CAREER: Architectures of Responsiveness: Coordinating Primitives for Soft Robot Locomotion in Confined Spaces

This Faculty Early Career Development Program (CAREER) project will create a vocabulary for robot motion in confined spaces, enabling new types of healthcare and infrastructure inspection systems. The soft bodies of animals such as earthworms allow them to travel through narrow spaces and to navigate complex arrangements of obstacles. Soft robots have the potential to emulate those capabilities through the combination of passive mechanical compliance and active response.

Like their biological analogs, these robots continually make and break contact with their surroundings as they move. The high number of degrees of freedom of structurally soft systems, combined with the dynamically changing contact points, makes centralized control of this type of robot motion an intractable problem. In contrast, studies of animals such as earthworms have shown that complex whole-body maneuvers may be generated as the collective action of simple movements of individual body segments.

Inspired by this example, this project will develop composable architectures of dynamically rich, but easily computed local response primitives that will combine to produce desired motions. These architectures will provide the basis for an on-going project to translate the locomotion strategies of soft-bodied animals into new classes of intelligent and responsive robots.

The research program is complemented by an educational plan to provide an alternate to traditional "pipeline" models for students who want to contribute to science -- including students learning from home. The goal is to create in-place connections between scientists, engineers, and students to increase awareness of scientific research, expand access to STEM careers, and foster public trust in communities that would otherwise rarely meet scientists.

The central hypothesis of this project is that building responsive movement primitives around saddle points in state space will enable an especially rich and versatile architecture of composable dynamic response. Using saddles instead of the typical stable attractors enables sensory input to influence trajectories without altering underlying dynamics, leading to more modular control within a larger network.

This will be demonstrated by developing tools to coordinate control primitives in architectures that scale spatially for many degrees of freedom, hierarchically for broad sensor information, and topologically for decision-making. The different scaling approaches will be validated and demonstrated by adding responsiveness to earthworm-like locomotion for confined spaces.

Spatial recruitment of neighboring segments in response to local failures will address the problem of slip. Hierarchical nested responsiveness will resolve competition between passive vs active turning. Topological switchable consensus will enable qualitatively different strategies (e.g. bracing with undulation versus peristaltic gaits) to suit different environments.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.