The Virtual Robot Test and Integration Laboratory (VRTIL) addresses the need to verify robot software using real data, produced by real sensors, in all its richness and imperfection. The project is embedding robot software in a synthetic environment in order to mimic its sensor interfaces. In an advance over the state of the art, these interfaces are windows into data bases that store the actual sensor data that a robot did read when it was in a particular state. The project goal is to reproduce the real world interface as exactly as possible using data. Two basic tools enable such ?virtualization? of a real robot in a real environment:
* Virtualized reality provides a means to produce synthetic views from virtual viewpoints that are near to an actual viewpoint for which a real image is available;
* High fidelity synthesized motion gathers large amounts of data to calibrate a model that is valid across all of state space.
Combining these ideas will result in a software test environment for robot software that has the realism of data logs and the responsiveness of a simulator. This may fundamentally transform the development of robots by lowering the cost and time of testing and debugging.
