Abstract
Soft manipulation fundamentally differs from traditional approaches to robotic manipulation and, as a result, much of the established wisdom of traditional manipulation must be questioned. Of course, most of the past results remain valid and form a conceptual basis for soft manipulation, but this basis must be partially revised and extended significantly to fully leverage the power of soft manipulation. In this special issue, we document some of our field’s early attempts towards defining this extended conceptual foundation in an attempt to devise a comprehensive and competent approach to robot manipulation.
The emerging field of soft manipulation transfers important insights from human manipulation into dexterous robotics. Studies of human manipulation revealed substantial differences in the ways humans and robots interact with their environment. For example, during manipulation, humans often leverage contact, not just with the manipulandum but also with the environment. This manifests itself in sliding objects across surfaces, pushing them against others, constraining them in corners, etc. The deliberate exploitation of these contact dynamics permits the delegation of aspects of perception and control to the interactions between hand and environment. For example, the height of a table does not need to be determined in absolute terms, when the hand can simply be lowered onto its surface until contact is achieved. This effectively outsources aspects of perception and control to the physical world, reducing the requirements on explicit perception and control, as we only have to perceive the rough location of the table and the contact forces acting on the robotic hand. This stands in stark contrast with traditional strategies in robotic manipulation, which regard the environment as an obstacle. They avoid environmental contact to prevent damage to the robot and the environment.
Design of soft mechanisms and hands
The transfer of human-like strategies to robot manipulation poses fundamental challenges. The contact-rich interactions of soft manipulation require hands and grippers with inherent compliance, to facilitate safe physical interactions and to leverage these interactions toward achieving robustness and dexterity. The clever design of soft hands is required to increase manipulation capabilities and at the same time reduce the requirements on perception and control. However, little is known about what constitutes a “clever” design in the context of a particular manipulation task.
Six papers in this special issue investigate the design of soft mechanisms, exploring different techniques, materials, and mechanisms. The paper “Design, implementation, and control of a deformable manipulator robot based on a compliant spine” by Bieze et al. explores a design principle for soft manipulators that combines tendon-based actuation with flexible spines. A closely related basic structure is the basis for the paper “Geometric constraint based modeling and analysis of a novel continuum robot with SMA initiated variable stiffness” by Yang et al., in which the compliance of the mechanism can be varied, a property essential in human hands. Still relying on the same spine-based design principle, Hussain et al. in their paper “Design and prototyping soft-rigid tendon driven modular grippers using interpenetrating phase composites (IPCs) materials” examine how varying the compliance through multi-material composition of structural components can lead to advantageous mechanism properties.
Pneumatic actuators represent an important alternative to spine-based designs of soft mechanisms. In “Multi-segment soft robotic fingers enable robust precision grasping,” Teeple et al. study how combining multiple pneumatic actuators can enable precision grasping. The fifth paper on the design of soft mechanisms for manipulation, “Design and characterization of a hybrid soft gripper with active palm pose control” by Subramaniam et al., investigates how the capabilities of hands with pneumatic, vacuum-actuated fingers can be extended through the use of an actuated palm.
In the final paper in this group, “Modeling and analysis of soft robotic fingers using the fin ray effect,” Shan et al. develop a model for “programming” desirable behavior into a mechanism: the fin ray effect causes a finger to bend around the object that it is touching, rather than away from it.
The diversity of design approaches represented in these six papers illustrates the potential for innovation in gripper design. This innovation promises to lead to grasping and manipulation systems with extended dexterity and increased robustness.
Planning and control
The advent of soft mechanisms fundamentally challenges existing approaches to planning and control. Previously, mechanisms were frequently designed to execute motions and actuation commands precisely. The inherent compliance of soft mechanisms makes this very difficult, if not ultimately impossible. Instead of modeling the motion of the mechanism, it now becomes important to find adequate models for the interaction between the mechanism and its environment. Based on these models, new approaches to control and planning are required. These approaches should exploit the advantages offered by soft mechanisms.
Three papers in this special issue address the new challenges in planning and control that arise with the use of softness. Pozzi et al., in their paper “Hand closure model for planning top-grasps with soft robotic hands,” propose a characterization of successful grasps that requires simple perception and control but achieves robust grasping performance by leveraging the advantages of inherent compliance of the hand. The paper “Robust navigation of a soft growing robot by exploiting contact with the environment” by Greer et al., examines the problem of motion planning for a soft robot that can be deflected and deformed by features in the environment. In “Model-free vision-based shaping of deformable plastic materials,” Cherubini et al. explore novel planning in the context of soft, plastic environments with non-prehensile actions.
As the earlier papers did for soft mechanisms, these three papers explore innovative approaches demonstrating novel opportunities that result from the consideration of softness and compliant contact. It is conceivable, maybe even likely, that the algorithms for adequately leveraging the advantages of softness in manipulation will have to differ rather fundamentally from existing approaches to planning and control.
Perception
The new designs for mechanisms, the inherent compliance, and the drastic change in algorithms for planning and control naturally lead to different requirements and opportunities for perception. Soft mechanisms have a very large number of passive degrees of freedom, making it often impossible to determine the exact state of the mechanism from built-in sensors, such as encoders. This poses novel challenges to perception. Hang et al. in their article “Hand–object configuration estimation using particle filters for dexterous in-hand manipulation” present an example of how these challenges might be addressed. Given that the relative state between manipulandum and hand is important for successful dexterity but can often not be supported by encoders embedded into the hand, this article proposes a method for estimating and tracking the hand/object configuration in service of manipulation planning.
Models of underlying physics
The increased focus on contact-based interactions in soft manipulation brings about the need for novel models that capture the essentials of related physical processes and make them available in a computationally efficient way for planning and control. Ghazaei Ardakani et al. derive and validate a hybrid dynamical system capable of predicting the effects of interaction forces on a sliding object. They present this in the article “Quasi-static analysis of planar sliding using friction patches.”
Applications
Soft manipulation promises to extend the range of robotic technology to a number of new application domains. This is a result of the increased manipulation capabilities, but also of the increased safety in interactions with humans. Intrinsically soft mechanisms absorb the energy of an impact, reducing the danger of injury and thus enabling safe human/robot interaction. The article “On the role of stiffness and synchronization in human–robot handshaking” examines the iconic application of handshaking, enabled by the safety of soft devices. In this article, Mura et al. demonstrate that a variety of factors contribute to the human-like feel in a handshake between humans and robots.
The potential and the challenges of soft manipulation
Soft manipulation is an emerging field. This special issue collects the state of the art in a number of exciting domains, all contributing to the establishment of a conceptual, algorithmic, and design foundation for soft manipulation. The papers included in this issue reveal the tremendous potential of soft manipulation for extending robotic manipulation abilities and for applying the resulting manipulation systems in novel application domains. But what also becomes apparent are the many open challenges that need to be addressed before soft manipulation can develop its full potential.
We believe that the questions raised by soft manipulation are exciting and novel. Ever since robotics became a scientific discipline, it has predominantly progressed in relatively separate subfields, such as control, planning, design, perception, reasoning, and machine learning. In contrast, the promise of soft manipulation can only be unlocked through the tight and task-specific integration of all required components. We have seen in this issue how design, planning and control, perception, and underlying modeling techniques need to change to adapt to the characteristics and capabilities of other components of the integrated manipulation system. For example, the advantages of a soft hand, i.e., its inherent compliance and ability to adapt to features in the environment, can only be fully leveraged with novel approaches to control and perception, such as those presented here. Soft manipulation requires an integrated, interdisciplinary view of manipulation, leaving us with hard unanswered questions, but also with the perspective of creating robotic manipulation systems that greatly outperform the systems we are familiar with today. We hope that this special issue can contribute to organizing our community appropriately to face this great challenge.