Simultaneous Localization and Mapping

The SLAM process (English Simultaneous Localization and Mapping; German Simultaneous localization and mapping ) is a method by which a mobile robot simultaneously create a map of its environment and can appreciate his pose inside this card.

Task

One of the basic skills of a mobile robot is to be able to orient themselves, so knowing how his environment looks like and where it is located. Is a map of the surrounding area are available, a robot with its sensors such as ultrasound or lidar can position it. Is the absolute position of the robot is known, a map can be built. The robot measures the relative position of possible obstacles to it and can then determine with his well-known position, the absolute position of the obstacles, which is then added to the map.

SLAM is thus a chicken and egg problem, since neither the card nor the position is known, but these are to be estimated simultaneously.

Application

For many locations of mobile robots there are no maps, and no way, the absolute position, eg via GPS to estimate. Without SLAM would have prior to use one card can be created, which can delay and cost of using. Therefore, it is important depending on the application that a robot is able to autonomously explore a new environment and to create a card that he can use later for navigation.

The SLAM method is an active area of ​​research within the field of robotics, which is edited extensively by many research groups.

Approaches

There are many different approaches there being basic similarities. Since a robot usually can only see part of the environment, the map is built up incrementally: first, no card is present and the position of the robot defines the origin of its coordinate system. This is trivially the absolute position of the robot is known and the first measurement of the environment can be entered directly into the card. Then the robot moves and measures again its surroundings. If the robot has not moved too far, he is a part of the already known environment again - but also a previously unknown area for the first time - measured. From the overlap of the new measurement with the previous card, the movement of the robot can be calculated again so that the absolute position is known, and thus the new measurement can be integrated in the card. In this approach, the map is incrementally expanded until the entire area is measured.

Since the determination of the movement of the robot between two measurements, however, is never exactly the calculated position of the robot from the true will vary continuously, whereby the quality of the map decreases. For the card still remains consistent, the algorithm must be able to recognize when an already well-known part of the environment is measured again (loop closing).

SLAM method

  • EKF SLAM (Extended Kalman Filter) SEIF ( Sparse Extended Information Filter )
  • UKF ( Unscented Kalman Filter)
  • FastSLAM
  • Grid - based methods with Rao - Blackwellized Particle Filters
  • DP - SLAM
  • Graph Slam, TORO, HOG -Man, TreeMap
  • Relaxation techniques
  • Smoothing techniques

The solution of the solution of the SLAM conditioned data association problem, i.e., it must be determined, which correspond to (ambient) characteristics. This problem is particularly difficult, since this does not extract features with absolute certainty. Scan matching method come from without features, since they take into account all the scans, or point cloud, and then use graph -based techniques.

Pose (computer vision) Iterative and incremental development Sebastian Thrun Wolfram Burgard Udacity
731508
de