Search

CN-121973226-A - Mechanical arm path planning method based on multi-source track fusion and learning strategy

CN121973226ACN 121973226 ACN121973226 ACN 121973226ACN-121973226-A

Abstract

The invention relates to the technical field of robot path planning, in particular to a robot arm path planning method based on multi-source track fusion and learning strategies. The method firstly utilizes a plurality of traditional path planning algorithms to generate multi-source candidate tracks, and realizes unified representation of the tracks through path progress alignment and space resampling. And then constructing a multi-observation track fusion model based on uncertainty modeling, and recursively fusing the multi-source tracks by adopting a Kalman filtering method to obtain a stable and reliable fusion track. And further taking the fusion track as an expert demonstration data training path generation strategy, and optimizing the strategy by combining reinforcement learning, so as to realize autonomous generation and optimization of the mechanical arm path in a complex environment. The method and the system can fully utilize the information advantages of different planning algorithms, improve the stability and safety of path planning, improve the training efficiency and the environment adaptability of learning strategies, and are suitable for robot path planning and autonomous control tasks in complex environments.

Inventors

  • LI DAN
  • ZUO WENTONG
  • ZHANG JIANQIU

Assignees

  • 复旦大学

Dates

Publication Date
20260505
Application Date
20260312

Claims (10)

  1. 1. A mechanical arm path planning method based on a multi-source track fusion and learning strategy is characterized by comprising the following steps: acquiring initial pose, target pose and environmental obstacle information in a robot motion environment; generating a plurality of candidate paths by using a plurality of path planning algorithms based on the environmental information; The method comprises the steps of carrying out unified representation processing on paths generated by different path planning algorithms so as to eliminate the difference between the number of path points and sampling density; constructing a track fusion model based on multi-source path information, and carrying out fusion processing on the candidate paths to obtain a fusion track; constructing expert demonstration data by utilizing the fusion track, and training a path to generate a strategy model; and optimizing the path generation strategy model through a reinforcement learning method to obtain a final path planning strategy.
  2. 2. The method for planning a path of a mechanical arm based on a multi-source trajectory fusion and learning strategy according to claim 1, wherein the path planning algorithm comprises a path planning algorithm based on graph search and a path planning algorithm based on random sampling.
  3. 3. The method for planning a path of a manipulator based on a multi-source trajectory fusion and learning strategy of claim 1, wherein the unified representation process comprises path progress normalization and spatial resampling.
  4. 4. The method for planning a path of a mechanical arm based on a multi-source trajectory fusion and learning strategy according to claim 3, wherein the path progress normalization is realized by path arc length calculation, and the expression is as follows: Wherein the method comprises the steps of Indicating path execution progress.
  5. 5. The mechanical arm path planning method based on the multi-source track fusion and learning strategy according to claim 1, wherein the track fusion model adopts a multi-source track fusion method based on Kalman filtering.
  6. 6. The method for planning a path of a manipulator based on a multi-source trajectory fusion and learning strategy of claim 5, wherein different path sources are weighted by observing a noise covariance matrix during trajectory fusion.
  7. 7. The method for planning a path of a mechanical arm based on multi-source trajectory fusion and learning strategy according to claim 1, wherein the path generation strategy model is trained by simulating a learning method to learn a mapping relationship between a robot state and a path target point.
  8. 8. The method for planning the path of the mechanical arm based on the multi-source track fusion and learning strategy according to claim 1, wherein the path generation strategy model is continuously optimized by interacting with the environment through a reinforcement learning method.
  9. 9. An electronic device comprising one or more processors, storage means for storing one or more programs, wherein the programs, when executed by the processors, cause the processors to perform the method of any of claims 1-8.
  10. 10. A computer readable storage medium having stored thereon an executable program, characterized in that the program, when executed by a processor, causes the processor to perform the method according to any of claims 1 to 8.

Description

Mechanical arm path planning method based on multi-source track fusion and learning strategy Technical Field The invention belongs to the technical field of robot path planning, and particularly relates to a robot arm path planning method based on a multi-source track fusion and learning strategy. Background Along with the rapid development of intelligent manufacturing and robot technology, industrial mechanical arms are widely applied in the fields of electronic assembly, automatic production, logistics transportation and the like. In the actual task execution process, the mechanical arm needs to generate a safe, feasible and efficient motion path in a complex environment with obstacles, so that the path planning technology becomes a key problem in the robot system, and the performance of the path planning technology directly influences the working efficiency and the safety of the robot system. At present, the mechanical arm path planning method mainly comprises a path planning method based on a traditional algorithm and a learning-based method. The traditional path planning algorithm such as A, dijkstra and other graph searching methods can search feasible paths in discrete space, and have the advantage of higher stability, and the RRT, PRM and other sampling methods can search the feasible space in a random sampling mode, so that the method has better adaptability in a high-dimensional environment. However, a single planning algorithm generally has a fixed search strategy, and it is difficult to simultaneously consider multiple indexes such as path length, smoothness, safety and the like in a complex environment, and stability and adaptability of a planning result are limited to a certain extent. In recent years, learning methods such as simulated learning, reinforcement learning and the like are gradually applied to a robot path planning task, and a mapping relation from an environment state to a path decision is established through learning expert demonstration data, so that the autonomous planning capability of the robot in a complex environment is improved. However, this type of approach typically relies on high quality expert demonstration data, and the performance of the learning strategy is prone to degradation when the demonstration data is of insufficient quality or insufficiently distributed. Therefore, how to effectively utilize multi-source path information generated by various path planning algorithms and construct high-quality demonstration data while ensuring the path stability, thereby further improving the path planning capability of the robot and becoming an important problem in the current path planning research of the robot. The present invention has been made in view of this. Disclosure of Invention In order to solve the technical problems, the invention adopts the basic conception of the technical scheme that: Aiming at the problems that the existing mechanical arm path planning method has insufficient path stability, limited adaptive capacity of a single planning algorithm, high-quality demonstration data dependence of a learning method and the like in a complex environment, the invention provides a mechanical arm path planning method based on a multi-source track fusion and learning strategy. The method is used for optimizing by combining track information generated by various path planning algorithms and combining a learning strategy, so that the stability, robustness and environment adaptability of path planning are improved. In order to achieve the above purpose, the invention adopts the following technical scheme: the invention provides a mechanical arm path planning method based on a multi-source track fusion and learning strategy, which is characterized by comprising the following steps of. 1. Multi-source path generation In the environment of a given starting point and target point, multiple candidate paths are generated using multiple path planning algorithms. The path planning algorithm comprises a graph search-based method and a sampling-based method, and paths generated by different algorithms can be expressed as follows: Wherein the method comprises the steps of Represent the firstThe path generated by the individual planning algorithm is,Representing the position of the path point in three-dimensional space. 2. Path alignment and resampling Because the number of the path points and the sampling density generated by different path planning algorithms are different, the paths need to be uniformly represented. Firstly, normalizing the path progress by using the path arc length: Wherein the method comprises the steps of Representing the normalized progress of path execution. And then spatial resampling is carried out in a unified progress interval, and paths are uniformly represented as a path sequence with fixed length: so that paths generated by different planning algorithms remain structurally consistent. 3. Multi-source trajectory fusion After the path alignment and r