Robot motion path planning method
Source: - Edit: Chong Chak Time: 2020/3/10
The performance indicators of robot motion path planning include: real-time, safety, and reachability. In a dynamic environment, because the environmental information changes all the time, if the mobile robot has poor real-time performance and lags behind the dynamic environment, it may cause obstacle avoidance failure. Safety and accessibility are also important. A method with poor performance indicators will be eliminated even if it can make the mobile robot out of a perfect trajectory. Some methods do not have advanced theories, but the calculations are simple, real-time, and safe, and there is room for them. How to make performance indicators better is an important direction of various algorithm research.
Intelligent mobile robot path planning mainly includes fuzzy control, neural network, genetic algorithm and their mutual combination and other methods. Intelligent methods can simulate human experience, approach nonlinearity, have self-organization, self-learning functions, and have certain fault tolerance. The application of these methods to path planning will make mobile robots more flexible and intelligent in dynamic environments.
Multi-sensor information fusion of mobile robots is also a relatively active research field today. The information required by mobile robots for path planning in a dynamic environment is obtained from sensors. It is difficult for a single sensor to guarantee the accuracy and reliability of the input information. The information obtained by multiple sensors has redundancy, complementarity, real-time and low cost, and can quickly analyze the field environment in parallel. Specific methods include the weighted average method that uses the probability method to express information, Bayesian estimation method, multi-Bayesian method, Kalman filter method, and statistical decision theory method. There are DS evidence reasoning that uses propositional methods to express information, fuzzy logic, production rules, and information processing methods that imitate biological neural networks, artificial neural network methods, and so on.
The behavior-based path planning method is a bottom-up method of constructing a system, that is, the path planning is decomposed into a series of relatively independent small systems, in the running state, through the competition mechanism to obtain the control of the robot, and interact with the environment Finally reached the goal. The function/behavior-based robot control structure combines the advantages of the two, both the necessary rationality based on the function control structure, and the quick response based on the behavior control structure.
1 The classification of the mobile robot path planning technology is
based on the robot's recognition of the surrounding environment information and the degree of information mastery, as well as the recognition of different types of obstacles. The robot path planning can be divided into four categories: the first category, in the known In a familiar environment, the path of the mobile robot is planned according to the location of the static obstacle; the second type is to plan the path of the mobile robot according to the location of the static obstacle in the unknown and unfamiliar environment; the third type, In the known and relatively familiar environment, the path of the mobile robot is planned according to the operating state of the dynamic obstacle; the fourth category, in the unknown and relatively unfamiliar environment, the path of the mobile robot is determined according to the operating state of the dynamic obstacle Plan. According to the robot's ability to master the surrounding environment, the path planning technology can be divided. The first type is to plan the path of the mobile robot on the basis of the verification of the surrounding environment information, and the planned path is the global path; The second type is to plan the path of the robot based on sensor information, and the planned path is a local path [1]. The path planning methods of mobile robots can generally be divided into two types, namely traditional methods and intelligent methods.
2 Global path planning method
2.1 Topological method
The path planning method is mainly to divide the planned space and form subspaces with topological characteristics. At the same time, the topological network is constructed and the detailed topological path from the start point to the end point is explored. Then according to the path planning of the topological path, the final geometric path required is obtained. The topological path planning method is mainly based on the dimensionality reduction method, that is, the high-dimensional and more complex spatial geometric path finding method is transformed into a low-dimensional and relatively simple topological space identification and connection method. The most obvious advantage of this method is to make full use of the topological characteristics, thereby effectively reducing the actual search space [2]. The complexity of the algorithm is only related to the number of obstacles. Therefore, the main problem is that when the number of obstacles increases, reasonable measures should be taken to correct the existing topology network and achieve the improvement of graphics speed. purpose.
2.2 Visualization method
This method of path planning is to treat the robot as a point, then make a reasonable combination, and connect the robot to the target point and the vertices of the polygonal obstacle. In the process of connecting points, it is necessary to ensure that the straight line is visible, that is, the line between the target point and the vertices of the polygonal obstacles and the vertices of the obstacles cannot pass through the obstacles. In this way, the search for the optimal path can be effectively converted into the shortest distance of the visible straight line from the starting point to the target point. Using optimization algorithms can not only simplify the view, but also reduce the time required for searching. The biggest advantage of this method is that it can effectively shorten the path, but the size of the robot itself is ignored in the calculation. In actual operation, when the robot passes an obstacle, it is very easy to get too close to the obstacle or touch it. Will extend the search time. This situation is not uncontrollable. Appropriate application of tangent diagrams and Voronoi diagrams can improve the visualization method [3]. The tangent diagram uses arcs to represent the tangent of obstacles. The arc can be used to indicate that the robot can walk on the shortest path from the starting point to the target point, but will not touch the obstacle when it approaches the obstacle. However, this method also has certain shortcomings, that is, if the position setting deviation occurs during the control process, the robot will collide with the obstacle. The application principle of the Voronoi diagram is to represent the arc with a path away from the obstacle. This representation method will increase the distance of the path and the position error, but this method will reduce the probability of collision between the robot and the obstacle.
2.3 Grid method
The grid method divides the environment in which the robot runs, and divides the entire environment into grid units. In the space where the robot runs, the position and size of obstacles remain unchanged. In the actual work of the robot, there are no obstacles. Changes. The size of the grid is the same. The two-dimensional workspace of the robot is divided by the grid. If there are no obstacles in a grid, then the grid can be regarded as a free grid; if there are obstacles in a grid Objects, then this grid is an obstacle grid. In the grid, free space and obstacles are represented by the integration of grid blocks. There are two marking methods for obstacle grids and free grids: Cartesian coordinate method and serial number method. Under normal circumstances, the working environment of the robot is represented by a quadtree and an octree, and then an optimization algorithm is used to search for the optimal path. This method uses the grid as the calculation unit to record environmental information. The smaller the grid granularity, the more accurately the obstacle can be represented. However, this will occupy a larger storage space and also increase the search index of the algorithm range. However, if the grid granularity is too large, the path planning will not be accurate. From this point of view, the determination of the grid granularity is an important consideration of this method.
3 Local path planning method
3.1 Artificial potential field method
This method is a virtual force method, which simulates the movement of the mobile robot in the environment into the movement of an artificial force field. During this movement, the force generated between the obstacle and the mobile robot is regarded as repulsive force, and the force generated between the obstacle and the target point is regarded as gravitational force. When algorithms are used to generate potential around these two forces, the mobile robot will be subjected to abstract forces to achieve the purpose of walking around obstacles. The artificial potential field method has a simple internal structure and can better control the lower layers, so it is widely used in obstacle avoidance and smooth trajectory control. Because the local optimal solution always exists, the deadlock phenomenon frequently occurs, which causes the mobile robot to stop walking before reaching the local optimal point. In order to effectively solve the above problems, the algorithm needs to be improved. Reasonably optimize the definition of the potential field equation, which can effectively eliminate the local extremum in the potential field [4].
3.2 Genetic Algorithm
Genetic algorithms need to ensure that the fitness function must be positive, and does not require the function to be derivative or continuous. In addition, this method is a parallel algorithm, which enables its own implicit parallelism to be applied in the global search. Most optimization algorithms using genetic algorithms can perform single-point search, which makes it easy for calculations to enter the local optimum. But genetic algorithm is a multi-point search algorithm, so it is easy to calculate the global optimal solution.
3.3 Neural network method
Neural network method is to perform behavior space through perception space. However, it is difficult to use mathematical equations to express this mapping relationship. However, it can be expressed using neural network methods. Regarding sensor data as the network input, the direction of the person's expected movement can be determined as the network output. In this way, the original sample set can be represented by a set of data, and the final sample set can be obtained by processing the repeated and macrograph samples.
4 Future prospects of robot path planning technology The path planning technology of
mobile robots is mainly accomplished through computers, sensors and control technology. With the continuous advancement of technology, there will be more and more new algorithms, so the path planning technology of robots will also Will continue to be improved. However, in the future research process, it is necessary to continuously improve the performance index of its path planning and realize the path planning of multiple mobile robot systems. It is also necessary to appropriately incorporate multi-sensor information in path planning, focus more on path planning for mobile robots, and develop more calculation methods to promote the development and improvement of mobile robot path planning technology.
5 Conclusion
In summary, the mobile robot path planning technology has achieved considerable results, but there are still many deficiencies in its global path and local path planning methods, and methods that can be applied to all situations have not been developed. Therefore, in-depth research on its path planning technology is needed to promote the further development of this technology.