Mobile robot technology has a wide range of applications; it can be used in everything from flying and swimming to running on land. Examples include industrial robots such as material handling robots; commercial robots such as unmanned vehicles, drones, food delivery robots, and tour guide robots; and consumer robots such as the widely used robotic vacuum cleaner.
The core technologies of mobile robots revolve around three aspects: "perception", "decision-making" and "execution".
Key Technology 1:
SLAM (Simultaneous Localization and Mapping) refers to the process by which a moving object simultaneously calculates its own position and builds a map of its environment based on information from sensors. Currently, SLAM is primarily used in robotics, virtual reality, and augmented reality. Its applications include localization of the sensor itself, as well as subsequent path planning and scene understanding.
The implementation and complexity of SLAM vary greatly depending on the type of sensor and its installation method. Based on the sensor, SLAM is mainly divided into two categories: laser-based and vision-based. Laser-based SLAM has been researched for a longer period and is relatively mature in both theory and engineering. Vision-based solutions are currently used by a few manufacturers in actual products.
SLAM research has been underway for nearly thirty years since its inception in 1988. Early SLAM research focused on using filter theory to minimize noise in the pose of moving objects and landmarks on the map. After the turn of the 21st century, researchers began to borrow methods from SfM (Structure from Motion) to solve SLAM problems based on optimization theory. This approach has achieved considerable success and has become dominant in the field of visual SLAM.
Laser sensors: Laser sensors can directly obtain direct distance information relative to the environment, thereby achieving direct relative positioning. Absolute positioning and trajectory optimization of laser sensors can be performed based on relative positioning.
Visual sensors: Visual sensors struggle to directly obtain direct distance information relative to the environment. Instead, they must estimate their own pose changes using two or more frames of images, and then calculate their current position by accumulating these pose changes. This method is more similar to direct odometry for localization, i.e., visual odometry. The odometry measurements, after integration, are equivalent to the localization information directly obtained by a laser sensor; this is the front end of the graph-optimized SLAM framework. The back end's optimization of localization and pose trajectory is essentially the same as that of laser sensor optimization, both based on the theoretical framework of optimal estimation.
Key Technology Two: Planning.
Planning includes path planning and motion planning. The technologies related to planning are relatively mature.
Commonly used path planning algorithms for mobile robots include A* and D*; commonly used motion planning algorithms include PID, VFF, DWA, and PTG.
Key technology three: control.
Key Technology 4: Structural Design and Hardware Design.
(Original title: Four Fundamental Technologies for Autonomous Mobile Robots)
