10 March 2024
The compliant nature of soft robots is appealing to a wide range of applications in high-impact areas such as medicine, the food industry, and human-robot interaction. However, this compliant property also poses several control challenges, e.g., how to deal with infinite degrees of freedom and highly nonlinear behaviours.
Therefore, the design of controllers for these systems can be significantly more complex than the strategies used for rigid robots, representing one of the current bottlenecks in this area. Several approximation approaches are adopted to deal with this challenge and can be classified into model-based and learning-based strategies.
Model-based control approaches are suitable for exploiting physical properties and prior system knowledge. This often yields controllers that are more energy-efficient than learning-based controllers. Nonetheless, these approaches are sensitive to model mismatches, and they are often analytically complex due to the highly nonlinear nature of soft robots. On the other hand, learning-based control approaches are robust concerning model mismatches, disturbances, and sometimes changing environments. Moreover, the analytical complexity is significantly reduced as most learning strategies are model-free.
In this work, EMERGE partners from Delft University of Technology and collaborators propose a hybrid controller for a pneumatic-actuated soft robot. To this end, a model-based feedforward controller is designed and combined with a correction torque calculated via Gaussian process regression. Then, the proposed model-based and hybrid controllers are experimentally validated, and a detailed comparison between controllers is presented. Notably, the experimental results highlight the potential benefits of adding a learning approach to a model-based controller to enhance the closed-loop performance while reducing the computational load exhibited by purely learning strategies.
Read the paper in the link below.
More Information
