Advanced control for soft pneumatic robots

Abstract

Men's long-cherished dream of a human-robot society has never been so close with the recent rapid development of soft robotics. Composed of soft materials or deformable structures, soft robotics could easily achieve safe human-robot interactions by their intrinsic compliance. Soft pneumatic robotics are the ones that use pneumatic as actuation methods and have been widely studied for their easy fabrication and excellent adaptation ability.

As soft pneumatic robotics are evolving fast, emerging realistic control problems are bottlenecking their further developments, such as lack of proper sensors, immature control methods, and complex pneumatic control implementations. Although tons of efforts have been devoted to designing suitable sensors, they are mostly still in the preliminary laboratory phase, not ready for wide use. The pneumatic control implementation is either limited by conserved system dynamics and restricted developing freedoms on commercialized platforms, or struggling with the inadequate performance and developing burden on self-built platforms.

This thesis aims to tackle these two problems to promote the realistic deployment of soft pneumatic robots, by first proposing three compliance-oriented approaches which take the angle from deep utilization of the compliance of soft pneumatic robotics to simplify the control problems, and second introducing an advanced pneumatic control platform which possesses high performance and easy usage.

The proposed control methodology consists of three aspects, including adjusting the compliance for interaction behavioral shaping, averaging compliance effect for in-hand manipulation, and analyzing compliance information for external loads sensation. For the first aspect, compliance adjustment is exploited to shape the behavior of interaction on two soft pneumatic joints respectively using a novel mass-based position and stiffness control method. Secondly, the averaged compliance effect is exploited to realize complex in-hand manipulation tasks on an anthropomorphic soft pneumatic hand with simple control strategies. Lastly, the compliant behavior of an extensible soft pneumatic arm is studied to approximate external loads by extracting information from the compliant behaviors. These aspects demonstrate that deep investigation and full utilization of the compliance of soft pneumatic robots could help to gain advanced control performance with less demanding control requirements.

Apart from the compliance-oriented control methodologies, an advanced pneumatic control platform named "PneuDrive" is developed. It has the features of high performance for implementing advanced algorithms, large extension ability for various soft robotic systems, real-time communication capability and user-friendly interfaces for easy programming. This platform is open sourced online.

With the proposal of compliance-oriented control methods and the open source of advanced pneumatic control platform "PneuDrive", this thesis hopes to deliver some inspiring thoughts and useful utilities to accelerate the pace of soft pneumatic robotics entering into human life.