Algorithm 1: Depth Information Extraction
The principle is to use two parallel cameras to triangulate each point in space. By matching the positions of the imaging points in the left and right cameras, the distance of the corresponding 3D point in space is calculated.
The robot wants to obtain the three-dimensional coordinates of the target through several images. A more important task in binocular vision technology is to perform image matching. First, the matching points of the object in the left and right images are identified, and then the disparity and depth information of each point are obtained.
Binocular stereo vision has advantages such as simple and inexpensive equipment, high accuracy and speed, and the ability to calculate distance and depth information without contacting objects. It has important applications in drone power line inspection and industrial construction robots.
Algorithm 2: Positioning and Navigation
Robot navigation is a relatively complex system, involving the following technologies:
• Visual odometry (VO)
• Mapping, utilizing VO and depth maps;
• Relocation: Identifying the current location from a known map;
• Closed-loop detection eliminates VO (Vibration Detection) closure-loop error;
• Global navigation;
• Visual obstacle avoidance;
• Scene tagging: Identifies objects in a room and adds tags to them.
Simply put, it involves optical processing of the robot's surrounding environment. First, a camera is used to collect image information, which is then compressed and fed back to a learning subsystem composed of neural networks and statistical methods. The learning subsystem then links the collected image information with the robot's actual position to complete the robot's autonomous navigation and positioning function.
This method, known as SLAM (Simultaneous Localization and Mapping), best reflects the intelligence level of mobile robots. The ability to simultaneously map and locate is widely considered a key prerequisite for robots to achieve autonomy.
Currently, commonly used SLAM technologies are mainly divided into two categories: one is VSLAM based on visual sensors, and the other is laser SLAM based on laser sensors.
Visual SLAM specifically refers to using depth cameras such as cameras and Kinect for indoor navigation and exploration. To date, indoor visual SLAM is still in the research stage and far from being practically applied. In contrast, laser SLAM technology is relatively mature and is currently the most stable, reliable, and high-performance SLAM method.
Algorithm 3: Obstacle Avoidance
Navigation addresses the problem of guiding a robot towards a target. When a robot lacks a map, the method used to approach the target is called visual obstacle avoidance. Obstacle avoidance algorithms solve the problem of avoiding static and dynamic obstacles based on data from visual sensors, while still maintaining movement towards the target, enabling real-time autonomous navigation.
There are many obstacle avoidance algorithms. Traditional navigation obstacle avoidance methods such as the visibility method, grid method, and free space method can handle obstacle avoidance problems when obstacle information is known. However, when obstacle information is unknown or the obstacle is movable, traditional navigation methods generally cannot solve the obstacle avoidance problem well or cannot avoid obstacles at all.
In real life, in most cases, the environment in which robots operate is dynamic, variable, and unknown. To solve these problems, people have introduced algorithms from fields such as computer science and artificial intelligence.
Meanwhile, thanks to the improvement of processor computing power and the development of sensor technology, it has become easier to perform some complex algorithm calculations on the mobile robot platform, which has led to a series of intelligent obstacle avoidance methods, among which the more popular ones include: genetic algorithm, neural network algorithm, and fuzzy algorithm.
