The GRITS Lab uses self-assembly with a fixed number of modules. Pickem’s research has shifted from working with homogeneous and centralized architectures to heterogeneous and decentralized approaches.
“The main reason is that I want to create a general framework for reconfiguration and scale my approach up to very high numbers of modules, to incorporate thousands or even millions of modules as robots become smaller and smaller,” he said. Existing hardware prototype robots are still fairly large, so at smaller scales more modules will be needed to assemble a given shape. Eventually, centralized control won’t be an option because communication and planning costs will become too high. Planning for so many centrally controlled robots won’t be possible on a reasonable time scale, and communicating among robots, and between robots and the central planner, will become a real issue.
“My vision for my own research is to create a general-purpose robot,” said Pickem. “Like a human being, the robot would adapt to the task, instead of adapting the task to the robot. If it needs to open a door, it forms an arm. If it needs to go fast, it forms wheels. I want to bridge the gap between what the task requires and how the robot is structured.”
There is no killer app yet for modular, self-reconfiguring robots, said Pickem. This is partly because hardware implementations are bulky, unreliable, and available only in small numbers. Their use is mainly targeted to environments with limited access, or that are otherwise space-limited in some way. For example, a search-and-rescue robot could pack itself really small, get into damaged buildings, and then expand into a larger form to do more tasks, like lift fallen walls. “For example, the robot could enter the building as a snake, and reconfigure into a shape that can walk or drive on wheels,” he said.
An aerospace application could be a modular robot for planetary exploration. In home automation, they might be used for reusable and reconfigurable furniture. Although modular robots won’t replace robots used in manufacturing processes, once they can be made small enough they may replace certain manufacturing processes altogether. “There will be a place for both, since not all goods we use can or should be made from modular robots,” said Pickem.
Miniature modular robots could eventually assemble into consumer products. “For example, a car could change its shape, color, model, or interior design, according to your daily needs or desires,” said Pickem. Instead of manufacturing parts to be assembled into furniture, one could buy a number of modular robots that self-assemble into the desired shape.
“Obviously, these robots would still have to be manufactured, but the classical manufacturing model where each product requires a specific manufacturing process would be replaced by a single process that builds only these modular robots.” Some products -- those that need high structural stability, don’t need to change shape, or aren’t long-lived -- will still be manufactured in the conventional way.