Summary:
The research discussed in this talk is focused on algorithmic issues that arise in highly-distributed robotics systems at the nanoscale. The fundamental problem to be addressed is how to program large numbers of limited-capability robots so that they assemble themselves into complex nanostructures with specified shapes. Unexpected connections between geometric modeling and distributed robotics emerge from this work.
Nanorobots have dimensions comparable to those of biological cells, and are expected to have remarkable applications in health care and environmental monitoring. For example, they might serve as programmable artificial cells for early detection and destruction of pathogens. The motivation for this project, however, does not stem from environmental or biomedical applications, but rather from the need to build nanostructures efficiently. Assembly of nanoscale components with Scanning Probe Microscopes (SPMs), which is used extensively today, is inherently a sequential process, ill-suited to mass production. Passive self-assembly, that is, the autonomous assembly of building blocks driven by natural phenomena such as termal agitation or surface tension, is an efficient, massively parallel process, but typically cannot produce the asymmetric structures needed by many applications. Active self-assembly, the approach proposed here, uses nanorobots as building blocks, and programs them through simple local rules to generate specified global shapes. Once the desired configuration is achieved, the constituent robots themselves become the structure. Active self-assembly is a massively parallel process that does not appear to have the drawbacks of its passive counterpart.
Individual nanorobots of the future are expected to have limited capabilities: small memories and processors, few sensors, and communications primarily based on chemical interactions, which require contact between robots. A robot per se will not do much, but large teams of such robots may achieve complex behaviors. In the work reported in this talk nanorobots teams are modeled as swarms of robots in which each robot has a simple shape, is controlled by a small finite-state machine, and is capable of moving autonomously and exchanging a small number of messages with other robots only when they are in contact. The results to be presented show that it is possible to use such models to build wires (or, equivalently, to plan collision-free paths) between given locations, as well as to construct various shapes. The ultimate goal of the project is to demonstrate that essentially arbitrary nanostructures can be built by programming such robot swarms.