Information Technology Reference
In-Depth Information
a detailed identification of the place where the mobile agent is localized or to rec-
ognize the several possible goals, indoor navigation requires managing different
data provided by diverse sensors, and combining the type of information each
sensor is able to give. In addition, the environment can provide complementary
information for navigation by means of sensors that are usually devoted to other
purposes (such as people location). In this way, the combination of the informa-
tion obtained from the agent's own sensors and from the context may enhance
navigational abilities. The data provided by a sensor can be scalars (such as the
output of sonars, infrared, laser, ...), vectors (containing information about the
position or the trajectory of the mobile agent provided by odometric sensors),
booleans (provided by contact sensors or bumpers), waveform signals (such as
pictures supplied by digital cameras or sounds from microphones), angles (from
compasses), etc. [26, 11]. Proximity sensors are commonly used to handle geo-
metrical concepts such as distances, position identification, etc. Among the vast
diversity of sensors, vision systems are frequently used for indoor navigation to
detect patterns in the environment, by means of artificial vision techniques. Im-
ages are potentially rich information sources [22] but computationally complex
and expensive. Different information can be extracted from color images: shape
of objects, color information, optical flow, etc. Textures and colors of the dif-
ferent environmental elements including people, walls, doors, floor, etc., can be
used to navigate safely by avoiding obstacles or as natural landmarks that help
the agent to locate itself globally in the environment.
2 Navigating Techniques from Mobile Robotics
Mobile Robotics has developed several methods for indoor navigation [42]. Many
of them are based on centralized models of the environment: maps. Data provided
by the robot's own sensors (frequently complemented by sensors located in the
environment) are used to match the current state to the model in order to plan
the path to the goal and perform related tasks. Alternatives to map-based navi-
gation strategies are biologically inspired navigation methods (Behavior-based)
that imitate navigational cues observed in animals [30]. This section reviews
these two approaches starting with the later that is the approach taken in the
example described in Section 3.
2.1
Behavior-Based Navigation
Behavior-based (BB) systems appeared in 1986, when R.A. Brooks proposed a
bottom-up approach for robot control that imposed a new outlook for develop-
ing intelligent embodied agents capable of navigating in real environments and
performing complex tasks. He introduced the Subsumption Architecture [13, 15]
and developed multiple robotic creatures capable of showing different behaviors
not seen before in real robots [21, 49, 16]. Behavior-based systems are originally
inspired by biological systems. Even the most simple animals show navigation
capabilities with a high degree of performance. For those systems, navigation
Search WWH ::




Custom Search