Cerebellar-inspired algorithm for adaptive control of nonlinear dielectric elastomer-based artificial muscle

Area of Science:

• Robotics
• Neuroscience
• Materials Science

Background:

• Electroactive polymer actuators are crucial for soft robotics but exhibit challenging nonlinear behaviors like creep.
• Biological control systems effectively manage complex motor tasks, offering potential solutions for artificial actuators.

Purpose of the Study:

• To investigate the efficacy of a cerebellum-inspired control scheme for managing nonlinear dielectric elastomer actuators.
• To explore how biological control principles can enhance the performance of artificial muscle technologies.

Main Methods:

• A control system model was developed, representing the cerebellum as an adaptive filter and the brainstem as an approximate inverse plant model.
• Recurrent connections between the cerebellar and brainstem models allowed for sensory error-based motor command adjustment.
• Semi-linear basis functions and transfer of training mechanisms were implemented, mimicking biological muscle recruitment and reflex adaptation.

Main Results:

• The bio-inspired control scheme achieved accurate tracking of displacement commands within the actuator's nonlinear range.
• Implementing semi-linear functions or transfer of training mechanisms mitigated issues related to actuator creep and reduced the need for increased control output.
• The control system demonstrated robustness in handling the inherent complexities of dielectric elastomer actuators.

Conclusions:

• A cerebellum-based adaptive-inverse control strategy is a viable and effective method for controlling nonlinear electroactive polymer actuators in soft robotics.
• The study highlights the importance of nonlinear basis functions, recruitment principles, and training transfer in biological motor control.
• This bio-inspired approach offers a promising direction for advancing the capabilities and control of artificial muscle technologies.