Localization. Why it’s one of the most critical aspects for your robot not to lose track.
So what are the different forms of localization? The first major challenge is the robot knowing where it is, using its own sensors, called self or ego-localization. The other is the robot knowing where it is, using external references, called landmark localization. Let’s look at ego-localization first. We list here the most critical ones used by most robots in typical logistics or manufacturing applications:
- Dead reckoning (odometer): this simplest form of localization uses the rotations of the wheels to estimate how the robot has moved. Each wheel has an encoder which is used to estimate the number of turns and whether it goes negative (backward) or positive (forward) it counts the distance the robot has made, the robot takes decisions based on what it thinks it is doing. Using dead reckoning is something like walking in a forest without a map. You may think you know how many steps you have taken but it’s not super accurate as you could still get lost. The issue with Dead Reckoning is that factors like wheel slip or encoder drift are not taken into account and estimates of where the robots are, are not always very reliable.
- Using an inertial sensor: In another approach to localization, an inertial sensor is used to correct Dead Reckoning by fusing it with the heading of the robot relative to the earth’s magnetic field and the robot’s linear and angular velocities (aka the gyroscope). Complex mathematical models fuse these various sources of information to provide an estimate of the position, speed and heading of the robot. This approach only helps a robot to know where it is relative to where it came from but it cannot help the robot understand where it is relative to its environment (e.g. a factory floor or in a house). So you still need to add a map so the robot can understand where it is and where it needs to go to.
- Map localization: So that brings us to a very popular approach called map localization. Here the robot uses the sensors and looks for the region around it, the robot has scanners and these scanners or lasers are scanning the region in a 360° mode, one can create a grid around it and the robot can figure out what it sees with the scanner. It knows there is a corner on the map, sees another corner and based on this data it can understand exactly where it is. A great way to find the exact location.
Besides these three forms of localization, robots can use landmark localization which is particularly useful in large complex environment.
The idea here is that certain unique objects (aka landmarks or beacons) whose positions in the world are known a priori, can be placed at specific locations. Anything that allows the robot to estimate the position and orientation of the object using its sensors can be used as landmarks.
For example, a robot can estimate the position of a reflector using its LIDARs or the position of a printed tag placed on the wall using a color camera. Even radio frequency techniques can be used. For example, low energy electronic Ultra-Wide Band (UWB) beacons, one placed on the robot and the rest in the robot’s world can help the robot to position itself even if the beacons are hidden or not in the line of sight of the robot’s sensors. There are other approaches too like Bluetooth direction finding introduced in Bluetooth 5.1.
Landmark localization can help robots localize in crowded or cluttered environments in combination with the robot’s own localization. As with many things in life, there is no one-size-fits-all. A sensor that works in one environment may not be suitable for another. So most robot vendors provide tried and tested approaches like these.
To understand how Tractonomy Robotics uses different forms of localization, contact us via this link and discover how the integration of the latest localization aspects can help your operation.