizona State University have developed an innovative approach for enhancing the locomotion control of a worm robot by integrating neural networks with the Koopman operator framework and fractional order control techniques. Leveraging fractional sliding mode control, robustness against disturbances is improved and chattering is reduced. Inspired by biological wave-like locomotion, the worm robot design is optimized using dynamic parameters for efficient propulsion. The approach facilitates advanced control strategies in a latent space, significantly enhancing tracking accuracy and reducing control efforts as demonstrated through simulations. By modeling the nonlinear dynamics of a segmented worm robot and transforming these into a linear framework using the Koopman operator, the method enables efficient control and prediction. Simulation results demonstrate superior tracking performance and reduced error compared to traditional control methods.

 

This technology represents a novel approach to linearize and predict the complex nonlinear dynamics of worm robots.

 

For more information about this opportunity, please see