CN118083808A - Dynamic path planning method and device for crown block system - Google Patents
Dynamic path planning method and device for crown block system Download PDFInfo
- Publication number
- CN118083808A CN118083808A CN202410489193.5A CN202410489193A CN118083808A CN 118083808 A CN118083808 A CN 118083808A CN 202410489193 A CN202410489193 A CN 202410489193A CN 118083808 A CN118083808 A CN 118083808A
- Authority
- CN
- China
- Prior art keywords
- node
- current
- sample
- crown block
- track
- 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 64
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 62
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 50
- 238000012549 training Methods 0.000 claims abstract description 36
- 238000005457 optimization Methods 0.000 claims abstract description 33
- 230000008569 process Effects 0.000 claims abstract description 20
- 239000013598 vector Substances 0.000 claims description 113
- 238000013528 artificial neural network Methods 0.000 claims description 15
- 125000004122 cyclic group Chemical group 0.000 claims description 14
- 238000004590 computer program Methods 0.000 claims description 11
- 239000011159 matrix material Substances 0.000 claims description 9
- 230000006870 function Effects 0.000 description 13
- 238000004364 calculation method Methods 0.000 description 7
- 238000004891 communication Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 230000003068 static effect Effects 0.000 description 4
- 238000010801 machine learning Methods 0.000 description 3
- 230000001133 acceleration Effects 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 235000008694 Humulus lupulus Nutrition 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000002441 reversible effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a dynamic path planning method and a dynamic path planning device for an overhead travelling crane system, which are characterized in that an optimized heuristic network is obtained after training a multi-layer convolutional neural network based on sample track environment information, an overhead travelling crane state of a sample overhead travelling crane and sample shortest running time between any two sample nodes, and the optimized heuristic network is integrated into an A-Star algorithm as a heuristic function so as to plan a path for a current overhead travelling crane based on a track map of the overhead travelling crane system and the optimized heuristic network in an initial state; in the advancing process, if the track state of any non-walking path point in the current optimal path changes, updating the current optimal path based on the optimization heuristic network and the track subgraph, searching and guiding an A-Star algorithm by utilizing the pre-trained optimization heuristic network, predicting path cost by integrating space environment information and motion constraint of the crown block, improving searching accuracy and improving the planning accuracy of the crown block system path.
Description
Technical Field
The invention relates to the technical field of path planning, in particular to a dynamic path planning method and device for an overhead travelling crane system.
Background
The static path planning algorithm is an algorithm that performs path planning with global environment information known and without consideration of dynamic obstacles. In the overhead travelling crane path planning scenario, the static path planning algorithm is mainly used for planning a path from a starting point to an end point for an overhead travelling crane in a fixed obstacle and a known working environment. Since the working environment of the crown block in the crown block carrying scene is known, a path planning algorithm of the crown block generally adopts a static planning algorithm, such as Dijkstra algorithm, A-Star algorithm and the like.
However, in an actual crown block working environment, dynamic obstacles may exist, such as a track with sudden failure, a crown block blockage, etc., which are unknown in the planning stage, and may cause the planned path to become infeasible or the handling time to reach the preset requirement when actually performed. If the static planning algorithm is called again when the working environment of the crown block changes, the calculation load is increased, and the real-time performance is difficult to meet the requirement of the crown block requiring quick response. More importantly, current heuristic path planning algorithms, such as the a-Star algorithm, employ heuristic functions, such as the usual manhattan distance and euclidean distance, in order to reduce the search space to increase the path search efficiency, but these conventional heuristic functions are not suitable for the scenario of dynamic path planning of the crown block. The reason is that: 1. the environment has complexity, and the simple linear distance or grid distance calculation is insufficient for accurately reflecting the cost of the actual path due to the complex factors; 2. crown block systems have environmental constraints in that crown blocks generally need to follow an orbital motion and cannot travel in reverse, and therefore, heuristic functions based on straight-line distances (e.g., euclidean distances) may not provide effective path guidance; 3. the crown block has its own specific motion constraints, which are not considered by the current heuristic functions. Therefore, the planning accuracy of the existing path planning algorithm in the crown block system still has a large improvement space.
Disclosure of Invention
The invention provides a dynamic path planning method and device for an overhead travelling crane system, which are used for solving the defect that the path planning precision of the overhead travelling crane system is not high enough in the prior art.
The invention provides a dynamic path planning method for an overhead travelling crane system, which comprises the following steps:
based on a track map of a crown block system and an optimization heuristic network, carrying out path planning on a current crown block by utilizing an A-Star algorithm to obtain a current optimal path, and controlling the current crown block to travel along the current optimal path;
In the travelling process of the current crown block, if the track state of any non-travelling path point in the current optimal path changes, updating the current optimal path based on the optimized heuristic network and the track subgraph; the track subgraph is obtained by deleting a path from the starting point to the end point of the current crown block, wherein the path comprises a non-walking path point with a changed track state; the optimized heuristic network is used for predicting the shortest expected running time from any node to the end point of the current crown block;
The optimization heuristic network is obtained by training the multi-layer convolutional neural network based on sample track environment information, a crown block state of a sample crown block and a sample shortest running time between any two sample nodes; the sample track environment information comprises a sample track map and an obstacle track, and the crown block state of the sample crown block comprises motion parameters of the sample crown block in different motion states.
According to the dynamic path planning method for the crown block system provided by the invention, the multi-layer convolutional neural network is trained based on sample track environment information, the crown block state of a sample crown block and the sample shortest running time between any two sample nodes, and the method specifically comprises the following steps:
And a data acquisition step: deleting nodes corresponding to the obstacle tracks in the sample track map and edges connecting the nodes corresponding to the obstacle tracks to obtain a sample track subgraph; acquiring a crown block state vector formed by motion parameters of the sample crown block in different motion states;
single training step: acquiring an adjacency matrix of the sample orbit subgraph, an orbit type of each node, an orbit environment vector formed by a current node and a sample target node, and inputting the orbit environment vector and the crown block state vector into the multi-layer convolutional neural network to obtain the shortest expected running time from the current node to the sample target node output by the multi-layer convolutional neural network;
parameter updating: adjusting parameters of the multi-layer convolutional neural network based on the difference between the shortest sample travel time from the current node to the sample target node and the shortest expected travel time;
Iterative steps: and selecting any child node or brother node of the current node in the sample track subgraph as the current node, and repeatedly executing the single training step and the parameter updating step.
According to the dynamic path planning method for the crown block system provided by the invention, the track environment vector and the crown block state vector are input into the multi-layer convolutional neural network to obtain the shortest predicted running time from the current node to the sample target node, which is output by the multi-layer convolutional neural network, specifically comprising the following steps:
Constructing a historical shortest predicted running time vector based on the shortest predicted running time from each historical node output by the multi-layer convolutional neural network to the sample target node; wherein each vector value in the historical shortest expected travel time vector corresponds to the shortest expected travel time from one historical node to the sample target node;
Coding the historical shortest expected running time vector based on a cyclic neural network, and determining a time sequence association vector corresponding to the historical shortest expected running time vector;
And inputting the orbit environment vector, the crown block state vector and the time sequence association vector into the multi-layer convolutional neural network to obtain the shortest expected running time from the current node to the sample target node, which is output by the multi-layer convolutional neural network.
According to the dynamic path planning method for the crown block system provided by the invention, the shortest historical expected running time vector is encoded based on the cyclic neural network, and the time sequence associated vector corresponding to the shortest historical expected running time vector is determined, which concretely comprises the following steps:
Inputting the shortest historical predicted running time vector into the cyclic neural network to obtain an initial association vector formed by hidden states output by the hidden layer of each time step of the cyclic neural network; wherein each vector value in the initial association vector corresponds to each history node;
Determining the weight corresponding to each history node based on the shortest running time of the sample from each history node to the current node; the shortest running time from any historical node to the current node is shorter, and the corresponding weight of any historical node is larger;
and determining a time sequence association vector corresponding to the shortest historical expected running time vector based on the weight corresponding to each historical node and the initial association vector.
According to the dynamic path planning method for the crown block system, which is provided by the invention, the path planning is carried out for the current crown block by utilizing an A-Star algorithm based on the track map and the optimized heuristic network of the crown block system, so as to obtain the current optimal path, and the method specifically comprises the following steps:
Initializing: adding the starting point of the current crown block into a first list, and setting the cost value of the starting point as a preset value; calculating a cost value of each node in the track map; the cost value of any node is the sum of the g value of any node and the heuristic value, the heuristic value of any node is the shortest expected running time from any node to the end point of the current crown block, the shortest expected running time from any node to the end point of the current crown block is determined based on the optimized heuristic network, and the g value of any node is the running time from the starting point to any node;
An expansion step: selecting a node with the minimum cost value from the first list as a current expansion node, and transferring the current expansion node to a second list; for any child node of the current expansion node in the track map, if the any child node is not in the first list and the second list, adding the any child node into the first list, and if the any child node is in the first list, updating the cost value of the any child node based on the sum of the g value of the current expansion node and the running time of the current expansion node to the any child node and the smaller value of the g value of the any child node;
Iterative steps: repeating the expanding step until the end point of the current crown block is added to the second list;
An optimal path determining step: and determining a current optimal path based on the nodes in the second list.
According to the dynamic path planning method for the crown block system provided by the invention, the updating of the current optimal path based on the optimized heuristic network and the track subgraph specifically comprises the following steps:
Parameter updating: clearing the first list and the second list; adding the current node of the current crown block to a first list, and setting the cost value of the current node of the current crown block as a preset value; based on the actual running time from the starting point of the current crown block to the current node of the current crown block, updating the g value of each node in the track subgraph; determining ancestor nodes of non-walking path points with changed track states in the track subgraph as update nodes, and updating cost values of the update nodes based on the optimization heuristic network;
An expansion step: selecting a node with the minimum cost value from the first list as a current expansion node, and transferring the current expansion node to a second list; for any child node of the current expansion node in the track subgraph, if the any child node is not in the first list and the second list, adding the any child node into the first list, and if the any child node is in the first list, updating the cost value of the any child node based on the sum of the g value of the current expansion node and the running time of the current expansion node to the any child node and the smaller value of the g value of the any child node;
Iterative steps: repeating the expanding step until the end point of the current crown block is added to the second list;
An optimal path determining step: and determining an updated current optimal path based on the nodes in the second list.
The invention also provides a dynamic path planning device facing the crown block system, which comprises:
The initial planning unit is used for planning a path for a current crown block by utilizing an A-Star algorithm based on a track map of the crown block system and an optimized heuristic network to obtain a current optimal path, and controlling the current crown block to travel along the current optimal path;
The dynamic planning unit is used for updating the current optimal path based on the optimized heuristic network and the track subgraph if the track state of any non-walking path point in the current optimal path changes in the advancing process of the current crown block; the track subgraph is obtained by deleting a path from the starting point to the end point of the current crown block, wherein the path comprises a non-walking path point with a changed track state; the optimized heuristic network is used for predicting the shortest expected running time from any node to the end point of the current crown block;
The optimization heuristic network is obtained by training the multi-layer convolutional neural network based on sample track environment information, a crown block state of a sample crown block and a sample shortest running time between any two sample nodes; the sample track environment information comprises a sample track map and an obstacle track, and the crown block state of the sample crown block comprises motion parameters of the sample crown block in different motion states.
According to the dynamic path planning device for the crown block system provided by the invention, the multi-layer convolutional neural network is trained based on sample track environment information, the crown block state of a sample crown block and the sample shortest running time between any two sample nodes, and the device specifically comprises:
And a data acquisition step: deleting nodes corresponding to the obstacle tracks in the sample track map and edges connecting the nodes corresponding to the obstacle tracks to obtain a sample track subgraph; acquiring a crown block state vector formed by motion parameters of the sample crown block in different motion states;
single training step: acquiring an adjacency matrix of the sample orbit subgraph, an orbit type of each node, an orbit environment vector formed by a current node and a sample target node, and inputting the orbit environment vector and the crown block state vector into the multi-layer convolutional neural network to obtain the shortest expected running time from the current node to the sample target node output by the multi-layer convolutional neural network;
parameter updating: adjusting parameters of the multi-layer convolutional neural network based on the difference between the shortest sample travel time from the current node to the sample target node and the shortest expected travel time;
Iterative steps: and selecting any child node or brother node of the current node in the sample track subgraph as the current node, and repeatedly executing the single training step and the parameter updating step.
The invention also provides an electronic device, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor realizes the dynamic path planning method facing the crown block system when executing the program.
The present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements a dynamic path planning method for an overhead travelling crane system as described in any one of the above.
The invention also provides a computer program product comprising a computer program which when executed by a processor implements a dynamic path planning method for an overhead travelling crane system as described in any one of the above.
According to the dynamic path planning method and device for the crown block system, the optimal heuristic network is obtained after training the multi-layer convolutional neural network based on sample track environment information, the crown block state of the sample crown block and the sample shortest running time between any two sample nodes, the optimal heuristic network is integrated into an A-Star algorithm as a heuristic function, and in an initial state, the path planning is carried out for the current crown block by utilizing the A-Star algorithm based on a track map of the crown block system and the optimal heuristic network, so that the current optimal path is obtained, and the current crown block is controlled to run along the current optimal path; in the advancing process, if the track state of any non-walking path point in the current optimal path changes, updating the current optimal path based on the optimization heuristic network and the track subgraph, and searching and guiding an A-Star algorithm by utilizing the optimization heuristic network obtained by pre-training, so that the space environment information and the motion constraint of the crown block are integrated to predict the path cost, the searching accuracy is improved, and the planning precision of the crown block system path is improved.
Drawings
In order to more clearly illustrate the invention or the technical solutions of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are some embodiments of the invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
Fig. 1 is a schematic flow chart of a dynamic path planning method for an overhead travelling crane system;
FIG. 2 is a flow chart of the optimized heuristic network training method provided by the invention;
fig. 3 is a schematic structural diagram of a dynamic path planning device facing an overhead travelling crane system provided by the invention;
fig. 4 is a schematic structural diagram of an electronic device provided by the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is apparent that the described embodiments are some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Fig. 1 is a schematic flow chart of a dynamic path planning method for an overhead travelling crane system, as shown in fig. 1, the method includes:
step 110, planning a path for a current crown block by utilizing an A-Star algorithm based on a track map of a crown block system and an optimization heuristic network to obtain a current optimal path, and controlling the current crown block to travel along the current optimal path;
Step 120, in the traveling process of the current crown block, if the track state of any non-traveling path point in the current optimal path changes, updating the current optimal path based on the optimized heuristic network and the track subgraph; the track subgraph is obtained by deleting a path from the starting point to the end point of the current crown block, wherein the path comprises a non-walking path point with a changed track state; the optimized heuristic network is used for predicting the shortest expected running time from any node to the end point of the current crown block;
The optimization heuristic network is obtained by training the multi-layer convolutional neural network based on sample track environment information, a crown block state of a sample crown block and a sample shortest running time between any two sample nodes; the sample track environment information comprises a sample track map and an obstacle track, and the crown block state of the sample crown block comprises motion parameters of the sample crown block in different motion states.
Specifically, in the initial state, the path planning can be performed for the current crown block by utilizing an A-Star algorithm and combining with the current track map of the crown block system to obtain the current optimal path. The global track map of the whole track system can be converted into a directed map, each node in the directed map corresponds to any track in the track system, and if any first track and any second track can pass through the first track and the second track in turn according to a specified running direction, the node corresponding to the first track in the directed map has an edge pointing to the node corresponding to the second track, and then only a sub-graph taking the starting point of the current crown block in the directed map as a root node is reserved as the track map. However, considering that the a-Star algorithm is a heuristic path planning algorithm, the common heuristic functions are manhattan distance and euclidean distance, but the linear distance or grid distance calculation in these traditional heuristic functions is not enough to accurately reflect the cost of the actual path in the crown block system; and the heuristic function may not provide effective path guidance under the condition that the crown block needs to satisfy environmental constraints and self-motion constraints. In this regard, the embodiment of the invention optimizes and improves the heuristic function in the A-Star algorithm, and utilizes the optimized heuristic network obtained by pre-training to conduct search guidance on the A-Star algorithm so as to overcome the problems of the existing heuristic function. The optimized heuristic network is used for predicting the shortest expected running time from any node to the end point of the current crown block, namely a heuristic value (h value) in an A-Star algorithm.
Specifically, a search process of an optimization heuristic network guidance A-Star algorithm suitable for crown block dynamic path planning can be constructed by utilizing a machine learning technology, a decision process can be learned and optimized from data by utilizing machine learning, the method is suitable for processing complex environments, and the characteristics of motion constraint and environment constraint of crown blocks can be effectively integrated, so that the accuracy of cost calculation is improved. In addition, considering that heuristic functions are used to calculate the cost of different nodes starting from the starting node to the same target node during the application of the A-Star algorithm, it can be understood as a number of similar but slightly different graph theory problems, and for this scenario, the machine learning model will be more efficient. Therefore, the multi-layer convolutional neural network can be trained based on the sample track environment information, the crown block state of the sample crown block and the sample shortest running time between any two sample nodes, and an optimized heuristic network is obtained. The sample track environment information comprises a sample track map and obstacle tracks (the obstacle tracks can be randomly arranged), and the crown block states of the sample crown block comprise motion parameters of the sample crown block in different motion states. It can be seen that the optimization heuristic network is based on spatial environment information in the training process, including relevant information of a sample track map (i.e. a sub-graph of a directed graph of a sample global track map in which a root node is a starting point of a sample crown block), and a current obstacle track, such as node attributes (track types), an adjacency matrix, etc. in the sample track map, and also considers crown block states of the sample crown block to satisfy motion constraints of the sample crown block when predicting the shortest expected travel time from the current node to the sample target node, so that a mapping between spatial environment information and motion constraints of the crown block and the shortest expected travel time from the current node to the sample target node can be established.
In some embodiments, as shown in fig. 2, the optimized heuristic network may be trained based on the following steps:
Data acquisition step 210: deleting nodes corresponding to the obstacle tracks in the sample track map and edges connecting the nodes corresponding to the obstacle tracks to obtain a sample track subgraph; acquiring a crown block state vector formed by motion parameters of the sample crown block in different motion states;
Single training step 220: acquiring an adjacency matrix of the sample orbit subgraph, an orbit type of each node, an orbit environment vector formed by a current node and a sample target node, and inputting the orbit environment vector and the crown block state vector into the multi-layer convolutional neural network to obtain the shortest expected running time from the current node to the sample target node output by the multi-layer convolutional neural network;
Parameter updating step 230: adjusting parameters of the multi-layer convolutional neural network based on the difference between the shortest sample travel time from the current node to the sample target node and the shortest expected travel time;
Iterative step 240: and selecting any child node or brother node of the current node in the sample track subgraph as the current node, and repeatedly executing the single training step and the parameter updating step.
And deleting nodes corresponding to the obstacle track and edges connecting the nodes corresponding to the obstacle track in the sample track map, and deleting nodes without father nodes except the starting point of the sample crown block in the sample track map to obtain a sample track child. It should be noted that, in the embodiment of the present invention, when there is an edge pointing to the solution B in the track map/sample track map, the node a is a parent node of the node B, the node B is a child node of the node a, and if any node C can reach the node D via at least one directed edge, the node C is an ancestor node of the node D. Meanwhile, an overhead travelling crane state vector formed by the motion parameters of the sample overhead travelling crane in different motion states is obtained, wherein the motion parameters of the sample overhead travelling crane in different motion states comprise the speed, the acceleration and the deceleration of the sample overhead travelling crane running on different types of tracks.
In the training process, the sample crown block randomly selects the travelling direction from the starting point, and a single training step is called each time the sample crown block arrives at a new track, and the shortest expected travelling time from the current node to the sample target node of the sample crown block is predicted by using the multi-layer convolutional neural network. The method comprises the steps of obtaining an adjacency matrix of a sample orbit subgraph, an orbit type of each node, an orbit environment vector formed by a current node and a sample target node, inputting the orbit environment vector and a crown block state vector into a multi-layer convolutional neural network, and obtaining the shortest expected running time from the current node to the sample target node output by the multi-layer convolutional neural network. Then, parameters of the multi-layer convolutional neural network are adjusted based on a difference between a sample shortest travel time from the current node to the sample target node and a model predicted shortest expected travel time. And selecting any child node or brother node of the current node in the sample track subgraph as the current node, and repeatedly executing the single training step and the parameter updating step until the multi-layer convolutional neural network converges, so that the optimized heuristic network is obtained.
In other embodiments, considering that in path planning, the costs between nodes are often not independent, but rather there is a certain timing correlation, especially in the a-Star algorithm, the node information searched before is searched backwards from the starting point, so that the path including the current node may be considered by the cost information of the node searched before, and therefore, the cost information of the node searched before may provide assistance for calculating the cost of the current node. Thus, embodiments of the present invention capture this correlation by encoding time series data so that the model can understand how changes in past node costs affect the costs of current and future nodes. Furthermore, encoding cost information of previous nodes into time series data may provide a rich feature set for the model. These features include not only the specific cost value of each node, but also statistics of rate of change, acceleration, etc. between costs. These additional features help the model more fully understand the distribution and dynamics of path costs. Thus, the shortest predicted travel time of each historical node to sample target node output before the multi-layer convolutional neural network may be incorporated in predicting the shortest predicted travel time.
Specifically, a historical shortest expected travel time vector may be constructed according to the output order of the multi-layer convolutional neural network based on the shortest expected travel time from each historical node to the sample target node output by the multi-layer convolutional neural network. Wherein each vector value in the historical shortest expected travel time vector is the shortest expected travel time of the corresponding historical node to the sample target node. And then, encoding the historical shortest expected running time vector based on the cyclic neural network, and determining a time sequence association vector corresponding to the historical shortest expected running time vector. Here, the shortest historical predicted running time vector can be input into the cyclic neural network to obtain an initial association vector formed by the hidden states output by the hidden layers of each time step of the cyclic neural network; wherein each vector value in the initial association vector corresponds to each history node. Then determining the weight corresponding to each history node based on the shortest running time of the sample from each history node to the current node; the shorter the shortest running time of any history node to the current node is, the more likely the current node is passed when the shortest predicted running time of the history node to the sample target node is predicted, the more useful the prediction result is, and the larger the weight corresponding to the history node is. Based on the weights and initial correlation vectors corresponding to each of the history nodes, a time-series correlation vector corresponding to the shortest predicted travel time vector of the history can be determined. For example, assuming that the respective history nodes 1,2, # i correspond to weights w1, w2, # wi, and the initial association vector is (v 1, v2, # vi), the timing association vector may be (w1×v1, w2×v2, # wi×vi). And then, inputting the orbit environment vector, the crown block state vector and the time sequence association vector into a multi-layer convolutional neural network to obtain the shortest expected running time from the current node to the sample target node output by the multi-layer convolutional neural network.
After the trained optimized heuristic network is obtained, the optimized heuristic network can be applied to a path planning process in an initial state and a dynamic planning process in a later stage.
In some embodiments, in an initial state, path planning can be performed for a current crown block by using an A-Star algorithm based on a track map of a crown block system and an optimized heuristic network to obtain a current optimal path. The specific flow may include the following steps:
Initializing: the starting point of the current crown block is added to the first list, and the cost value of the starting point is set to a preset value (which is a larger value). And calculating a cost value of each node in the track map. The cost value of any node is the sum of the g value of the node and the heuristic value, the heuristic value of any node is the shortest expected running time from the node to the end point of the current crown block, and the shortest expected running time from the node to the end point of the current crown block is determined based on an optimized heuristic network. In some embodiments, an overhead travelling crane state vector formed by motion parameters of a current overhead travelling crane in different motion states can be obtained, an adjacency matrix of a track map, a track type of each node, a track environment vector formed by the node and a destination point can be obtained, and the obtained track environment vector and overhead travelling crane state vector are input into an optimization heuristic network to obtain the shortest expected travelling time from the node to the destination point output by the network. The running time from the starting point to the node of the g value of any node can be calculated according to the track length of the track undergone by the path with the least hops from the starting point to the node in the track map and the average running speed of the track corresponding to the track type.
An expansion step: selecting a node with the minimum cost value from the first list as a current expansion node, and transferring the current expansion node to the second list; for any child node of the current expansion node in the track map, if the child node is not in the first list and the second list, adding the child node to the first list, and if the child node is in the first list, updating the cost value of the child node based on the sum of the g value (g_current) of the current expansion node and the running time (c) of the current expansion node to the child node (which can be calculated based on the track length and the average running speed of the track corresponding to the current expansion node) and the smaller value (min (g_current+c, g_child)) of the g value (g_child) of the child node.
Iterative steps: repeating the expanding step until the end point of the current crown block is added to the second list.
An optimal path determining step: and backtracking from the terminal point to the starting point according to the time sequence relation added into the second list based on the nodes in the second list, and obtaining the current optimal path.
Therefore, search guidance of the A-Star algorithm is conducted based on the optimized heuristic network, and accuracy of path cost calculation in the crown block system is improved. And after the current optimal path is obtained, controlling the current crown block to travel along the current optimal path. In the advancing process, if the track state of any non-walking path point in the current optimal path changes, for example, the track fails or is severely jammed, the current optimal path can be updated on the basis of an A-Star algorithm based on the optimized heuristic network and the track subgraph, namely, a dynamic planning stage is entered. The track subgraph is obtained by deleting a path from a starting point to an end point of a current crown block in the track map, wherein the path comprises a non-walking path point with a changed track state. It should be noted that if the complete a-Star algorithm is directly recalled here, the work of calculating g values and heuristic values of all nodes in the a-Star algorithm takes a long time, resulting in a decrease in real-time performance of dynamic path planning. In this regard, the embodiment of the invention improves the A-Star algorithm to improve the planning efficiency thereof, so that the method is more suitable for the dynamic planning stage.
Specifically, in the dynamic programming phase, the following steps may be adopted to perform path update:
Parameter updating: clearing the first list and the second list; and adding the current node of the current crown block to the first list, and setting the cost value of the current node of the current crown block to a preset value (a larger value). And updating the g value of each node in the track subgraph based on the actual running time from the starting point of the current crown block to the current node where the current crown block is located. The actual running time cr actually consumed by the current crown block from the starting point to the current node is recorded, and the current g value of each node in the track subgraph is subtracted by the actual running time cr to obtain the updated g value of each node in the track subgraph. And determining ancestor nodes of the non-walking path points with changed track states in the track subgraph as update nodes, and updating cost values of the update nodes based on the optimization heuristic network. That is, only the shortest expected travel time of the update node to the endpoint needs to be predicted as the heuristic value of the update node using the optimized heuristic network.
An expansion step: selecting a node with the minimum cost value from the first list as a current expansion node, and transferring the current expansion node to the second list; for any child node of the current expansion node in the track subgraph, if the child node is not in the first list and the second list, adding the child node into the first list, and if the child node is in the first list, updating the cost value of the child node based on the sum of the g value of the current expansion node and the running time of the current expansion node to the child node and a smaller value in the g value of the child node;
iterative steps: repeating the expanding step until the end point of the current crown block is added into the second list;
An optimal path determining step: and determining an updated current optimal path based on the nodes in the second list.
Therefore, in the path planning stage, the calculation result of the A-Star algorithm can be applied to a larger extent by introducing the parameter updating step, so that the calculation amount in the dynamic planning stage is reduced, and the path updating efficiency and instantaneity are improved.
In summary, according to the method provided by the embodiment of the invention, the multi-layer convolutional neural network is trained based on the sample track environment information, the crown block state of the sample crown block and the sample shortest running time between any two sample nodes to obtain the optimized heuristic network, and the optimized heuristic network is integrated into the A-Star algorithm as a heuristic function, so that in an initial state, the path planning is performed for the current crown block by using the A-Star algorithm based on the track map of the crown block system and the optimized heuristic network, so that the current optimal path is obtained, and the current crown block is controlled to run along the current optimal path; in the advancing process, if the track state of any non-walking path point in the current optimal path changes, updating the current optimal path based on the optimization heuristic network and the track subgraph, and searching and guiding an A-Star algorithm by utilizing the optimization heuristic network obtained by pre-training, so that the space environment information and the motion constraint of the crown block are integrated to predict the path cost, the searching accuracy is improved, and the planning precision of the crown block system path is improved.
The following describes a dynamic path planning device facing an overhead travelling crane system, and the dynamic path planning device facing the overhead travelling crane system and the dynamic path planning method facing the overhead travelling crane system described above can be referred to correspondingly.
Based on any of the above embodiments, fig. 3 is a schematic structural diagram of a dynamic path planning device for an overhead travelling crane system according to the present invention, where, as shown in fig. 3, the device includes:
The initial planning unit 310 is configured to plan a path for a current crown block by using an a-Star algorithm based on a track map of a crown block system and an optimization heuristic network, obtain a current optimal path, and control the current crown block to travel along the current optimal path;
the dynamic planning unit 320 is configured to update the current optimal path based on the optimized heuristic network and the track subgraph if the track state of any non-walking path point in the current optimal path changes during the traveling process of the current crown block; the track subgraph is obtained by deleting a path from the starting point to the end point of the current crown block, wherein the path comprises a non-walking path point with a changed track state; the optimized heuristic network is used for predicting the shortest expected running time from any node to the end point of the current crown block;
The optimization heuristic network is obtained by training the multi-layer convolutional neural network based on sample track environment information, a crown block state of a sample crown block and a sample shortest running time between any two sample nodes; the sample track environment information comprises a sample track map and an obstacle track, and the crown block state of the sample crown block comprises motion parameters of the sample crown block in different motion states.
According to the device provided by the embodiment of the invention, the optimal heuristic network is obtained after training the multi-layer convolutional neural network based on sample track environment information, the crown block state of the sample crown block and the sample shortest running time between any two sample nodes, and the optimal heuristic network is integrated into the A-Star algorithm as a heuristic function, so that in an initial state, the path planning is performed for the current crown block by utilizing the A-Star algorithm based on the track map of the crown block system and the optimal heuristic network, so that the current optimal path is obtained, and the current crown block is controlled to run along the current optimal path; in the advancing process, if the track state of any non-walking path point in the current optimal path changes, updating the current optimal path based on the optimization heuristic network and the track subgraph, and searching and guiding an A-Star algorithm by utilizing the optimization heuristic network obtained by pre-training, so that the space environment information and the motion constraint of the crown block are integrated to predict the path cost, the searching accuracy is improved, and the planning precision of the crown block system path is improved.
Based on any one of the above embodiments, the training of the multi-layer convolutional neural network based on the sample track environment information, the crown block state of the sample crown block, and the sample shortest running time between any two sample nodes specifically includes:
And a data acquisition step: deleting nodes corresponding to the obstacle tracks in the sample track map and edges connecting the nodes corresponding to the obstacle tracks to obtain a sample track subgraph; acquiring a crown block state vector formed by motion parameters of the sample crown block in different motion states;
single training step: acquiring an adjacency matrix of the sample orbit subgraph, an orbit type of each node, an orbit environment vector formed by a current node and a sample target node, and inputting the orbit environment vector and the crown block state vector into the multi-layer convolutional neural network to obtain the shortest expected running time from the current node to the sample target node output by the multi-layer convolutional neural network;
parameter updating: adjusting parameters of the multi-layer convolutional neural network based on the difference between the shortest sample travel time from the current node to the sample target node and the shortest expected travel time;
Iterative steps: and selecting any child node or brother node of the current node in the sample track subgraph as the current node, and repeatedly executing the single training step and the parameter updating step.
Based on any one of the above embodiments, the inputting the orbit environment vector and the crown block state vector into the multi-layer convolutional neural network, to obtain the shortest expected running time from the current node to the sample target node output by the multi-layer convolutional neural network, specifically includes:
Constructing a historical shortest predicted running time vector based on the shortest predicted running time from each historical node output by the multi-layer convolutional neural network to the sample target node; wherein each vector value in the historical shortest expected travel time vector corresponds to the shortest expected travel time from one historical node to the sample target node;
Coding the historical shortest expected running time vector based on a cyclic neural network, and determining a time sequence association vector corresponding to the historical shortest expected running time vector;
And inputting the orbit environment vector, the crown block state vector and the time sequence association vector into the multi-layer convolutional neural network to obtain the shortest expected running time from the current node to the sample target node, which is output by the multi-layer convolutional neural network.
Based on any one of the foregoing embodiments, the encoding the historical shortest expected running time vector based on the recurrent neural network, and determining a timing correlation vector corresponding to the historical shortest expected running time vector specifically includes:
Inputting the shortest historical predicted running time vector into the cyclic neural network to obtain an initial association vector formed by hidden states output by the hidden layer of each time step of the cyclic neural network; wherein each vector value in the initial association vector corresponds to each history node;
Determining the weight corresponding to each history node based on the shortest running time of the sample from each history node to the current node; the shortest running time from any historical node to the current node is shorter, and the corresponding weight of any historical node is larger;
and determining a time sequence association vector corresponding to the shortest historical expected running time vector based on the weight corresponding to each historical node and the initial association vector.
Based on any one of the above embodiments, the track map and the optimization heuristic network based on the crown block system perform path planning for the current crown block by using an a-Star algorithm to obtain a current optimal path, which specifically includes:
Initializing: adding the starting point of the current crown block into a first list, and setting the cost value of the starting point as a preset value; calculating a cost value of each node in the track map; the cost value of any node is the sum of the g value of any node and the heuristic value, the heuristic value of any node is the shortest expected running time from any node to the end point of the current crown block, the shortest expected running time from any node to the end point of the current crown block is determined based on the optimized heuristic network, and the g value of any node is the running time from the starting point to any node;
An expansion step: selecting a node with the minimum cost value from the first list as a current expansion node, and transferring the current expansion node to a second list; for any child node of the current expansion node in the track map, if the any child node is not in the first list and the second list, adding the any child node into the first list, and if the any child node is in the first list, updating the cost value of the any child node based on the sum of the g value of the current expansion node and the running time of the current expansion node to the any child node and the smaller value of the g value of the any child node;
Iterative steps: repeating the expanding step until the end point of the current crown block is added to the second list;
An optimal path determining step: and determining a current optimal path based on the nodes in the second list.
Based on any one of the foregoing embodiments, the updating the current optimal path based on the optimized heuristic network and the track subgraph specifically includes:
Parameter updating: clearing the first list and the second list; adding the current node of the current crown block to a first list, and setting the cost value of the current node of the current crown block as a preset value; based on the actual running time from the starting point of the current crown block to the current node of the current crown block, updating the g value of each node in the track subgraph; determining ancestor nodes of non-walking path points with changed track states in the track subgraph as update nodes, and updating cost values of the update nodes based on the optimization heuristic network;
An expansion step: selecting a node with the minimum cost value from the first list as a current expansion node, and transferring the current expansion node to a second list; for any child node of the current expansion node in the track subgraph, if the any child node is not in the first list and the second list, adding the any child node into the first list, and if the any child node is in the first list, updating the cost value of the any child node based on the sum of the g value of the current expansion node and the running time of the current expansion node to the any child node and the smaller value of the g value of the any child node;
Iterative steps: repeating the expanding step until the end point of the current crown block is added to the second list;
An optimal path determining step: and determining an updated current optimal path based on the nodes in the second list.
Fig. 4 is a schematic structural diagram of an electronic device according to the present invention, as shown in fig. 4, the electronic device may include: processor 410, memory 420, communication interface (Communications Interface) 430, and communication bus 440, wherein processor 410, memory 420, and communication interface 430 communicate with each other via communication bus 440. The processor 410 may invoke logic instructions in the memory 420 to perform a method of dynamic path planning for an overhead traveling crane system, the method comprising: based on a track map of a crown block system and an optimization heuristic network, carrying out path planning on a current crown block by utilizing an A-Star algorithm to obtain a current optimal path, and controlling the current crown block to travel along the current optimal path; in the travelling process of the current crown block, if the track state of any non-travelling path point in the current optimal path changes, updating the current optimal path based on the optimized heuristic network and the track subgraph; the track subgraph is obtained by deleting a path from the starting point to the end point of the current crown block, wherein the path comprises a non-walking path point with a changed track state; the optimized heuristic network is used for predicting the shortest expected running time from any node to the end point of the current crown block; the optimization heuristic network is obtained by training the multi-layer convolutional neural network based on sample track environment information, a crown block state of a sample crown block and a sample shortest running time between any two sample nodes; the sample track environment information comprises a sample track map and an obstacle track, and the crown block state of the sample crown block comprises motion parameters of the sample crown block in different motion states.
Further, the logic instructions in the memory 420 described above may be implemented in the form of software functional units and may be stored in a computer readable storage medium when sold or used as a stand alone product. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a usb disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
In another aspect, the present invention also provides a computer program product comprising a computer program stored on a non-transitory computer readable storage medium, the computer program comprising program instructions which, when executed by a computer, enable the computer to perform a method of dynamic path planning for an overhead travelling crane system provided by the methods described above, the method comprising: based on a track map of a crown block system and an optimization heuristic network, carrying out path planning on a current crown block by utilizing an A-Star algorithm to obtain a current optimal path, and controlling the current crown block to travel along the current optimal path; in the travelling process of the current crown block, if the track state of any non-travelling path point in the current optimal path changes, updating the current optimal path based on the optimized heuristic network and the track subgraph; the track subgraph is obtained by deleting a path from the starting point to the end point of the current crown block, wherein the path comprises a non-walking path point with a changed track state; the optimized heuristic network is used for predicting the shortest expected running time from any node to the end point of the current crown block; the optimization heuristic network is obtained by training the multi-layer convolutional neural network based on sample track environment information, a crown block state of a sample crown block and a sample shortest running time between any two sample nodes; the sample track environment information comprises a sample track map and an obstacle track, and the crown block state of the sample crown block comprises motion parameters of the sample crown block in different motion states.
In still another aspect, the present invention further provides a non-transitory computer readable storage medium having stored thereon a computer program, which when executed by a processor, is implemented to perform the above-provided dynamic path planning method for an overhead travelling crane system, the method comprising: based on a track map of a crown block system and an optimization heuristic network, carrying out path planning on a current crown block by utilizing an A-Star algorithm to obtain a current optimal path, and controlling the current crown block to travel along the current optimal path; in the travelling process of the current crown block, if the track state of any non-travelling path point in the current optimal path changes, updating the current optimal path based on the optimized heuristic network and the track subgraph; the track subgraph is obtained by deleting a path from the starting point to the end point of the current crown block, wherein the path comprises a non-walking path point with a changed track state; the optimized heuristic network is used for predicting the shortest expected running time from any node to the end point of the current crown block; the optimization heuristic network is obtained by training the multi-layer convolutional neural network based on sample track environment information, a crown block state of a sample crown block and a sample shortest running time between any two sample nodes; the sample track environment information comprises a sample track map and an obstacle track, and the crown block state of the sample crown block comprises motion parameters of the sample crown block in different motion states.
The apparatus embodiments described above are merely illustrative, wherein the elements illustrated as separate elements may or may not be physically separate, and the elements shown as elements may or may not be physical elements, may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art will understand and implement the present invention without undue burden.
From the above description of the embodiments, it will be apparent to those skilled in the art that the embodiments may be implemented by means of software plus necessary general hardware platforms, or of course may be implemented by means of hardware. Based on this understanding, the foregoing technical solution may be embodied essentially or in a part contributing to the prior art in the form of a software product, which may be stored in a computer readable storage medium, such as ROM/RAM, a magnetic disk, an optical disk, etc., including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method described in the respective embodiments or some parts of the embodiments.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims (10)
1. The dynamic path planning method for the crown block system is characterized by comprising the following steps of:
based on a track map of a crown block system and an optimization heuristic network, carrying out path planning on a current crown block by utilizing an A-Star algorithm to obtain a current optimal path, and controlling the current crown block to travel along the current optimal path;
In the travelling process of the current crown block, if the track state of any non-travelling path point in the current optimal path changes, updating the current optimal path based on the optimized heuristic network and the track subgraph; the track subgraph is obtained by deleting a path from the starting point to the end point of the current crown block, wherein the path comprises a non-walking path point with a changed track state; the optimized heuristic network is used for predicting the shortest expected running time from any node to the end point of the current crown block;
The optimization heuristic network is obtained by training the multi-layer convolutional neural network based on sample track environment information, a crown block state of a sample crown block and a sample shortest running time between any two sample nodes; the sample track environment information comprises a sample track map and an obstacle track, and the crown block state of the sample crown block comprises motion parameters of the sample crown block in different motion states.
2. The method for planning a dynamic path for an overhead travelling crane system according to claim 1, wherein the training of the multi-layer convolutional neural network based on the sample track environment information, the overhead travelling crane state of the sample overhead travelling crane and the sample shortest running time between any two sample nodes specifically comprises:
And a data acquisition step: deleting nodes corresponding to the obstacle tracks in the sample track map and edges connecting the nodes corresponding to the obstacle tracks to obtain a sample track subgraph; acquiring a crown block state vector formed by motion parameters of the sample crown block in different motion states;
single training step: acquiring an adjacency matrix of the sample orbit subgraph, an orbit type of each node, an orbit environment vector formed by a current node and a sample target node, and inputting the orbit environment vector and the crown block state vector into the multi-layer convolutional neural network to obtain the shortest expected running time from the current node to the sample target node output by the multi-layer convolutional neural network;
parameter updating: adjusting parameters of the multi-layer convolutional neural network based on the difference between the shortest sample travel time from the current node to the sample target node and the shortest expected travel time;
Iterative steps: and selecting any child node or brother node of the current node in the sample track subgraph as the current node, and repeatedly executing the single training step and the parameter updating step.
3. The method for planning a dynamic path for an overhead traveling crane system according to claim 2, wherein the inputting the orbit environment vector and the overhead traveling crane state vector into the multi-layer convolutional neural network obtains a shortest estimated traveling time from the current node to the sample target node output by the multi-layer convolutional neural network, specifically comprising:
Constructing a historical shortest predicted running time vector based on the shortest predicted running time from each historical node output by the multi-layer convolutional neural network to the sample target node; wherein each vector value in the historical shortest expected travel time vector corresponds to the shortest expected travel time from one historical node to the sample target node;
Coding the historical shortest expected running time vector based on a cyclic neural network, and determining a time sequence association vector corresponding to the historical shortest expected running time vector;
And inputting the orbit environment vector, the crown block state vector and the time sequence association vector into the multi-layer convolutional neural network to obtain the shortest expected running time from the current node to the sample target node, which is output by the multi-layer convolutional neural network.
4. The method for planning a dynamic path for an overhead travelling crane system according to claim 3, wherein the coding the historical shortest expected travelling time vector based on the cyclic neural network, and determining a time sequence associated vector corresponding to the historical shortest expected travelling time vector specifically comprises:
Inputting the shortest historical predicted running time vector into the cyclic neural network to obtain an initial association vector formed by hidden states output by the hidden layer of each time step of the cyclic neural network; wherein each vector value in the initial association vector corresponds to each history node;
Determining the weight corresponding to each history node based on the shortest running time of the sample from each history node to the current node; the shortest running time from any historical node to the current node is shorter, and the corresponding weight of any historical node is larger;
and determining a time sequence association vector corresponding to the shortest historical expected running time vector based on the weight corresponding to each historical node and the initial association vector.
5. The dynamic path planning method for an overhead traveling crane system according to claim 1, wherein the path planning is performed for a current overhead traveling crane by using an a-Star algorithm based on a track map of the overhead traveling crane system and an optimization heuristic network, so as to obtain a current optimal path, and the method specifically comprises:
Initializing: adding the starting point of the current crown block into a first list, and setting the cost value of the starting point as a preset value; calculating a cost value of each node in the track map; the cost value of any node is the sum of the g value of any node and the heuristic value, the heuristic value of any node is the shortest expected running time from any node to the end point of the current crown block, the shortest expected running time from any node to the end point of the current crown block is determined based on the optimized heuristic network, and the g value of any node is the running time from the starting point to any node;
An expansion step: selecting a node with the minimum cost value from the first list as a current expansion node, and transferring the current expansion node to a second list; for any child node of the current expansion node in the track map, if the any child node is not in the first list and the second list, adding the any child node into the first list, and if the any child node is in the first list, updating the cost value of the any child node based on the sum of the g value of the current expansion node and the running time of the current expansion node to the any child node and the smaller value of the g value of the any child node;
Iterative steps: repeating the expanding step until the end point of the current crown block is added to the second list;
An optimal path determining step: and determining a current optimal path based on the nodes in the second list.
6. The dynamic path planning method for an overhead travelling crane system according to claim 5, wherein updating the current optimal path based on the optimized heuristic network and the track subgraph specifically comprises:
Parameter updating: clearing the first list and the second list; adding the current node of the current crown block to a first list, and setting the cost value of the current node of the current crown block as a preset value; based on the actual running time from the starting point of the current crown block to the current node of the current crown block, updating the g value of each node in the track subgraph; determining ancestor nodes of non-walking path points with changed track states in the track subgraph as update nodes, and updating cost values of the update nodes based on the optimization heuristic network;
An expansion step: selecting a node with the minimum cost value from the first list as a current expansion node, and transferring the current expansion node to a second list; for any child node of the current expansion node in the track subgraph, if the any child node is not in the first list and the second list, adding the any child node into the first list, and if the any child node is in the first list, updating the cost value of the any child node based on the sum of the g value of the current expansion node and the running time of the current expansion node to the any child node and the smaller value of the g value of the any child node;
Iterative steps: repeating the expanding step until the end point of the current crown block is added to the second list;
An optimal path determining step: and determining an updated current optimal path based on the nodes in the second list.
7. A dynamic path planning device for an overhead travelling crane system, comprising:
The initial planning unit is used for planning a path for a current crown block by utilizing an A-Star algorithm based on a track map of the crown block system and an optimized heuristic network to obtain a current optimal path, and controlling the current crown block to travel along the current optimal path;
The dynamic planning unit is used for updating the current optimal path based on the optimized heuristic network and the track subgraph if the track state of any non-walking path point in the current optimal path changes in the advancing process of the current crown block; the track subgraph is obtained by deleting a path from the starting point to the end point of the current crown block, wherein the path comprises a non-walking path point with a changed track state; the optimized heuristic network is used for predicting the shortest expected running time from any node to the end point of the current crown block;
The optimization heuristic network is obtained by training the multi-layer convolutional neural network based on sample track environment information, a crown block state of a sample crown block and a sample shortest running time between any two sample nodes; the sample track environment information comprises a sample track map and an obstacle track, and the crown block state of the sample crown block comprises motion parameters of the sample crown block in different motion states.
8. The dynamic path planning device for an overhead traveling crane system according to claim 7, wherein the training of the multi-layer convolutional neural network based on the sample track environment information, the overhead traveling crane state of the sample overhead traveling crane and the sample shortest traveling time between any two sample nodes specifically comprises:
And a data acquisition step: deleting nodes corresponding to the obstacle tracks in the sample track map and edges connecting the nodes corresponding to the obstacle tracks to obtain a sample track subgraph; acquiring a crown block state vector formed by motion parameters of the sample crown block in different motion states;
single training step: acquiring an adjacency matrix of the sample orbit subgraph, an orbit type of each node, an orbit environment vector formed by a current node and a sample target node, and inputting the orbit environment vector and the crown block state vector into the multi-layer convolutional neural network to obtain the shortest expected running time from the current node to the sample target node output by the multi-layer convolutional neural network;
parameter updating: adjusting parameters of the multi-layer convolutional neural network based on the difference between the shortest sample travel time from the current node to the sample target node and the shortest expected travel time;
Iterative steps: and selecting any child node or brother node of the current node in the sample track subgraph as the current node, and repeatedly executing the single training step and the parameter updating step.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements a dynamic path planning method for an overhead travelling crane system according to any of claims 1 to 6 when executing the program.
10. A non-transitory computer readable storage medium having stored thereon a computer program, wherein the computer program when executed by a processor implements a dynamic path planning method for an overhead travelling crane system according to any of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410489193.5A CN118083808B (en) | 2024-04-23 | 2024-04-23 | Dynamic path planning method and device for crown block system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410489193.5A CN118083808B (en) | 2024-04-23 | 2024-04-23 | Dynamic path planning method and device for crown block system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN118083808A true CN118083808A (en) | 2024-05-28 |
| CN118083808B CN118083808B (en) | 2024-06-25 |
Family
ID=91148125
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410489193.5A Active CN118083808B (en) | 2024-04-23 | 2024-04-23 | Dynamic path planning method and device for crown block system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118083808B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106444750A (en) * | 2016-09-13 | 2017-02-22 | 哈尔滨工业大学深圳研究生院 | Two-dimensional code positioning-based intelligent warehousing mobile robot system |
| CN109945873A (en) * | 2019-04-04 | 2019-06-28 | 东南大学 | A kind of mixed path planing method for indoor mobile robot motion control |
| CN112612267A (en) * | 2020-12-08 | 2021-04-06 | 苏州挚途科技有限公司 | Automatic driving path planning method and device |
| US20230168681A1 (en) * | 2021-11-30 | 2023-06-01 | B&R Industrial Automation GmbH | Trajectory planning with flexible replanning functionality - changed end point |
| CN117740011A (en) * | 2023-12-18 | 2024-03-22 | 纵目科技(重庆)有限公司 | Method and device for automatic driving |
-
2024
- 2024-04-23 CN CN202410489193.5A patent/CN118083808B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106444750A (en) * | 2016-09-13 | 2017-02-22 | 哈尔滨工业大学深圳研究生院 | Two-dimensional code positioning-based intelligent warehousing mobile robot system |
| CN109945873A (en) * | 2019-04-04 | 2019-06-28 | 东南大学 | A kind of mixed path planing method for indoor mobile robot motion control |
| CN112612267A (en) * | 2020-12-08 | 2021-04-06 | 苏州挚途科技有限公司 | Automatic driving path planning method and device |
| US20230168681A1 (en) * | 2021-11-30 | 2023-06-01 | B&R Industrial Automation GmbH | Trajectory planning with flexible replanning functionality - changed end point |
| CN117740011A (en) * | 2023-12-18 | 2024-03-22 | 纵目科技(重庆)有限公司 | Method and device for automatic driving |
Also Published As
| Publication number | Publication date |
|---|---|
| CN118083808B (en) | 2024-06-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Petrich et al. | Map-based long term motion prediction for vehicles in traffic environments | |
| Hug et al. | Particle-based pedestrian path prediction using LSTM-MDL models | |
| CN109597425B (en) | Unmanned aerial vehicle navigation and obstacle avoidance method based on reinforcement learning | |
| CN110348636B (en) | Path plan prediction method, apparatus, device and computer readable storage medium | |
| CN112148008A (en) | Real-time unmanned aerial vehicle path prediction method based on deep reinforcement learning | |
| Butyrev et al. | Deep reinforcement learning for motion planning of mobile robots | |
| CN115560771A (en) | Sampling-based path planning method and device and automatic driving equipment | |
| Liang et al. | Hierarchical reinforcement learning with opponent modeling for distributed multi-agent cooperation | |
| Gan et al. | Dp-a*: For path planing of ugv and contactless delivery | |
| CN118083808B (en) | Dynamic path planning method and device for crown block system | |
| CN115617034B (en) | Multi-agent environment exploration method and device, electronic equipment and storage medium | |
| CN114861368B (en) | Construction method of railway longitudinal section design learning model based on near-end strategy | |
| CN113790729B (en) | Unmanned overhead traveling crane path planning method and device based on reinforcement learning algorithm | |
| CN116009542A (en) | Dynamic multi-agent coverage path planning method, device, equipment and storage medium | |
| CN111912411B (en) | Robot navigation positioning method, system and storage medium | |
| Zhang et al. | Reinforcement Learning from Demonstrations by Novel Interactive Expert and Application to Automatic Berthing Control Systems for Unmanned Surface Vessel | |
| Jin et al. | A multi-objective multi-agent framework for traffic light control | |
| CN116225046A (en) | Unmanned aerial vehicle autonomous path planning method based on deep reinforcement learning under unknown environment | |
| Feng et al. | Mobile robot obstacle avoidance based on deep reinforcement learning | |
| Wang et al. | Efficient reinforcement learning for autonomous ship collision avoidance under learning experience reuse | |
| CN112987742A (en) | Robot path planning method and planning system | |
| Saeed et al. | Domain-aware multiagent reinforcement learning in navigation | |
| Silva et al. | Hierarchical search strategy for a team of autonomous vehicles | |
| CN113687657B (en) | Method and storage medium for multi-agent formation dynamic path planning | |
| Agafonov et al. | Agent-Based Traffic Signal Control Using a Reinforcement Learning Approach |
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 |