Our vision is that hard problems in robotics require that we fundamentally change the way robots are made and deeply embed sensing and control into their material. Such materials will require tight integration of sensing, actuation, and computation. Such materials can possibly be mass-produced cheaply and easily, thereby abstracting complex functionality into a simple interface. We call such materials “Robotic Materials“.
In order to motivate and test robotic materials, we are using the Baxter robot by Rethink Robotics, which has been a milestone in making robots simpler by relying on series-elastic actuators. While this approach makes the robot compliant and inexpensive, it does not provide the accuracy and analytical tractability that we are used from conventional robots. We believe that robots should be even simpler and take advantage of rich sensing, pre-processing and control provided by robotic materials. For example, a sensitive skin will help the robot to navigate inside complex objects such as a plant canopy. Soft fingers that can control their curvature and pressure allow for compliant grasps, as well as assessing an object’s properties beyond those that are visible. Fishing-line or McKibben actuators combined with variable stiffness materials, might provide efficient actuation.
On the other end of the spectrum, we are working on algorithms that allow conventional robots with large numbers of degree of freedom to plan faster in dynamic environments and predict the dynamics of complex objects they interact with. This will allow robots to work closely with humans and possibly even manufacture robotic materials. To this end, we are investigating estimation and planning algorithms that allow robots to assemble structures with very high precision – such as space telescope optical benches – using stock materials.
In order to create robotic materials, our lab is equipped with facilities to manufacture composites, electronics, and smart fabrics for wearable devices. While the latter has important applications in leisure, military, and assistive contexts, we believe that cloth has also important applications in robotics. In addition to established manufacturing techniques, cloth-based sensors and actuators can be easily donned and doffed, as well as registered to the robots surfaces. For example, NASA’s Robonaut is already wearing a jersey to protect mechanism and electronics, which is an ideal target for functionalization.
|Composite laboratory||Electronic workstation||Smart fabric workstation|
In addition to providing rich sensing and actuation to robots, robotic materials have a wide range of applications to equip systems, ranging from airplane wings to furniture and tableware, with the ability to sense at high resolution and bandwidth or to change their shape or appearance. Our lab is using two platforms to investigate the distributed algorithms that will drive robotic materials, including self-assembly, in a broader sense. To this end, we have developed a modular wall that can sense touch and change its color and opacity, as well as a swarm of “Droplets” that is powered via its floor and can communicate via infrared. All the systems that we use in the lab are also actively being used in classes, some of which are taught in a dedicated area of the lab.
|The amorphous facade||The Droplets||Common area and class room|
Our lab publishes on three different sub-fields of robotics: novel devices, robotic algorithms and systems, and the impact that these systems have on people. These thrusts inspire each other and can best understand as a loop: novel devices enable novel robotic systems and algorithms, eventually leading to novel applications and interactions, which can be studied systematically. These studies in turn might lead to novel device requirements. A good example for this design loop is “Flutter”, an assistive garment for the deaf, which has begun as a dress, became a robotic skin, and soon might enable robots to solve new manipulation tasks.