SLAM (simultaneous localization and mapping) is the incremental construction of an stochastic map of the environment while concurrently generating an estimate for the location of a mobile vehicle. The SLAM process needs an arbitrary point to start, then a mobile robot should be able to autonomously explore the environment with its sensors, gain knowledge about it, build an appropriate map and localise itself in this map, (relative localization, not absolute). The resulting model can then be employed by planning and navigational strategies to achieve the robot goal positions in an efficient manner. SLAM techniques and absolute localization is a very active research topic.
Map building and position estimation are basic tasks in mobile robot navigation with path planning. Depending on the physical properties of the walls that form the room where the robot is navigating, sonar sensors show different behaviours. Map building is an important topic in control architectures that employs planned paths, in the sense, the robot is not needing a knoledge of the environment when it starts to move. Through a first exploration, the robot is able to make an internal representation of its work environment. For indoor environments, a metric map is a more compacted representation of the environment. The squared grid maps are the most widely used of this kind of environment models. Moreover, employing grid map, path planning is quite simple.
Among the external sensoring field used in mobile robots, ultrasonic sensors for range measures have been widely employed in the last few years, mainly due to their simplicity, robustness and low cost. However sonar range readings are affected by many factors such as the wide angle of their radiation lobes, multiple reflections (specular reflections), diffuse reflections, fluctuations in the propagation medium and so on.
Ultrasonic sensors depend on two separate devices: an ultrasonic transducer and a detector. An ultrasonic transducer is a device that converts energy into an ultrasonic frequency. Ultrasonic transducers are usually made from piezoelectric crystals that can change size when a voltage is applied to them. When an alternating current is applied to a piezoelectric crystal, it vibrates extremely fast and produces an ultrasonic sound wave. The detector is also made of a piezoelectric crystal, but produces a voltage when an ultrasonic frequency comes in contact with it, effectively producing the opposite results. A sensor calculates the time that it takes in between broadcasting the ultrasonic frequency and receiving the incoming waves.
Ultrasonic sensors produce ultrasonic frequencies that humans cannot hear, making them ideal for quiet environments. They do not use much electricity, are simple in design, and are relatively inexpensive. Some piezoelectric crystals transmit and receive ultrasonic sound waves. Likewise, ultrasonic sensors can be used with both radio and sound waves.
Ultrasonic sensors do not have many disadvantages, but are limited in their capabilities. For example, density, consistency, and material can distort an ultrasonic sensor’s readings. One of the most common sources of error in range reading are multiple reflections that depend on object surfaces over which sonar wave is reflected. A classification between smooth wall environments (specular rooms) and rough wall environments is done by most of authors. For high incidence angles, multiple reflection are produced in the first type of walls (smooth walls) causing greater range readings than the real distances.
Author:
David Alejandro Trejo Pizzo. IEEE Member and researcher @ AIGROUP, working in the FIC Project. Student @ Universidad de Palermo.
References:
Toledo, F.J., et al. Map building with ultrasonic sensors of indoor environments using neural networks
Miró, J.V., et al. Vision-based using natural features in indoor environments.
Borenstein, J. & Koren, Y. Obstacle avoidance with ultrasonic sensors. IEEE Journal on Robotics and Automation, vol. 4, no. 2, pp. 213-218, 1988.