The paper demonstrated that a visually guided mobile robot can safely map, explore and navigate unknown indoor environments. We have shown that real--time stereo vision is a viable alternative to active sensing devices for these applications. Robot velocities were limited by the speed at which stereo could be computed. With our current system calculating stereo depth images at 2 Hz, the robot traveled at a speed of 0.5 m/s without degrading safe navigation or map updating.
Occupancy grid mapping provided a good base for raster-based path planning techniques. Planning techniques that combine shortest path searches with repulsion fields from nearby obstacles can strike a balance between the most direct and the most safe path. The exploration module mapped the entire robot environment via an efficient path.
While the Real World Interface robot base odometry proved to be adequate for our task, future work will include localization of robot position to complement odometry readings. Spinoza has an additional color camera on a pan tilt unit that we plan to use for tracking and identifying landmarks in a localization module. In addition, linking of local maps into a connected graph of nodes to represent larger spaces should improve the overall range and capabilities of the system.