Path Planning And Collision Avoidance For An Indoor Security Robot

Abstract

Any mobile robot which must operate in a dynamically changing indoor environment, such as an office, laboratory, or warehouse, must be able to detect and successfully avoid unexpected obstacles. Transient objects such as chairs, doors, trash cans, etc. change position or state frequently, and thus cannot be assigned a static representation in an "absolute" X -Y planview map of the workspace. The most simplistic path planning scheme therefore assumes there are no transient objects in this global model for the initial "find-path" operation. For collision avoidance purposes, a secondary "relative" model of the robot's immediate surroundings is created from real world sensor data collected as the robot is moving, and used to find a path around each individual obstruction as it is encountered. No information regarding the position of permanent objects is available in this smaller relative model, and the position of each transient object is forgotten as soon as it no longer obstructs the path. Conversely, if the absolute position of each detected obstruction is simply recorded in the global map, the resulting model eventually fills up with clutter and the find -path operation fails because no free path exists. This paper discusses a robust approach for map maintenance implemented on a prototype security robot, wherein transient objects are added to the global map as they are encountered, and removed from the model later if no longer detected at the same location. In this manner, subsequent find -path operations will avoid previously identified obstructions, and information on the location of both permanent as well as transient objects is available when reacting to the discovery of a new obstruction.

Directed Sonar Sensing for Mobile Robot Navigation

“…Hence, position estimation is only possible with a correlation-based technique, which is time-consuming and subject to the cell resolution of the grid. Gilbreath notes that one motivation for using a grid-type representation is that "it is hard to accurately glean polygonal information from inexpensive sensors mounted on a mobile robot" [57]. The results presented above provide strong evidence to the contrary.…”

“…Beckerman and Oblow [9], Gilbreath and Everett [57], Noborio et al [109] and Zelinsky [136] have used a grid representation to build sonar maps. Beckerman and Oblow have presented a map building algorithm that uses a consistent labeling scheme to resolve conflicts in a 6 inch cell resolution grid [9].…”

“…The vehicle Robart II, used by Gilbreath and Everett [57], uses two sonar rings to build grid-type maps, one positioned near the floor and the other about 4 feet off the ground. Mapping is based on the assumption that transient obstacles tend to be lower to the ground, and hence visible to the lower sonar ring, while permanent room features such as walls will be visible to the higher mounted ring.…”

Feature Extraction from a Broadband Sonar Sensor for Mapping Structured Environments Efficiently

“…However, the problems associated with ultrasound (large beamwidth and specular reflections) mean that it is almost impossible to extract line segments directly (Crowley 1985(Crowley , 1989) to match external plans or maps. The brittleness of line segment-based feature extraction was a motivating force behind the grid representation for sonar maps (Beckerman and Oblow 1990;Elfes 1987;Gilbreath and Everett 1988;Noborio, Kondo, and Noda 1990;Zelinsky 1988). Grid-type representations are useful for obstacle avoidance (Borenstein and Koren 1990) but cannot be correlated easily with most external representations.…”

Surface Profile Determination from Multiple Sonar Data Using Morphological Processing

“…A measure for judging map accuracy was proposed by Raschke and Borenstein (1990), where improved mapping results were reported with this method. Gilbreath and Everett (1988), Zelinsky (1988), Beckerman, and Oblow (1990) have all used variations of the gridbased approach to construct sonar maps for path planning. The use of grids has been extended to 3-D sonar-range sensing for under water by Auran and Silven (1996).…”