Vision based model generation for indoor environments

Abstract

This paper presents our approach to retrieve a dependable three-dimensional description of a partially known indoor environment. W e describe the way the sensor data f r o m a video camera is preprocessed b y contour tracing to extract the boundary lines of the objects and how this information is transformed into a three-dimensional environmental model of the world. W e introduce a dynamic map that operates in a closed loop with various sensor systems improving their performance by filtering and contributing certain knowledge. The filtering relies on the capability of a mobile robot to gather sensor readings f r o m different positions. A n important part of our approach is the interaction between the dynamic map, storing and filtering the incoming information, and a module predicting missing sensor features based on structures andreference objects. This interaction helps to generate a more accurate model containing also poor detectable features, that are impossible to extract from a single sensor view. 0-7803-3612-7-4/97 $5.00 0 1997 IEEE

Visually Guided Cooperative Robot Actions Based on Information Quality

“…In SLAM a robot is localizing itself as it maps the environment. Researchers have addressed this problem for well-structured (indoor) environments and have obtained important results (Anousaki, 1999;Asada, 1990;Burschka, 1997;Castellanos, 1998;Choset, 2001;Kruse, 1996;Kuipers, 1991;Leonard, 1991;Thrun, 2000;Tomatis, 2001;Victorino, 2000;Yamauchi, 1998a). These algorithms have been implemented for several different sensing methods, such as camera vision systems (Castellanos, 1998;Hager, 1997;Park, 1999), laser range sensors (Tomatis, 2001;Yamauchi, 1998a), and ultrasonic sensors (Anousaki, 1999;Choset, 2001;Leonard, 1991;Yamauchi, 1998b).…”

“…Researchers have used model-based approaches, requiring previously known environments (Burschka, 1997;Cowan, 1988;Hutchinson, 1989;Kececi, 1998;Laugier, 1995). Target motions (if any) are assumed to be known (Laugier, 1995).…”

Visually built task models for robot teams in unstructured environments

“…One of these is concerned with sensor positioning-placing a sensor so that it can best observe some feature and selecting a sensing operation which will prove the most useful in object identification and localization. Researchers have limited their work to model-based approaches, requiring previously known environments [4,6,9]. Target motions (if any) are assumed to be known.…”

“…Mapping has been done in sequential brute force fashion [13,19]. Researchers have addressed the concept of map building using a single mobile vision system [1,4,10]. Often, sensor models and data uncertainty are not fully considered [1,10,11] or exploration schemes are not developed [3,14].…”

“…Often, sensor models and data uncertainty are not fully considered [1,10,11] or exploration schemes are not developed [3,14]. Structured environment and partial knowledge assumptions are also made [4,14]. This paper proposes an environment and task model building algorithm, to overcome the uncertainties in robot and camera location and orientation, for robot teams cooperatively working in an unstructured field environment.…”