From DRLWiki

Motivation

We are interested in designing "soft" robotic systems that show unprecedented levels of shape deformation, flexibility and robustness. To achieve this goal, we aim at combining chemistry, information theory, and control to create novel materials that embed sensing, computation, and actuation.

Approach

We are currently investigating a series of closed-chain kinematic systems, which have the ability to locomote and deform simply by changing the stiffness of their joints as well as suitable actuators. Unlike classical actuators such as electrical, combustion or steam-engine motors, changes in stiffness have the potential to be directly obtained from chemical reactions (e.g. thermal expansion, liquid-gas phase transitions or "smart'" fluids) inside the material.

We are also experimenting with closed-chain kinematic systems that can locomote by expanding and contracting parts of their body where pressure is potentially generated by a local chemical reaction.

Status

We have designed, simulated and analyzed closed-chain kinematic systems based on hexagons and octagons and developed control schemes that allow such structures to locomote on a 2D plane and collapse into a single line simply by changing the stiffness of their joints. We have also conducted preliminary experiments with novel 1D actuator concepts based on thermal expansion of fluids, electrolysis and phase transitions.

Currently, we are developing a distributed control layer for a novel robotic concept developed by Erik Schoenfeld at iRobot. The platform can locomote by selectively inflating parts of its body using a compressed air source and can potentially be miniaturized by leveraging chemical actuator technologies currently under development by our collaborators at the Whitesides group in Harvard.

iRobot prototype of a novel soft-robotic actuator. A prototype of a distributed control layer is shown in the background. The video shows initial experiments using open-loop and manual control.

Robot with embedded sensor (photocells) and control logic.

Our distributed control scheme relies on the fact that the pose of the robot can unambiguously identified by evaluating only a subset of the sensors (notably those close to where the robot body touches the ground). A chamber is inflated only if it is fully on the ground and if either its left or right neighbor (depending of the direction of motion) does not have contact to the ground.

Simulation of combined closed-chain kinematic systems

2-D Hexagon rolling laterally. (won't play?).

2-D Octagon rolling laterally. (won't play?).

Quadhex rolling laterally when its cross-section is a square. (won't play?).

Quadhex rolling laterally when its cross-section is a hexagon resting on an edge. (won't play?).

Quadhex rolling laterally when its cross-section is a hexagon resting on a vertex. (won't play?).

People (MIT)

• Nikolaus Correll
• Kyle Gilpin
• Daniela Rus

Collaborators

• iRobot Inc., Chris Jones (PI)
• Distributed Robotics Laboratory, MIT, Daniela Rus
• Microrobotics Laboratory, Harvard University, Robert Wood
• The Applied Math Lab, Harvard University, L. Mahadevan
• The Whitesides Research Group, Harvard University, George Whitesides