Comparison of FAST SLAM And Q SLAM

Simultaneous Localization and Mapping:
Simultaneous localization and mapping (SLAM) are used in a computational problem that constructs and updates the map of an unfamiliar environment and simultaneously keeps the agent track’s location in the location. It is used in computational geometry and robotics. It usually appears simple, but various algorithms are required to solve it. These algorithms solve it within a time that can be traceable for some environments. Some approximate solution approaches consist of the extended Kalman filter, GraphSLAM, particle filter, and Covariance intersection. These algorithms are applied to navigation, odometry for augmented reality and virtual reality, and robotic mapping. SLAM algorithms are used for tailoring the available resources at operational compliance. Therefore, the aim is never to achieve perfection. Self-driving cars, self-sufficient underwater vehicles, aerial vehicles that are unmanned, the latest domestic robots, and planetary rovers use published approaches.
SLAM Problem:
Simultaneous Localization and Mapping are needed.
 

For localization and mapping, the SLAM algorithms use the basic problems of Chicken or Egg. The SLAM task includes mapping the environment and to detect the robot pose concerning the environment. If the map is not available, then the robot finds it hard to localize itself. The location is necessary to build the map, which will help it to find its location.
To explore a static and unknown environment by providing the robot’s controls and based on the observations of nearby features, by SLAM, you can estimate the features map, pose, or the path of the robot.

Why is SLAM a hard problem?
 
 

There are various uncertainties as there could be an error in observation, an error in the pose, the error accumulated, and an error in the mapping.
The map and the robot path both are unknown. Any error in the robot path corresponds to the errors in the map.
Observations and landmarks are unknown in the mapping in the real world. Also, if the wrong data is picked, there could be catastrophic consequences. The error in the pose correlates to the data associations.

FastSLAM Algorithm:
 
The Flastlam algorithm uses the particle filter approach to the SLAM problem. It maintains a collection of particles. These particles comprise a map and the sampled robot path. Own local Gaussian represents the features of the map. A separate set of Gaussians Map features is created, which constitute the map. The Gaussians Map features are independent of the conditions.
How does the algorithm work?
First, the conditionally independent map features are given to the path. It factors one particle per path. This makes the features of the map independent. Then correlation is eliminated. The sample new pose of the FastSLAM is updated and the observation features are updated. This update can be performed online. It can solve both offline and online problems based on the SLAM. The instances include feature-based maps and grid-based algorithms.
FastSLAM 2.0 Algorithm:
FastSLAM 2.0 sample poses are based on measurement and control to avoid the problem.
Step 1: Sample the new poses by extending the path posterior.
Step 2: Observe the features and update them.
Step 3: Do the re-sampling.
Features of Fast-SLAM:
 

Every single particle can rely on itself. It supports decisions based on local data association.
The data association decision is more robust and is based on a per-particle basis.
It can provide a solution to online and offline SLAM problems.
The FastSLAM 1.0 is less effective in creating samples. However, FastSLAM 2.0 is more and at the cost of mathematical complexity.

Article Source: http://EzineArticles.com/10443463