sanan at cmu dot edu
The use of robots in assistive roles will be an increasingly significant application for robotics. This is fed by the growing aged populations in countries such as the US and Japan. The application of robots in these roles has unique demands from a mechanical and control design perspective, which have not been thoroughly addressed in the past.
Safety is key when robots and human work in close proximity, while accuracy is secondary. The paradigm shift thus needed in the way we design robots, is a unique departure from the way humans have thought about machines in the past. The robot must minimize the risk of injury to others during its operation. We are exploring the use of robots with flexible and soft structures for making robots inherently safe during physical interaction.
When we speak of safety, we will usually mean safety under uncontrolled impacts unless otherwise stated.
Rigid links have always been the basic component for most robots. The reasons for this are both technical and historical. Historically, it has always been a goal in machine design to develop machines using rigid components. This allowed relatively simple synthesis of mechanisms based on kinematic considerations. As these were machine parts, safety from the viewpoint of physical interaction was almost certainly never a consideration. As robot design largely borrowed from machine design, the same design principles were applied.
One basic approach to improving safety while still utilizing a rigid structure, is to use of a soft padding around the robot. While it is effective in improving safety, the amount of padding required to make a robot safe while still being reasonably fast is prohibitive. Researchers besides our group at RI, have focused mostly on the use of joint compliance. In this approach, the connectivitiy between the actuator and the link is modified by the introduction of compliance between the two. While the approach is useful, it assumes the same design principle of using rigid components.
In our research, we open our eyes to the possibilites of robots which throw away assumptions regarding the need for rigid components. This allows us to develop novel designs which may be more suited to phyical interaction with humans. Our problem is then to design and control robots in this extended design space, for safe physical human robot interaction
Refer to the Media section for videos of robots currently under development.