Introduction

Perception and sensing of the environment is an essential functional part of GIM, because there is not usually enough a priori information available for the task to be capable of being performed in an open loop form (as in the case of most industrial robot applications). The actual necessary information can be obtained only through sensing, when the robotic machine is moving or doing work autonomously. Most GIMs are designed for unstructured environments, such as the natural environment or disorganized urban environments. This emphasizes the role of sensing and often makes perception very demanding.

The other important primary task which all mobile robots have to perform is navigation. Navigation can be based on different methods and means. Recently there has been much progress in this field, both in new technology, such as micromechanical inertia sensors and laser scanners, and methods, such as estimation algorithms that allow the mapping of the environment concurrently with navigating. There remains, however, much to be done in this field to make the navigation methods more adaptive, reliable, and simple. Localization and perception are also closely-linked tasks; perception produces information about relative changes in position and orientation, while knowing the position makes perception easier, and attention and expectation can be utilized more effectively.

In order for robots to work with humans, there must exist mutual understanding between them. The robot works in the human world and carries out tasks defined by humans. This, however, does not mean that the robot has to have human cognition. On the contrary, the challenge is how to incorporate human cognition into perception and SLAM algorithms. A human is more capable of making high-level associations than a robot. On the other hand, a computer can perform a statistical analysis of any data set given. Thus, if the environment model supports it, it is possible to share abstractions between humans and robots - to teach the robot without explicit models.

Research challenges in navigation, perception, and SLAM aim at developing robust and practical methods for mobile work machines and robots to operate among people in outdoor environments. In this respect safety aspects are of top priority. Perception and navigation methods that enable flexible co-operation between man and machine and which simultaneously guarantee safe operation are the main research areas. In unstructured outdoor environments uneven terrain and cluttered surroundings make most of the current state-of-the-art navigation and SLAM methods difficult to apply. Some steps to solve the problems have already been taken by the previous works of Jorma Selkäinaho and Pekka Forsman. The former discusses adaptive navigation methods for outdoor environments and the latter 3D navigation and mapping of forest environments.

Another type of challenging environment is the scene of a catastrophe, for example a fire or an explosion in a large building. Such conditions are often unpredictable and difficult both from the perception and navigation points of view. In these environments simultaneous multi-agent operation by both humans and robots has been proposed as a key technology to be further developed for the most demanding situations.