Review of machine learning methods in soft robotics.

Area of Science:

• Robotics
• Artificial Intelligence
• Materials Science

Background:

• Soft robots offer unique flexibility and adaptability but face complex modeling, calibration, and control challenges due to material non-linearity and hysteresis.
• Traditional rigid robot control methods are often inadequate for the inherent complexities of soft robotic systems.
• Machine learning (ML) has emerged as a promising approach to overcome these limitations.

Purpose of the Study:

• To review and categorize existing machine learning techniques applied to soft robotics.
• To analyze the implementation trends of ML in various soft robotic applications, including sensors, actuators, and wearable robots.
• To identify current research limitations and summarize ML methods for soft robots.

Main Methods:

• Systematic review and categorization of machine learning approaches in soft robotics literature.
• Analysis of ML techniques based on their application in soft sensors, soft actuators, and soft wearable robots.
• Trend analysis of ML methods across different soft robot application domains.

Main Results:

• Machine learning techniques are increasingly adopted to manage the non-linear dynamics and hysteresis of soft robots.
• Categorization reveals diverse ML applications, from enhancing soft sensor accuracy to improving soft actuator control and enabling sophisticated soft wearable robots.
• Analysis highlights specific ML trends correlating with different soft robotic application types.

Conclusions:

• Machine learning provides effective solutions for the modeling, calibration, and control challenges inherent in soft robotics.
• The review offers a comprehensive overview of current ML methods, application trends, and identifies areas for future research in soft robotics.
• Further investigation into advanced ML algorithms can unlock new capabilities and broader applications for soft robotic systems.