«——[Introduction·] ——»
In modern society, robots have been studied in order to eliminate human labor, allowing machines to perform physical tasks that previously required human labor. However, people not only need machines to be able to perform physical tasks, but also to act like humans with thinking and decision-making, and in order for machines to become intelligent, AI knowledge becomes very important.
Path planning has always been considered the most common problem in robot navigation, and the robot must move from the starting position to the target position by avoiding obstacles. Things like miniature air cars, action robots, wall-climbing robots and underwater robots have all been tested with different algorithms.
«——[·Global Navigation Method.]——»
Navigation can be divided into two types: global navigation and local navigation, for the global navigation type, you need to know the environment a priori, also known as offline mode path planning.
For local navigation also known as online mode path planning, the robot determines its position and direction, and can use externally equipped sensors for motion control, such as infrared sensors, ultrasonic sensors, lidar and vision sensors, etc., which can automatically correct the robot's orientation through software.
Other techniques include neural networks, fuzzy logic, neural fuzziness, particle swarm optimization, genetic algorithms, and ant colony optimization, which are all algorithms used by researchers that have been successfully applied.
Artificial Potential Field (APF) is a technology inspired by nature, the basic idea of APF is to fill the robot environment with artificial potential field, in which obstacles are repelled by repulsion and the robot is attracted to the target by gravity.
The potential field depends on two forces, gravity and repulsion, the target produces gravitational pull on the robot and the obstacle produces a repulsive force, which is inversely proportional to the distance from the robot to the obstacle, pointing towards the obstacle.
In the potential field method, a gravitational field is created to reach the goal, and the potential field is usually defined throughout free space, and at each time step, the potential field is calculated at the robot position and the induced force applied by the field is calculated.
The main problem with APF is that the robot can fall into a local or global minimum problem, where the robot is trapped at two points where the forces cancel each other out, not allowing the robot to move further or even retreat.
The graphical search method is considered to be the easiest way for robots to find paths, it is a well-defined, efficient and less computationally complex method to find non-obstructive paths in a shorter time and with less computational complexity.
After building an environment for your bot, you can easily reach your goal by connecting paths, and the process continues until you get from node to node for a better and better solution.
When the robot reaches the desired target, the robot can continue to travel to the new location. The Dijkstra algorithm is considered to be a graphical search method for solving the optimal path problem with non-negative edge path cost, which can produce the shortest path, which is used for path cost queries from a single point to a single destination.
This algorithm is important in traffic information systems to track sources and destinations by tracking sources, and it is used to determine the relationship between the initial node, the shortest distance from other nodes in the diagram, and low cost. The main key of the algorithm is to repeatedly calculate the shortest distance from the initial point to the end point, while excluding longer distance paths.
A* is a search algorithm that can also be used to find paths, which constantly searches for unexplored locations in a graph.
When the target location is reached, the algorithm stops. If the goal is not achieved, all neighbors are set to the shortest path to be searched, and the A* algorithm is widely used in games for pathfinding.
Local navigation method In navigation technology, sensors are commonly used to control the direction and position of the robot, and LIDAR sensors are frequently used for automation purposes.
LIDAR can work independently compared to GPS systems, so it has the ability to map the environment. LIDAR can be used standalone, but when combined with other sensors, improved results are obtained.
For example, working with cameras can provide a powerful positioning tool that can be leveraged to map the local environment to locate and identify landmark locations, often referred to as SLAM.
With this technology, mobile robots can automatically correct their position and orientation. The use of motor encoder sensors in combination with LIDAR can also improve accuracy.
Vector field histogram (VFH) is another local navigation method for solving mobile robot path planning problems, and VFH's idea is based on the virtual force field approach. As the name suggests, it is a field, so an obstacle found within a certain distance from the vehicle exerts a repulsive force that moves the vehicle away from the obstacle and draws the vehicle to the target point using attraction.
VFH uses a deterministic grid similar to a radar screen, where obstacles found by the sensor are counted in a deterministic grid of corresponding coordinates. This means that a higher certainty value indicates that a real object has been detected within the sensor range. In real time, the mesh is constantly updated at every instant, so this method is suitable for sparse moving objects.
Similarly, autonomous guided vehicle systems (AGVs) can also be divided into different categories, and AGVs can be divided into two methods according to the number of loading units it can carry at the same time, i.e. single loading units or multiple loading units. The load is treated as a single unit, which brings the vehicle from the starting point to the unloading point.
In a single unit loading system, an idle vehicle is selected to perform the task of delivering the goods to the specified destination, and the vehicle then departs from its initial location to the pickup location to obtain the goods and returns to its unloading destination.
During the unloading of assigned tasks, the vehicle is not interrupted by any other tasks. In a multi-unit loading system, the loaded vehicle is interrupted during the ongoing task to pick up other additional loads.
By assigning vehicles to different tasks, the load of the task in progress and other additional loads carried by the vehicle are affected. Therefore, scheduling and scheduling function parameters are introduced in the controller to determine the allocation of vehicles to carry loads.
The path navigation of AGVs can be divided into static path determination and dynamic path determination, in which the vehicle travels between the start and end points using a predefined path.
In direct guidance systems, where embedded wires, tapes, radio frequency identification (RFID) chips, and deadband calculations are used to guide the vehicle, static paths can also be divided into two categories: unidirectional and bidirectional.
In a one-way system, the vehicle can only travel in one direction, while in a two-way system, the vehicle can travel in any direction; This is achieved through the use of U-turn points or two-way vehicles that can move forward or backward.
In a dynamic routing system, the vehicle autonomously determines its path by detecting and avoiding obstacles. In this system, the vehicle knows its destination through certain coordinate systems, and the vehicle uses an internal navigation system to reach the destination.
«——[Classification of navigation systems.] ——»
Wired navigation uses slots or wires cut and placed below the surface, and sensors are mounted on the bottom of the AGV that detect the location of the radio signal emitted to the wire, and this information is used to help the control circuit of the AGV follow the path.
The guided sensor vehicle follows the path of tape or line drawing with the help of a camera, the information is transmitted by radio communication, and the advantage of the guidance path is that it can be relocated and deleted anywhere.
Laser navigation is considered the most effective obstacle avoidance and path tracking technology, and it does not require any wires, tracks, and tracks to make motion. The beam sent and received from the sensor, the time it takes for the beam to propagate and return helps determine the distance and angle, which in turn helps the vehicle to move.
Navigation can be achieved by assigning and guiding tasks to the vehicle through the computer control system. For this purpose, base stations buried under the floor are used to help the vehicle verify its position and orientation.
This navigation is performed by gyro sensors, and the combination of laser sensors and gyroscope measurements provides another method of distance measurement, which is an efficient way to determine the shortest path allowed to path.
Vision-based AGVs use cameras to acquire environmental features and make decisions based on these features to navigate vehicles, while geo-guidance understands its environment through the use of location.
It uses a fixed reference point to identify any product in the warehouse and uses this to help the vehicle navigate, and in the case of automated guided vehicles, the management of vehicle tasks is also very important.
Fuzzy logic is of great significance in mobile robots and automated guided vehicle control, where they are best suited for inaccurate and imprecise information in sensor measurements and heuristic knowledge.
They can be combined with AGV systems to control motor movement and steering. Since fuzzy logic is a rule-based system, different rules can be created to drive AGVs.
During the fuzzing process, sensors on the AGV acquire the position of vehicles and obstacles, and then create a library of obstacle avoidance and path following rules, which are combined with fuzzy inference systems to obtain the desired direction of AGV motion.
Similarly, neural networks can be used in a similar way for AGV systems, which rely on biological nervous systems, such as the brain, to process information, which consists of artificial neurons that are interconnected and work together to produce a specific output.
Learning biological systems involves adjusting the highly interconnected synaptic connections present between neurons to adapt them to a given dataset. NGOs have a tendency to learn from their surroundings and are designed to learn experientially to improve their efficiency, while they are also well suited to environmental fluctuations.
These different navigation methods are used in conjunction with fuzzy logic controllers and neural networks to achieve better position and direction control and efficient results.
«——[Kinematics and dynamic analysis of mobile robots.] ——»
Over the past decade, the issue of autonomous vehicle control has been extensively studied and deeply investigated. Various soft computing techniques have been studied to develop and design autonomous mobile robots that guide mobile robots to target points and avoid obstacles through fuzzy logic methods.
A neural fuzzy controller was created for the incomplete differential drive robot for its navigation purposes, and to read the obstacle distance, four infrared sensors were used, and their distance information was fed into the controller to maintain the speed of the motor.
Two fuzzy controllers combined with vehicle control direction assist the driver on the highway, sometimes playing the role of the driver to reduce accidents. It uses a CMOS sensor as a path recognition device to draw the path of the centerline through image processing.
Over the past few years, researchers have proposed different soft computing techniques to solve obstacle avoidance problems and robot navigation in different environments.
The introduction of fuzzy set theory was first proposed by Professor Lofty Zadeh of the University of California in 1965. Fuzzy logic theory has a wide range of applications in signal and information processing, control engineering, pattern recognition, decision making, and expert systems.
The system produces highly efficient independent intelligent systems, fuzzy logical systems that are naturally inspired by human reasoning and whose work is based on perception. In fuzzy logic, a clear set of input values accumulates in the fuzzy step and together is transformed into a set of fuzzy sets with a set of inference rules.
During defuzzification, eight rule-based fuzzy controllers are designed for mobile robots using member functions to transform the resulting output into a set of clear sets, for path tracking and obstacle avoidance.
To achieve obstacle avoidance and robot navigation, the Takagi-Sugeon fuzzy controller based on the gradient method is used to adjust, with different member function parameters to obtain the best results for navigation purposes.
And the ease of use of NNs and the tendency to expand learning and generalization capabilities make them a popular method that is popular in practical applications. Problems such as facial expressions, handwritten letter recognition, finding the shortest route for travel, and making a logistics plan can only be solved by using NN.
«——[·Conclusion·] ——»
Robot navigation technology is an important research direction in the field of robotics, which has a wide range of applications in industrial manufacturing, service robots, military and other fields.
With the continuous development of computer technology and the improvement of computing power, robot navigation technology is also constantly evolving, and gradually becomes a complex system with a complete closed-loop from sensing technology to control technology, decision-making technology to planning technology.
In the future, with the rapid development of artificial intelligence technology, robot navigation technology will be further evolved and improved, bringing more application scenarios and economic benefits to all walks of life.
Bibliography:
1.Contreras-Cruz, M. A., Ayala-Ramirez, V., & Hernandez-Belmonte, U. (2015).
2.Guan-Zheng, T., Huan, J., & Sloman, M. (2007). Path planning for mobile robots using ant colony system and Dijkstra's algorithm.
3.Purian, R., & Sadeghian, A. (2013). A fuzzy controller and ant colony optimization algorithm to solve the optimal path problem of a mobile robot in an unknown dynamic environment.
4.Ganapathy, S., Jie, L., & Parasuraman, S. (2010). Ant colony optimization-based navigation algorithms for mobile robots.
5.Meier, L., Tullumi, E., Stauffer, A., Dornberger, R., & Hanne, T. (2017). Backup path planning using dispersed information with ant colony optimization.