Support Vector Machines and Features for Environment Perception in Mobile Robotics
Environment perception is one of the most challenging and underlying task which allows a mobile robot to perceive obstacles, landmarks and extract useful information to navigate safely. In this sense, classification techniques applied to sensor data may e
- PDF / 11,456,525 Bytes
- 32 Pages / 430 x 660 pts Page_size
- 85 Downloads / 148 Views
Abstract. Environment perception is one of the most challenging and underlying task which allows a mobile robot to perceive obstacles, landmarks and extract useful information to navigate safely. In this sense, classification techniques applied to sensor data may enhance the way mobile robots sense their surroundings. Amongst several techniques to classify data and to extract relevant information from the environment, Support Vector Machines (SVM) have demonstrated promising results, being used in several practical approaches. This chapter presents the core theory of SVM, and applications in two different scopes: using Lidar (Light Detection and Ranging) to label specific places, and vision-based human detection aided by Lidar.
1 Introduction Learning models of the environment [2] is required in tasks such as localization, path-planning, and human-robot interaction. Other examples of application of learning systems to environment perception are the following: Industry Product Manufacturing Learning algorithms are important in a production line to check if a product contains some error. The application of these methods increases the productivity and guarantees higher quality of the product. Security Important security systems are based on learning methods. Two example applications are burglar detection in banks and casinos systems, and terrorist detection in airports. These systems try to detect dangerous people by automatically identifying their face. Biological Sciences For example, learning systems are intensively used on classification of protein types, based on the DNA sequence from which they are generated. Research is also being carried out for the application of learning systems for understanding how the human brain works. The process of constructing models by explicitly considering and coding aspects of the environment is not trivial. Thus, methodologies which aim at creating automatic perception and system control are necessary. L.C. Jain et al. (Eds.): Comp. Intel. Para.: Innov. Applications, SCI 137, pp. 219–250, 2008. c Springer-Verlag Berlin Heidelberg 2008 springerlink.com
220
R. Ara´ ujo et al.
The problem of learning to perceive the environment from sensor data is an active research topic and it is usually frequent that certain systems may be able to perceive or comprehend the environment in a way similar to human beings. If this is achieved, a new set of applications can be proposed, e.g., intelligent systems may interact with users using a new perspective. In robotics, the robot would be able to build models and to react to the world using new approaches. Machine learning techniques have been used to tackle the particular problem of building generic models from the input space in order to recognize patterns which are implicitly present within data. Amongst the various approaches available on the literature, SVM has been receiving attention in the last few years. These machines represent learning structures that, after being trained to accomplish classification tasks (pattern recognition) or approx
Data Loading...
