SLAM

A 3D scanner is useful for smaller objects, with surrounding sensors facing inward. For larger objects and stationary models, the sensors are usually inside facing out; see Figure 9.26(b). A common example of a stationary model is the inside of a building. Scanning such models is becoming increasingly important for surveying and forensics. This is also the classical robotics problem of mapping, in which a robot carrying sensors builds a 2D or 3D representation of its world for the purposes of navigation and collision avoidance. Robots usually need to estimate their locations based on sensors, which is called the localization problem. Robot localization and tracking bodies for VR are fundamentally the same problems, with the main distinction being that known motion commands are given to robots, but the corresponding human intent is not directly given. Robots often need to solve mapping and localization problems at the same time, which results in the simultaneous localization and mapping problem; the acronym SLAM is widely used. Due to the similarity of localization, mapping, and VR tracking problems, deep connections exist between robotics and VR. Therefore, many mathematical models, algorithms, and sensing technologies overlap.

Consider the possible uses of a large, stationary model for VR. It could be captured to provide a virtual world in which the user is placed at the current time or a later time. Image data could be combined with the 3D coordinates of the model, to produce a photorealistic model (recall Figure 2.14 from Section 2.2). This is achieved by texture mapping image patches onto the triangles of the model.

Steven M LaValle 2016-12-31