“Simultaneous localization and mapping” is a relatively recent advancement in techniques for robot-human collaboration. SLAM stands for simultaneous localization and mapping. The robot uses this technology to explore an unfamiliar environment by moving around it. Periodically, the robot makes measurements or takes pictures of its surroundings to build up a map of the area. The map then continuously updates itself as more data becomes available until, finally, it can use that information to find a way home. It’s a technique originally invented in in-depth perception research.
What Does SLAM Do?
Computer vision provides a way of processing images that allows computers to determine where objects are in them, independent of the camera’s viewpoint. Computer vision uses color, texture, light intensity, and depth perception. The input for these perceptual elements to the computer is images captured by a camera. The output of these perceptual elements is a three-dimensional point cloud. A point cloud represents where each pixel is located on the x, y, and z-axis in the image. Using machine learning algorithms, the computer recognizes objects within its environment. It then produces a map that depicts how each object interacts with other objects in the environment around it.
Simultaneous localization maps an image onto a 2D space from 2 viewpoints (left and right eyes). The algorithm is made up of two parts: a model of the 3D world and a localization algorithm. These components are combined so that when an object is seen by one viewpoint (left or right), it is offset by the same distance in 2D space. The process works like how your eyes work from two different viewpoints when you see an object from one eye. If a person standing behind you were to turn her head to look at you, her head would appear to move left for her left eye and right for her right eye because the camera was looking at it from 2 different viewpoints. That is known as binocular parallax. For instance, when you look at Dioram SLAM, you will realize that to determine where objects are in 3D space, SLAM uses 2D image matching and camera calibration. This process allows a computer to learn where to place each pixel in its point cloud.
How Does It Work?
The process of SLAM depends on the existence of a point cloud that a previous step in the process has generated. These points must already be determined correctly before the SLAM algorithm can work correctly. That is why SLAM must be done in the reverse order to how it is normally done. It must first be determined “on which visual feature of a previously acquired image, the computer should search for points.” The steps involved include:
The algorithm used to do this can be found in Google’s paper. The paper explains that if an object is located on either side of a previous point, you should not use it as a reference point when determining the new pixel location. Rather, they use a nearby similar color or an average of nearby colors. This process uses correspondence learning and generative adversarial networks (GAN). A computer vision technique commonly used in the SLAM algorithm is that of stereo matching. This roof-top algorithm uses difference thresholds to determine whether the object being viewed out of the left or right eye is in front or behind.
Applications of SLAM
Disaster Assistance
Charitable organizations, government agencies, and disaster relief organizations use simultaneous localization and mapping to improve their ability to provide immediate assistance following a natural disaster or human-induced events such as an explosion, terrorist attack, or earthquake. This technology enables emergency personnel to enter foreign areas and find survivors following a natural disaster. Using cameras mounted on unmanned aerial vehicles (UAVs), emergency response organizations can discover survivors who might otherwise be missed by untrained search parties or people simply looking the wrong way. Cameras mounted on UAVs provide a bird’s eye view of disaster areas, and the simultaneous localization and mapping technology provide exact locations of survivors. This information can then direct emergency responders to specific locations, only saving time and money while helping as many people as possible.
Geo-tagging
Using simultaneous localization and mapping technology enables users to geo-tag or “label” their information, such as photos, videos, and location data. This labeling allows users to categorize and organize information in a structured manner. Several services provide this functionality, including Flickr and Google’s visual search feature. The term “geo-tagging” also means tagging objects and people. That is done by assigning tags based on a location acquired through the GPS receiver in photos or videos you take with your cell phone. Examples of tagging people and places include:
Tracking and Monitoring
Simultaneous localization and mapping are used in monitoring and tracking mobile devices, such as cell phones or hand-held computers, for example, to keep track of the location of a business employee. With this technology, if an employee gets lost, an employer can immediately find him. Tracking individuals provide a valuable service to employers by allowing them to monitor the whereabouts of their employees. That can also help with government-mandated paperwork that employees may have to complete. For example, driver’s license applications typically require proof of current residence. This technology enables license agencies to instantly display the location of a motor vehicle operator’s home.
Supply Chain Management
Simultaneous localization and mapping allow businesses to manage the location of their assets and inventory in real-time. These assets include machines, trucks, containers, and other items used in the manufacturing process and products that are being delivered. The technology can track shipments by following them through a warehouse or a shipping yard, for example. You can also use it to monitor truck drivers when delivering merchandise from one depot to another or from one store location to another. That can be particularly helpful when merchandise must be delivered on a certain day. The technology allows businesses to know exactly where their assets are at any given time. That will enable them to minimize out-of-stock situations, overages, shrinkage, theft, and delays in delivery.
Automotive Navigation
In their latest effort to turn Google Maps into a complete tool of mobile phone technology, Google has integrated SLAM technology. That provides the user with 3D off-road navigation and real-time ghosting of road info.
SLAM also allows drivers to see a digital overlay of traffic information and local landmarks visible through their car’s windshield.
Simultaneous localization and mapping (SLAM) is a sensor-driven technique for autonomously creating two-dimensional maps of one’s environment and simultaneously registering oneself within that map. It does this by taking advantage of the ever-present laws of physics (e.g., gravity) to build an accurate map. That can be accomplished in real-time with the help of a single camera or through scans from two cameras if necessary for more detail. You can apply this technology in different sectors, including disaster assistance, geo-tagging, tracking and monitoring, supply chain management, and automotive navigation.
Table of Contents
