Ant-Like Microrobots - Fast, Small, and Under Control

Abstract

The goal of this multidisciplinary project is to develop the first cooperative network of small robots. It involves the concurrent investigation of control and estimation algorithms, actuators, sensors and platforms that support mobility and collaboration subject to severe power, size and weight constraints.

The following is a partial description of our major research achievements:

• We have developed new algorithms for control and navigation. In particular, we have investigated receding horizon control principles for robot coordination that are robust to inclusion and removal of agents, and are amenable to provable and tractable methods to determine performance degradation as a function of computational constraints. We have also proposed methods for distributed navigation that are rooted on the design of controlled MDPs via entropy maximization principles.

• We have proposed design methods for very low power mixed--signal circuit architectures that can provide accurate localization information. Mixed--signal strategies have also been proposed and tested for simple cooperative control tasks, such as rendezvous and leader--follower.

• We have discovered new microfabrication methods that enabled the creation of actuators at very small scales. In particular, we were able to fabricate dielectric elastomer actuators via micromachining techniques. These actuators can be integrated with on--chip circuitry and support a contraction of over 100nm. A microfabrication technique has also been proposed and tested to create small--scale legs by adding elastomers to silicon MEMS. An actuation scheme based on electroosmotic pumping has also been investigated that is guaranteed to lead to stable performance.

• We developed, constructed and tested a new low cost centimeter--scale microrobotic platform "Tiny TeRP: A Tiny Terrestrial Robotic Platform". Besides a new chassis that allows for great maneuverability, it contains communication and computation modules capable of supporting cooperative control tasks. The platform hardware and software are fully released under open source licenses.

• We discovered a new parametrization that significantly extends the class of distributed control problems for which optimal design is tractable using convex methods.

We are currently testing and improving the proposed localization systems. We are also transitioning some of our navigation algorithms into the platforms. Plans for a new platform are also being developed.

Education and outreach

As a result of this project, the Institute for Systems Research supported the creation of a new laboratory where our prototypes and algorithms can be developed and tested. Since its creation, the new laboratory has also been heavily used to support undergraduate projects in the area of robotics, with an emphasis on small platforms.