CN115723129A - Mechanical arm continuous operation motion planning method - Google Patents
Mechanical arm continuous operation motion planning method Download PDFInfo
- Publication number
- CN115723129A CN115723129A CN202211437040.3A CN202211437040A CN115723129A CN 115723129 A CN115723129 A CN 115723129A CN 202211437040 A CN202211437040 A CN 202211437040A CN 115723129 A CN115723129 A CN 115723129A
- Authority
- CN
- China
- Prior art keywords
- mechanical arm
- path
- planning
- continuous operation
- algorithm
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 32
- 238000005070 sampling Methods 0.000 claims abstract description 50
- 239000000203 mixture Substances 0.000 claims abstract description 35
- 238000012549 training Methods 0.000 claims abstract description 24
- 238000009826 distribution Methods 0.000 claims abstract description 18
- 238000010845 search algorithm Methods 0.000 claims abstract description 8
- 238000001514 detection method Methods 0.000 claims description 15
- 238000009827 uniform distribution Methods 0.000 claims description 11
- 238000013138 pruning Methods 0.000 claims description 8
- 230000000750 progressive effect Effects 0.000 claims description 7
- 238000012545 processing Methods 0.000 claims description 6
- 238000009499 grossing Methods 0.000 claims description 5
- 230000000007 visual effect Effects 0.000 claims description 5
- 239000012636 effector Substances 0.000 abstract description 5
- 238000013507 mapping Methods 0.000 abstract description 2
- 238000010276 construction Methods 0.000 abstract 1
- 238000010586 diagram Methods 0.000 description 6
- 230000003044 adaptive effect Effects 0.000 description 3
- 230000001934 delay Effects 0.000 description 2
- 238000003064 k means clustering Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 230000036544 posture Effects 0.000 description 2
- 238000004904 shortening Methods 0.000 description 2
- 238000012952 Resampling Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
Images
Landscapes
- Manipulator (AREA)
Abstract
The invention discloses a planning method for continuous operation and movement of a mechanical arm. The method comprises the following steps: and solving the inverse kinematics of the mechanical arm according to the D-H model of the mechanical arm to realize the mapping from the pose of the end effector of the mechanical arm to the joint of the mechanical arm. Teaching continuous operation tasks by using an expert planner, collecting teaching tracks and discretizing the teaching tracks into a series of path points, wherein the path points are used as training set data; learning the distribution of training set data through a Gaussian mixture model to realize the construction of the self-adaptive sampler; exploring the configuration space of the mechanical arm by adopting a lazy probability road map, and sampling by utilizing a constructed self-adaptive sampler; based on the constructed undirected graph, a graph search algorithm is adopted to query the path, whether the path is collided or not is detected, if so, collision points are removed, and candidate paths are searched. The method can accelerate the path solving speed and improve the efficiency of the mechanical arm continuous operation motion planning.
Description
Technical Field
The invention belongs to the technical field of motion planning of mechanical arms, and particularly relates to a continuous operation motion planning method of a mechanical arm.
Background
With the continuous development of science and technology in recent years, mechanical arms are widely applied to the fields of life, production and the like and used for executing continuous operation tasks such as article taking and placing, part assembling and the like. This also imposes increasingly stringent demands on the precision and efficiency of the operation of the robot arm. For the mechanical arm to complete such operation process, the mechanical arm is often required to accurately and quickly reach a specified position, then grab an object and place the object at other positions, and the mechanical arm needs to avoid collision with the environment and the mechanical arm. The paths of the mechanical arm for completing the planning task are many, the planning time is spent as little as possible, a short enough path close to the optimal path is found, and the path is always the target of the mechanical arm motion planning.
The traditional motion planning algorithm comprises an artificial potential field method, an ant colony algorithm, an A-star algorithm and the like, and most of the algorithms are low in convergence speed, are easy to fall into local extreme values and are not easy to expand to a high-dimensional space such as a multi-axis mechanical arm. The motion planning algorithm based on sampling can effectively solve the problem of motion planning in a high-dimensional space, and does not need to accurately model the environment. The PRM algorithm is a sampling-based motion planning algorithm, exploration of a mechanical arm configuration space is realized by constructing an undirected graph, then the shortest path can be effectively solved by adopting the graph search-based algorithm, and the algorithm has the property of multi-query. It is worth noting that when the working space changes, such as the movement of an object, the introduction of new obstacles, which make the constructed good graph structure useless. The lazy prm algorithm delays collision detection to a path solving stage, can effectively cope with environmental changes, and improves the composition speed, but the lazy prm algorithm globally adopts uniform distribution for sampling, so that the path planning speed is low, and the path quality is not high enough.
Disclosure of Invention
The method for planning the continuous operation motion of the mechanical arm by introducing the teaching prior can accelerate the path planning speed of the mechanical arm facing to the continuous operation task and improve the path quality.
A method for planning continuous operation movement of a mechanical arm comprises the following steps:
step 1.1, modeling the geometric relation of each joint of the mechanical arm by adopting a D-H method, establishing a connecting rod coordinate system of the mechanical arm, and calculating and obtaining a D-H model of the mechanical arm according to the D-H parameters of the mechanical arm;
and step 1.2, according to the D-H model of the mechanical arm constructed in the step 1.1, analyzing and calculating inverse kinematics of the mechanical arm, and mapping the pose of the end effector of the mechanical arm to the joint space of the mechanical arm by using the inverse kinematics.
And 1.3, detecting the number of objects in the working space and the position and posture of the objects through a vision sensor device. Modeling a continuous operational task T, where T i For a single task in T, it involves an object O detected by a vision sensor i And the initial grabbing pose P of the end effector of the mechanical arm start And placing pose P goal Where i is the number of objects detected by the vision sensor. Initial pose P of mechanical arm when operating object by inverse kinematics start With object pose P goal Resolving into starting poses and target poses in a plurality of groups of mechanical arm joint spaces;
step 1.4, taking the initial position of the mechanical arm in the step 1.3 as the initial position of a joint space of the mechanical arm, taking the target position of the mechanical arm movement as the target position of a configuration space, teaching a planning task by adopting a progressive optimal movement planner, solving a corresponding teaching track, collecting the teaching track solved by the planner, and discretizing the teaching track to be taken as training set data;
and step 1.5, initializing parameters of the Gaussian mixture model by using a K-means algorithm according to the training set data in the step 1.4, wherein the number of Gaussian components is selected according to AIC, and then iteratively training the Gaussian mixture model until the model parameters are converged.
And step 1.6, generating sample points by combining the Gaussian mixture distribution trained in the step 1.5 with the uniform distribution, and constructing an undirected graph data structure by adopting a lazy probability road graph method.
And step 1.7, based on the undirected graph constructed in the step 1.6, solving the continuous planning task by adopting a graph search algorithm, performing collision detection on the solved path, inquiring a candidate path if collision occurs, then performing smooth processing on the path by adopting a pruning algorithm, and realizing continuous operation motion planning of the mechanical arm by utilizing a series of joint control commands contained in the path.
Step 1.2, calculating inverse kinematics by using the D-H model of the mechanical arm established in the step 1.1, and providing an initial point and a target point for subsequent continuous motion planning of the mechanical arm; continuous operation tasks require the mechanical arm to plan to the initial position of an object, then grab the object and place the object to the target position, so the grabbing and placing poses at the tail end of the mechanical arm are resolved into corresponding initial joint angles and target joint angles by adopting the inverse kinematics of the mechanical arm, and the joint angles are used as the initial point and the target point of the planning problem.
Step 1.3 needs modeling of continuous operation tasks of the mechanical arm, wherein the modeling is completed through a visual sensor device, and the visual sensor is used for detecting the number of objects and the pose of the objects.
Step 1.4, teaching a continuous operation task of the mechanical arm, and collecting a taught motion trail as training set data; the teaching track is obtained by solving a progressive optimal motion planning algorithm, and can be called through a motion planning open source library (OMPL), and meanwhile, the teaching track is discretized and the configuration is extracted to establish a training set.
Step 1.5, training a Gaussian mixture model by using the collected teaching data to provide a basis for offset sampling of a subsequent motion planner; because the training of the Gaussian mixture model is easy to fall into a local extreme value, the training data are clustered by adopting a K-means algorithm, the initialization of the parameters of the Gaussian mixture model is realized, the number of Gaussian components is selected according to the AIC criterion, and finally the parameters of the model are iteratively estimated by utilizing an EM algorithm.
Step 1.6, generating sampling points by utilizing the trained Gaussian mixture distribution and uniform distribution so as to explore the configuration space of the mechanical arm; firstly, setting a sampling threshold value, firstly, sampling a random number in 0-1 distribution, comparing the random number with the sampling threshold value, and when the random number is less than or equal to the sampling threshold value, setting the sampling threshold value to be 0.5; biased sampling is carried out by utilizing Gaussian mixture distribution, and when the random number is larger than a sampling threshold value, sample points are generated by adopting uniform distribution, so that the generalization of sampling is improved; setting the initial sampling number and the number of nearest neighbor nodes, constructing an undirected graph by using a lazy probability roadmap algorithm, and delaying the collision detection of the nodes and the edges to a path solving stage.
Step 1.7, solving a continuous planning problem by using the constructed undirected graph structure, giving a group of mechanical arm initial configuration and target configuration, solving a feasible path by using a graph search algorithm, performing collision detection on the path, and deleting failed nodes and edges and searching candidate paths if the path has nodes and edges which collide; carrying out smoothing treatment on the solved collision-free path by adopting pruning calculation, removing redundant nodes in the path, and further optimizing the path length; and directly calling the constructed undirected graph structure in the next planning process, and solving the path of the motion planning by using an A-star algorithm.
The invention utilizes the visual sensor equipment to detect the number and the pose of the objects to be operated in the working space, builds the modeling of the continuous operation tasks of the mechanical arm on the basis of the number and the pose, and achieves the motion control of the mechanical arm from the continuous operation tasks so as to achieve the purposes of accelerating the path planning speed and improving the path quality. Aiming at continuous operation tasks of a mechanical arm, firstly, an inverse kinematics model of the mechanical arm is constructed according to D-H parameters of the mechanical arm, and then the continuous operation tasks are mapped into a plurality of groups of starting and target joint configurations through inverse kinematics to be used as starting points and end points of a planning problem. And then calling a progressive optimal motion planning algorithm in an open source motion planning library (OMPL) as an expert planner, teaching a planning task by using the expert planner, collecting a corresponding teaching track, discretizing the teaching track to be used as training data of a Gaussian mixture model, initializing parameters of an EM algorithm by using a k-means clustering algorithm, introducing an AIC criterion, and selecting the number of Gaussian components according to the criterion. After the model is trained, the converged Gaussian mixture distribution is used as an adaptive sampler for biased sampling. Setting a sampling threshold, sampling a random number in 0-1 distribution before sampling each time, comparing the probability value with the sampling threshold, performing biased sampling by using Gaussian mixed distribution when the probability value is less than or equal to the sampling threshold, and generating sample points by adopting uniform distribution when the probability value is greater than the sampling threshold, thereby realizing balanced exploration and utilization. Then setting the initial sampling number and the number of nearest neighbor nodes, constructing an undirected graph structure by using a LazyPRM algorithm, storing the nodes and edges which collide with the environment in the data structure, inquiring the shortest path by using an A-graph search algorithm, performing collision detection on the path, removing invalid nodes and edges and searching candidate paths if the path collides with the environment, and then performing smoothing processing on the path, removing redundant path points and further shortening the length of the path. In the subsequent planning problem, the constructed undirected graph is directly used for searching the path without resampling and composition, so that the continuous planning of the mechanical arm is realized.
Drawings
FIG. 1 is a schematic view of a robot arm continuous operation motion planning
FIG. 2 is a flow chart of continuous planning of a robotic arm incorporating teaching priors
FIG. 3 is a schematic diagram of learning the distribution of teaching data and constructing an adaptive sampler
FIG. 4 is a flow chart of a robot arm continuous operation motion planning solution
FIG. 5 is a schematic diagram of path collision detection and path smoothing
FIG. 6 is a diagram of a collision-free path planned in a 2D environment according to the present invention
FIG. 7 is a schematic diagram of a robotic arm employing the present invention to accomplish a continuous operation motion planning task
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all embodiments. All other embodiments, which can be obtained by a person skilled in the art without making any creative effort based on the embodiments of the present invention, belong to the protection scope of the present invention.
As shown in fig. 1, the method of the present invention is schematically illustrated, a motion planning method of teaching learning is adopted, and an expert planner is used to teach continuous operation tasks of a mechanical arm, so as to serve as prior knowledge, thereby achieving the purposes of improving subsequent motion planning efficiency, improving planning path quality, and guiding the mechanical arm to complete continuous operation tasks efficiently without collision. Fig. 2 contains the overall process of the method, which mainly includes: and collecting teaching data as prior knowledge and solving a mechanical arm path. Firstly, the kinematics of the mechanical arm is analyzed, then a progressive optimization algorithm in an OMPL library is called as an expert planner, a planning task is taught and teaching data is collected, and a Gaussian mixture model is trained by utilizing the teaching data. And then, sampling is combined by utilizing Gaussian mixture distribution and uniform distribution, an undirected graph structure is constructed by utilizing a LazyPRM algorithm, the undirected graph data structure comprises nodes and edges, sampling points are used as the nodes of the undirected graph, the nodes are connected with the nodes according to the nearest neighbor principle to form the edges of the undirected graph, and collision detection is not performed on the nodes and the edges in the graph in the composition stage. On the basis, a graph search algorithm-A is adopted to solve a collision-free path, and finally a pruning algorithm is utilized to carry out smooth processing on the path, wherein the specific implementation mode of the pruning algorithm is that collision detection is carried out on the solved path, if the path collides with the environment, invalid nodes and edges are removed, candidate paths are searched, then the path is subjected to smooth processing, redundant path points are removed, and therefore the purpose of further shortening the length of the path is achieved.
Modeling the geometric relation of each joint of the mechanical arm by adopting a D-H method, establishing a connecting rod coordinate system of the mechanical arm, and calculating and obtaining a D-H model of the mechanical arm according to the D-H parameters of the mechanical arm;
the joint space planning of the mechanical arm is more efficient than the Cartesian space planning, when the mechanical arm grabs and places an object, the grabbing and placing postures are relative to a Cartesian coordinate system, the object needs to be mapped to the joint space to be more efficiently planned, and D-H modeling needs to be firstly adopted for the machine to realize the step.
Analyzing and calculating inverse kinematics of the mechanical arm according to the constructed D-H model of the mechanical arm, and resolving continuous operation tasks into a plurality of groups of mechanical arm initial positions and target positions by using the inverse kinematics of the mechanical arm;
the constructed D-H model of the mechanical arm is used for analyzing and calculating the inverse kinematics of the mechanical arm, continuous operation tasks of the mechanical arm on an object usually comprise a plurality of grabbing and placing operations, and grabbing and placing poses at the tail end of the mechanical arm are mapped into joint space through the inverse kinematics, so that a series of grabbing and placing joint configurations of the mechanical arm are obtained and are used as a planned starting point and a planned target point.
Taking the initial position of the mechanical arm as the initial position of a mechanical arm configuration space, taking the target position of the mechanical arm movement as the target position of the configuration space, calling a progressive optimal movement planner in an OMPL (open unified planning phase) library as an expert movement planner, teaching a planning task, collecting a teaching track solved by the planner, and discretizing the teaching track to be used as training set data;
the quality of the teaching track affects the training effect of a subsequent Gaussian mixture model, and the track planned by the progressively optimal mechanical arm motion planning algorithm is close to the optimal track, so that the progressively optimal motion planning algorithm in the OMPL library is called as an expert planner to teach the operation motion planning task of the mechanical arm. The teaching of the expert planner and the collection of teaching tracks are carried out off line, and the data are subsequently utilized as training data of a Gaussian mixture model, so that the teaching prior is introduced.
Training a Gaussian mixture model by using the collected teaching data to provide a basis for offset sampling of a subsequent motion planner; the training of the Gaussian mixture model is easy to fall into a local extreme value, so that the training data are clustered by adopting a K-means algorithm, the initialization of parameters of the Gaussian mixture model is realized, the number of Gaussian components is selected according to an AIC (advanced information center) criterion, finally, the parameters of the model are iteratively estimated by utilizing an EM (effective electromagnetic modeling) algorithm, the maximum iteration frequency is set to be 100, and when the EM algorithm finishes iteration, the Gaussian mixture model is considered to be converged.
Due to the fact that the topological structure of the configuration space of the mechanical arm is complex, a Gaussian mixture model is used for modeling a collision-free track in the joint space of the mechanical arm, and the configuration of a teaching track is extracted through a probability method. We iteratively estimate the parameters of the gaussian mixture model using an EM algorithm that can maximize the likelihood that previous motion planning solutions were sampled from the distribution. Then, initializing parameters of the EM algorithm by adopting a k-means clustering algorithm, introducing an AIC criterion, and selecting the number of Gaussian components according to the criterion. Sampling from the trained model creates sample points at task-dependent locations to bias sampling. The process of collecting teaching data and training a gaussian mixture model to construct an adaptive sampler is shown in fig. 3.
Generating sampling points by utilizing the trained Gaussian mixture distribution and uniform distribution so as to explore the configuration space of the mechanical arm; firstly, setting a sampling threshold value to be 0.5, sampling a random number p in 0-1 distribution before sampling each time, comparing the random number p with the sampling threshold value, performing biased sampling by utilizing Gaussian mixture distribution when p is less than or equal to the sampling threshold value, generating sample points by adopting uniform distribution when the random number p is greater than the sampling threshold value, generalizing introduced teaching priori knowledge in such a way, and still effectively planning when a new task instance is given. And then setting the number of initial sampling points and the number of nearest neighbor nodes, and constructing an undirected graph by using a lazy probability roadmap algorithm, wherein the algorithm delays collision detection to a path query stage, and is still effective when a working space is partially changed.
Detecting the number and pose of objects in a working space by a vision sensor device and modeling a continuous operational task T, where T i Is a subtask in T, which includes the object O detected by the vision sensor i And initial grabbing pose P of mechanical arm end effector start And the placing pose P goal Where i is the number of objects detected by the vision sensor. Then, each operation task t is processed by inverse kinematics i Resolving the pose of the end effector into an initial point x in the joint space of the mechanical arm start To the target point x goal Given a starting point x based on a lazy probabilistic road map biased towards sampling start To the eyesPunctuation x goal Searching for the shortest path by using A-graph search algorithm, as shown in FIG. 4, the continuous operation task T comprises a plurality of sub-tasks T i Each subtask corresponds to an object to be operated O i ，P start Are respectively P goal The initial pose and the final pose of the mechanical arm when the mechanical arm operates an object are mapped into initial and target joint configurations through inverse kinematics, and finally path search is carried out through a probability road map based on deflection sampling.
As shown in fig. 5, when a new obstacle appears in the environment, the method performs collision detection on the path by calling a flexible collision detection library (FCL), can remove failed nodes and edges in the lazy probability path graph, then queries a candidate path by using a graph search algorithm, and finally performs smoothing processing on the path by using a pruning algorithm to remove redundant nodes in the path. The specific implementation manner of the pruning algorithm is to connect any two nodes, and if the nodes are not collided, remove the redundant node between the two nodes, thereby further optimizing the path length. Fig. 6 shows a path diagram obtained by planning in a 2D environment according to the present invention.
Aiming at the planned collision-free path, a trajectory optimization algorithm in a motion planning open source library (OMPL) is adopted for trajectory planning, so that continuous operation motion control of the mechanical arm is realized, a composite diagram of continuous motion planning of the mechanical arm during a shelf stacking task is shown in FIG. 7, the task comprises a plurality of grabbing and placing subtasks, and when a new obstacle appears in the environment, the mechanical arm can still carry out efficient planning and generate a collision-free trajectory.
For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and for relevant points, reference may be made to the description of the method section.
The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the above, the present disclosure should not be construed as limiting the invention.
Claims (6)
1. A mechanical arm continuous operation motion planning method is characterized by comprising the following steps:
step 1.1, modeling the geometric relation of each joint of the mechanical arm by adopting a D-H method, establishing a connecting rod coordinate system of the mechanical arm, and calculating and obtaining a D-H model of the mechanical arm according to the D-H parameters of the mechanical arm;
step 1.2, analyzing and calculating inverse kinematics of the mechanical arm according to the D-H model of the mechanical arm constructed in the step 1.1, and resolving continuous operation tasks into multiple groups of mechanical arm initial positions and target positions by using the inverse kinematics of the mechanical arm;
step 1.3, modeling is needed to be carried out on continuous operation tasks of the mechanical arm, wherein the modeling is completed through visual sensor equipment, and the visual sensor is used for detecting the number of objects and the pose of the objects; resolving the starting pose and the target pose of the mechanical arm when operating the object into a plurality of groups of starting positions and target positions in the joint space of the mechanical arm by using inverse kinematics;
step 1.4, taking the initial position of the mechanical arm in the step 1.3 as the initial position of a mechanical arm configuration space, taking the target position of the mechanical arm movement as the target position of the configuration space, teaching a planning task by adopting an expert movement planner, collecting a teaching track solved by the planner, and discretizing the teaching track to be used as training set data;
step 1.5, initializing parameters of the Gaussian mixture model by using a K-means algorithm according to the training set data in the step 1.4, wherein the number of Gaussian components is selected according to an AIC criterion, and then iteratively estimating the parameters of the Gaussian mixture model by using an EM algorithm;
step 1.6, generating sample points by combining the Gaussian mixture distribution trained in the step 1.5 with uniform distribution, and constructing an undirected graph data structure by adopting a LazyPRM algorithm;
and step 1.7, solving continuous planning tasks by adopting a graph search algorithm based on the undirected graph constructed in the step 1.6, performing collision detection on a solved path, inquiring a candidate path if collision occurs, then performing smooth processing on the path by adopting a pruning algorithm, and realizing continuous operation motion planning of the mechanical arm by utilizing a series of joint control commands contained in the path.
2. The method for planning the continuous operation and movement of the mechanical arm according to claim 1, wherein: step 1.2, calculating inverse kinematics by using the D-H model of the mechanical arm established in the step 1.1, and providing an initial point and a target point for subsequent continuous motion planning of the mechanical arm; the continuous operation task requires the mechanical arm to plan to the initial position of the object, then the object is grabbed and placed to the target position, so the grabbing and placing poses at the tail end of the mechanical arm are resolved into corresponding initial joint angles and target joint angles by adopting inverse kinematics of the mechanical arm, and the initial joint angles and the target joint angles are used as the initial point and the target point of the planning problem.
3. The method for planning the continuous operation and movement of the mechanical arm according to claim 1, wherein: step 1.4, teaching a continuous operation task of the mechanical arm, wherein the teaching is implemented in a specific mode that a progressive optimal motion planner in an OMPL library is called to solve the planning task, and then a taught motion track is collected to serve as training set data; the track is obtained by solving a progressive optimal motion planning algorithm, and the teaching track is discretized and the key configuration is extracted to establish a training set.
4. The method for planning the continuous operation and movement of the mechanical arm according to claim 1, wherein: step 1.5, training a Gaussian mixture model by using the collected teaching data to provide a basis for offset sampling of a subsequent motion planner; and clustering the training data by adopting a K-means algorithm to realize the initialization of Gaussian mixture model parameters, selecting the number of Gaussian components according to an AIC criterion, and finally, iteratively estimating the parameters of the model by utilizing an EM algorithm, setting the maximum iteration number to be 100, and when the EM algorithm finishes iteration, determining that the Gaussian mixture model converges.
5. The method for planning the continuous operation and movement of the mechanical arm according to claim 1, wherein: step 1.6, generating sampling points by utilizing the trained Gaussian mixture distribution and uniform distribution so as to explore a configuration space of the mechanical arm; setting the number of initial sampling points as N and the number of nearest neighbor nodes as K; the undirected graph data structure comprises nodes and edges, wherein sampling points are used as the nodes of the undirected graph, the nodes are connected with the nodes according to the nearest neighbor principle to form the edges of the undirected graph, and collision detection is not carried out on the nodes and the edges in the graph in the composition stage; then setting the sampling threshold value to be 0.5; before each sampling, a random number p is sampled in 0-1 distribution, the random number p is compared with a sampling threshold value, when the p is smaller than or equal to the sampling threshold value, gaussian mixed distribution is used for biased sampling, and when the random number p is larger than the sampling threshold value, uniform distribution is adopted to generate sample points.
6. The method for planning the continuous operation and movement of the mechanical arm according to claim 1, wherein: step 1.7, solving a continuous planning problem by using the constructed undirected graph structure, giving a group of mechanical arm initial configuration and target configuration, solving a feasible path by using a graph search algorithm, calling a flexible collision detection library to perform collision detection on the path, deleting failed nodes and edges in the graph if the path has nodes or edges which collide, searching a collision-free path by using an A-algorithm through the rest nodes in the graph; carrying out smoothing treatment on the solved collision-free path by using a pruning algorithm, wherein the specific implementation mode is that any two nodes are connected, and if the collision is not caused, redundant nodes between the two nodes are removed, so that the path length is further optimized; and directly calling the constructed undirected graph in the next planning process, and solving the path of the motion planning by using an A-star algorithm.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211437040.3A CN115723129B (en) | 2022-11-17 | 2022-11-17 | Continuous operation motion planning method for mechanical arm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211437040.3A CN115723129B (en) | 2022-11-17 | 2022-11-17 | Continuous operation motion planning method for mechanical arm |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115723129A true CN115723129A (en) | 2023-03-03 |
| CN115723129B CN115723129B (en) | 2024-06-21 |
Family
ID=85296148
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211437040.3A Active CN115723129B (en) | 2022-11-17 | 2022-11-17 | Continuous operation motion planning method for mechanical arm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115723129B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116935252A (en) * | 2023-07-10 | 2023-10-24 | 齐鲁工业大学(山东省科学院) | Mechanical arm collision detection method based on sub-graph embedded graph neural network |
| CN117067220A (en) * | 2023-10-16 | 2023-11-17 | 广州上诺生物技术有限公司 | Neural network-based blood bank mechanical arm collision-free path planning method |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108237534A (en) * | 2018-01-04 | 2018-07-03 | 清华大学深圳研究生院 | A kind of space collision free trajectory method of continuous type mechanical arm |
| CN108356819A (en) * | 2018-01-17 | 2018-08-03 | 西安交通大学 | Based on the industrial machinery arm Collision Free Path Planning for improving A* algorithms |
| US20190184560A1 (en) * | 2017-01-19 | 2019-06-20 | Beijing University Of Technology | A Trajectory Planning Method For Six Degree-of-Freedom Robots Taking Into Account of End Effector Motion Error |
| CN110509279A (en) * | 2019-09-06 | 2019-11-29 | 北京工业大学 | A kind of trajectory path planning method and system of apery mechanical arm |
| CN110744543A (en) * | 2019-10-25 | 2020-02-04 | 华南理工大学 | Improved PRM obstacle avoidance motion planning method based on UR3 mechanical arm |
| CN113352319A (en) * | 2021-05-08 | 2021-09-07 | 上海工程技术大学 | Redundant mechanical arm obstacle avoidance trajectory planning method based on improved fast expansion random tree |
-
2022
- 2022-11-17 CN CN202211437040.3A patent/CN115723129B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190184560A1 (en) * | 2017-01-19 | 2019-06-20 | Beijing University Of Technology | A Trajectory Planning Method For Six Degree-of-Freedom Robots Taking Into Account of End Effector Motion Error |
| CN108237534A (en) * | 2018-01-04 | 2018-07-03 | 清华大学深圳研究生院 | A kind of space collision free trajectory method of continuous type mechanical arm |
| CN108356819A (en) * | 2018-01-17 | 2018-08-03 | 西安交通大学 | Based on the industrial machinery arm Collision Free Path Planning for improving A* algorithms |
| CN110509279A (en) * | 2019-09-06 | 2019-11-29 | 北京工业大学 | A kind of trajectory path planning method and system of apery mechanical arm |
| CN110744543A (en) * | 2019-10-25 | 2020-02-04 | 华南理工大学 | Improved PRM obstacle avoidance motion planning method based on UR3 mechanical arm |
| CN113352319A (en) * | 2021-05-08 | 2021-09-07 | 上海工程技术大学 | Redundant mechanical arm obstacle avoidance trajectory planning method based on improved fast expansion random tree |
Non-Patent Citations (1)
| Title |
|---|
| 于建均等: "基于RNN的机械臂任务模仿系统", 北京工业大学学报, vol. 44, no. 11, 11 October 2018 (2018-10-11), pages 1401 - 1408 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116935252A (en) * | 2023-07-10 | 2023-10-24 | 齐鲁工业大学(山东省科学院) | Mechanical arm collision detection method based on sub-graph embedded graph neural network |
| CN116935252B (en) * | 2023-07-10 | 2024-02-02 | 齐鲁工业大学(山东省科学院) | Mechanical arm collision detection method based on sub-graph embedded graph neural network |
| CN117067220A (en) * | 2023-10-16 | 2023-11-17 | 广州上诺生物技术有限公司 | Neural network-based blood bank mechanical arm collision-free path planning method |
| CN117067220B (en) * | 2023-10-16 | 2024-01-02 | 广州上诺生物技术有限公司 | Neural network-based blood bank mechanical arm collision-free path planning method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115723129B (en) | 2024-06-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115723129B (en) | Continuous operation motion planning method for mechanical arm | |
| Chi et al. | A generalized Voronoi diagram-based efficient heuristic path planning method for RRTs in mobile robots | |
| CN111546347B (en) | Mechanical arm path planning method suitable for dynamic environment | |
| Ni et al. | Pointnet++ grasping: Learning an end-to-end spatial grasp generation algorithm from sparse point clouds | |
| Danielczuk et al. | Object rearrangement using learned implicit collision functions | |
| CN111251295B (en) | Visual mechanical arm grabbing method and device applied to parameterized parts | |
| Kiatos et al. | Robust object grasping in clutter via singulation | |
| US9411335B2 (en) | Method and apparatus to plan motion path of robot | |
| CN109960880A (en) | A kind of industrial robot obstacle-avoiding route planning method based on machine learning | |
| Akinola et al. | Dynamic grasping with reachability and motion awareness | |
| CN111872934A (en) | Mechanical arm control method and system based on hidden semi-Markov model | |
| CN112605973A (en) | Robot motor skill learning method and system | |
| CN114147708B (en) | Mechanical arm obstacle avoidance path planning method based on improved longhorn beetle whisker search algorithm | |
| CN109940614B (en) | Mechanical arm multi-scene rapid motion planning method integrating memory mechanism | |
| Nam et al. | Fast and resilient manipulation planning for object retrieval in cluttered and confined environments | |
| Li et al. | Phasic self-imitative reduction for sparse-reward goal-conditioned reinforcement learning | |
| Harada et al. | Initial experiments on learning-based randomized bin-picking allowing finger contact with neighboring objects | |
| CN112428274A (en) | Space motion planning method of multi-degree-of-freedom robot | |
| CN116117822A (en) | RRT mechanical arm track planning method based on non-obstacle space probability potential field sampling | |
| Xu et al. | Efficient object manipulation to an arbitrary goal pose: Learning-based anytime prioritized planning | |
| Imtiaz et al. | Comparison of Two Reinforcement Learning Algorithms for Robotic Pick and Place with Non-Visual Sensing | |
| Lai et al. | Ltr*: Rapid replanning in executing consecutive tasks with lazy experience graph | |
| Yevsieiev et al. | Route constructing for a mobile robot based on the D-star algorithm | |
| Kang et al. | Grasp planning for occluded objects in a confined space with lateral view using monte carlo tree search | |
| Golluccio et al. | Objects relocation in clutter with robot manipulators via tree-based q-learning algorithm: Analysis and experiments |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |