This EArly-concept Grant for Exploratory Research (EAGER) project will create new sensor-based control algorithms to enable soft robots to use a novel soft optical sensor in order to react in real time to changes in the robot and the environment. Soft robots -- that is, robots made of soft materials such as rubber -- have the potential to revolutionize the way robots are designed, manufactured and deployed. Rapid prototyping techniques can produce new robots in a matter of hours, not months, robot design can be customized to fit both the robot's environment and task, and these robots are inherently safe, enabling close interaction with people. However, there are currently few examples of reliable, long duration use of soft robots in complex environments. Reasons for this lack of robust behavior include the difficulty of modeling the robot's interaction with the environment, for example the way the legs bend when walking on sand vs. concrete, the higher failure rate of soft actuators versus rigid actuators, and relative lack of good soft sensors versus traditional rigid sensors. This project addresses all three of these challenges, because the soft sensors are physically robust and, together with the new control algorithms, greatly reduce the need for accurate models.
The objective of this EAGER project is to collect preliminary data and assess the feasibility of co-designing control and sensor placement for foam-based robots embedded with stretchable optical waveguides which can be used to sense deformations in the soft robot?s body. The outcomes of this project will advance the vision of soft robots that can be predictably and reliably controlled in a variety of environments. The novel soft optical sensors can be embedded throughout the actuator in a plurality of configurations; therefore, the placement can be designed with the control objective in mind. Due to the difficulty in exact modeling of soft actuators, especially those that can be created in different shapes, in order to create reliable and repeatable control for soft robots, sensors must be embedded in the body of the soft robot. The information the sensors provide, together with simplified models of robot motion, should be used to control the robot. This project will produce (i) a characterization of the information that can be obtained from a stretchable optical waveguide embedded in a foam actuator, (ii) physical limits on the quality, spatial resolution, and type of information that can be collected from multiple embedded sensors due to manufacturing constraints and the cross-influence of the sensors on the foam actuator and each other, and (iii) an algorithm that given a fixed multi-legged robot returns the required sensor placements and the feedback controller that will create predictable motions over a specific terrain. The real breakthrough that will enable reliable and robust control of soft robots will require a radical new approach to combining sensing, actuation and control, where the control objective, together with physical limitations, determines sensor placement and feedback control.
