Monthly Technical Report

November 1998


We have completed the task of defining realistic models and data structures for

  1. representing the map of a building being explored,
  2. describing the perception and mobility capabilities of the robots in a parametric form that is usable by our future software modules,
  3. expressing the motion plans to be computed by these modules, and
  4. defining a parametric model of a target's behavior.
A report describing these models and data structures will be emailed at the beginning of January 1999.

We are continuing to refine and implement the next-best-view techniques we have developed for model construction (in two dimensions) by a team of mobile robots equipped with range-finders. We expect that a software demonstration (in a simulated environment) will be ready in time for the Second Quarterly TMR IPR meeting in Denver Colorado in January 1998. The goal of our algorithms is to minimize the time needed for constructing a model. We make the robots navigate through different areas of the environment and exchange model data whenever appropriate. We also ensure that each robot can precisely localize itself by matching its current sensory data with the model built so far.

We have developed some new techniques for a team of robots to sweep an environment so as to find any set of targets that are hiding it. We ensure that the robots maintain a communication network at all instants during the sweep. Our algorithms tailor their performance to the complexity of their environment; thus, we use a small number of robots for simple environments but might use more robots for complicated workspaces. We are planning software implementations of these techniques.

In summary: