CN117782084A - Path planning system and method for converter station inspection robot - Google Patents
Path planning system and method for converter station inspection robot Download PDFInfo
- Publication number
- CN117782084A CN117782084A CN202311616077.7A CN202311616077A CN117782084A CN 117782084 A CN117782084 A CN 117782084A CN 202311616077 A CN202311616077 A CN 202311616077A CN 117782084 A CN117782084 A CN 117782084A
- Authority
- CN
- China
- Prior art keywords
- path
- inspection robot
- bev
- data
- converter station
- 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.)
- Pending
Links
- 238000007689 inspection Methods 0.000 title claims abstract description 98
- 238000000034 method Methods 0.000 title claims abstract description 37
- 238000004422 calculation algorithm Methods 0.000 claims description 40
- 230000001133 acceleration Effects 0.000 claims description 34
- 230000033001 locomotion Effects 0.000 claims description 25
- 230000006870 function Effects 0.000 claims description 22
- 239000011159 matrix material Substances 0.000 claims description 21
- 238000013507 mapping Methods 0.000 claims description 20
- 238000004891 communication Methods 0.000 claims description 18
- 238000005457 optimization Methods 0.000 claims description 17
- 230000002457 bidirectional effect Effects 0.000 claims description 15
- 238000000605 extraction Methods 0.000 claims description 15
- 238000010586 diagram Methods 0.000 claims description 13
- 238000009826 distribution Methods 0.000 claims description 11
- 230000008859 change Effects 0.000 claims description 8
- 230000007246 mechanism Effects 0.000 claims description 7
- 230000008569 process Effects 0.000 claims description 7
- 230000002159 abnormal effect Effects 0.000 claims description 6
- 238000010276 construction Methods 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 6
- 238000003384 imaging method Methods 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 4
- 230000036461 convulsion Effects 0.000 claims description 4
- 238000001514 detection method Methods 0.000 claims description 4
- 230000005540 biological transmission Effects 0.000 claims description 3
- 230000000903 blocking effect Effects 0.000 claims description 3
- 238000009795 derivation Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 238000000844 transformation Methods 0.000 claims description 3
- 230000004927 fusion Effects 0.000 description 2
- 241000282414 Homo sapiens Species 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 239000011248 coating agent Substances 0.000 description 1
- 238000000576 coating method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 231100001261 hazardous Toxicity 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Landscapes
- Manipulator (AREA)
Abstract
The invention belongs to the field of direct current station/converter station inspection, and particularly relates to a path planning system and a path planning method of a converter station inspection robot.
Description
Technical Field
The invention belongs to the field of direct current station/converter station inspection, and particularly relates to a path planning system and method of a converter station inspection robot.
Background
The converter station is an important ring of stable operation of electric power, and the safety and stability of the converter station directly influence the normal operation of an electric power system, so that the converter station needs to be periodically inspected and maintained to ensure the normal operation of electric power equipment. However, converter stations often include hazardous environments such as high voltage equipment, potential gas leaks, high temperature areas, and the like. The manual inspectors can enter the environments and possibly face safety risks, based on the safety risks, robots are adopted in the station to replace human beings for inspection, operation cost can be reduced, personnel wages and related expenses are reduced, and production downtime is reduced. In consideration of safety and inspection efficiency, path planning is required before inspection and during inspection of the robot, firstly, the robot is guaranteed to quickly inspect a converter station according to the shortest path, so that time and resources are saved, secondly, the robot is guaranteed to avoid dangerous areas, contact with dangerous equipment is reduced, and safety is improved.
The invention patent with the bulletin number of CN 112254731A discloses a patrol robot, a patrol path planning method and a patrol path planning system, which comprises the following steps: the controller is used for controlling the driving device to drive the inspection robot to run; the distance sensor is used for collecting the distance between the base of the inspection robot and the ground in real time when the inspection robot runs in the current inspection path; the processor is used for sending a control instruction to the controller when the distance between the inspection robot base and the ground is greater than a first distance threshold; the controller is also used for re-planning the inspection path according to the control instruction and controlling the driving device to work so as to drive the inspection robot to run according to the re-planned inspection path. By the method, accurate measurement of an actual map is not needed, and the robot is redeployed after the robot inspection map is not needed to be modified and a virtual wall is arranged. When the no-pass area is changed, the coating can be removed at any time so as to change the no-pass area, and the no-pass area is simplified.
However, conventional path planning methods generally deal with global path planning, but may not be flexible enough in terms of obstacle avoidance, require additional local path planning to cope with obstacles, and may change the environment of the converter station, and conventional algorithms are generally not suitable for adapting to the dynamics of the present invention, and the conventional path planning methods require a long calculation time in terms of time, especially when the search space is large, may cause slow path planning speed and poor timeliness; the existing path planning method aims at the dynamic abnormal condition of an unknown environment, and cannot dynamically adjust the planned path in real time; in the search space, the usual a-algorithm may miss the optimal path. Therefore, a path planning strategy which is efficient, quick and dynamically adjustable in real time needs to be designed.
Disclosure of Invention
The invention aims to provide a path planning system and a path planning method for a converter station inspection robot, aiming at the problems in the prior art, and the method for carrying out path planning on the inspection robot by using the system greatly reduces the time for calculating the most effective path, ensures the dynamic obstacle avoidance in the inspection process, and effectively improves the inspection efficiency and reliability.
The technical scheme of the invention is as follows:
the path planning system of the converter station inspection robot is used for planning and controlling the walking path of the converter station inspection robot and is characterized by comprising a mechanical structure module, a power system module for providing motion for the converter station inspection robot, a control system module for operation, path planning and remote control and a communication module for transmitting communication data; the mechanical structure module comprises a chassis serving as a motion carrier, a sensor used for navigating and detecting objects in a station and measuring distance, and imaging equipment used for shooting and recording abnormal equipment in the process of inspection, wherein the inspection robot is arranged on the chassis, and the sensor and the imaging equipment are arranged on the inspection robot; the power system module comprises a battery for supplying power and a motor for supplying power; the control system module comprises an operation control unit, a path planning algorithm unit and a remote control unit; the communication module comprises communication equipment and a data transmission unit.
Specifically, the sensor comprises a laser radar sensor and an infrared sensor.
Specifically, the operation control unit integrates a processor and a control unit.
Specifically, the path planning algorithm fuses a bidirectional A-global path planning, a closed solution path optimization algorithm and a dynamic path real-time optimization algorithm based on time space attention coding.
Specifically, a path planning method of a converter station inspection robot, which uses the path planning system of the converter station inspection robot, comprises the following steps:
s1, a patrol robot controlled based on a remote control unit surrounds a circle on a patrol line of a converter station, and is familiar with the line;
s2, the inspection robot acquires environment data through a laser radar sensor, an infrared sensor and a camera device, and is used for constructing an environment map, identifying obstacles and determining the current position;
s3, constructing a preliminary point cloud map or grid map based on the acquired sensor data and image data, wherein the preliminary point cloud map or grid map is used for describing obstacles and passable areas in the environment;
s4, preprocessing data, namely preprocessing multi-mode data acquired by a laser radar sensor, an infrared sensor and a camera device, and performing data enhancement or normalization;
s5, data encoding, namely encoding multi-mode data acquired by a laser radar sensor, an infrared sensor and image pickup equipment into a form suitable for input of a transducer model, wherein the encoding comprises discretizing continuous numerical data, converting the discretized data into a vector form and serializing the data;
s6, data feature extraction, namely performing feature extraction on multi-mode data acquired by a laser radar sensor, an infrared sensor and a camera device, wherein cloud data acquired by the laser radar sensor and the infrared sensor are subjected to feature network extraction by adopting a Resnet18, and a visible light image shot by the camera device is subjected to feature extraction by adopting a Darknet53 network;
s7, fusing the multi-mode data acquired by the laser radar sensor, the infrared sensor and the camera equipment into a BEV format;
s8, extracting a feature F (x, y) through a multi-layer self-attention mechanism and position coding for an input multi-channel BEV image B (x, y): f (x, y) =transducer (B (x, y)), where F (x, y) represents a feature map, feature values at coordinates (x, y);
s9, after the self-attention mechanism and the position code extraction feature, forming a BEV Map mapping distribution diagram and target detection of the whole converter station inspection robot;
s10, calculating a path of the inspection robot by adopting a bidirectional A-x algorithm based on the obtained BEV Map mapping distribution diagram;
s11, the path planned by the bidirectional A-algorithm in the previous step is subjected to further path optimization through a polynomial-based closed-form solving algorithm and a transducer-based space-time dynamic fine adjustment;
s12, taking the path track inspected by the history of the robot, the path optimized by a polynomial closed solving algorithm and the current space-time environment state as inputs, sending the inputs to a time space attention coding and decoding network, and predicting the running state of the inspection robot at the next moment;
and S13, sending the sensor data, including the image and measured value information, to a platform for recording and backtracking inquiry based on the communication module at fixed time.
Specifically, the step S7 of fusing the multi-mode data collected by the laser radar sensor, the infrared sensor and the image capturing device into the BEV format includes the following steps:
s7-1: for radar point cloud data acquired by a laser radar sensor and an infrared sensor, projecting a three-dimensional point cloud onto a two-dimensional plane, and then rasterizing the plane to generate a height map;
s7-2: for radar data, converting distance and angle information into Cartesian coordinates, and then rasterizing on a BEV plane;
s7-3: for image data acquired by the image pickup device, projecting the image data onto a BEV plane to generate a color or intensity map;
s7-4: the BEV data are superimposed to form a multi-channel BEV image, let the laser radar BEV height map be H (x, y), the radar BEV distance map be R (x, y), and the camera BEV intensity map be I (x, y), then the multi-channel BEV image can be expressed as: b (x, y) = [ H (x, y), R (x, y), I (x, y) ], where B (x, y) represents the pixel values of the multi-channel BEV image at coordinates (x, y), and [ (j ] represents the channel overlap.
Specifically, the step S10 of calculating the planned path of the robot inspection by using the bidirectional a-algorithm based on the obtained BEV Map mapping distribution diagram includes the following steps:
s10-1: initializing a starting point and a finishing point, and respectively assigning an initial heuristic function value to the starting point and the finishing point, namely estimating the distance from the starting point to the finishing point;
s10-2: initializing priority open lists of a starting point and a terminal point, wherein the priority open lists are respectively used for storing nodes to be expanded;
s10-3: starting searching from the starting point and the end point at the same time, and selecting the node with the lowest priority each time for expansion;
s10-4: for the currently selected node, calculating and updating a heuristic function value h (n), a cost value g (n) and a total estimated value f (n), wherein f (n) =h (n) +g (n), g (n) is the actual path length from the starting point to the node, and f (n) is the heuristic function value plus the cost value;
s10-5: forward searching is carried out;
s10-6: performing reverse search, wherein the steps are the same as those of forward search;
s10-7: judging the return values of the step S10-6 and the step S10-5, if the return open list is empty and the target node is not reached yet, prompting a searching failure state, and selecting a forward searching path as a final path; if the search completion flag is returned, the full path is displayed.
Specifically, the forward searching steps are as follows:
(1) Taking out the optimal estimated distance node from the open list, if the open list is empty, indicating that all the nodes which can be searched in the forward direction have not reached the target node yet, returning a search failure result, otherwise, performing the step (2);
(2) Updating the optimal path node of forward search, judging whether the optimal node is in the optimal path of reverse search, if so, finding a path, and if not, adding the node into a forward search closing list;
(3) Processing the current optimal node, finding out a connecting point m of the current node, and searching for the next m point if the m node is in the forward closing list; otherwise, judging whether m is in the forward search open list, if not, respectively calculating g (m), h (m), and f (m);
(4) Setting the father node of m as n, and adding m into the forward search open list from small to large according to f value;
(5) Judging whether the distance from n to m is greater than g (m), if so, searching for the next connection point of n, otherwise, recalculating g (m), h (m), f (m),
until the node connected with n is searched.
Specifically, in the step S11, path optimization is performed by a polynomial closed-form solving algorithm, which specifically includes the following steps:
s11-1: the method comprises the steps of representing the number of segments of a path by k in a path planned by a bidirectional A-x algorithm, and representing the number of selected nodes by k+1, wherein the continuous track between two points of each segment is represented by an n-order polynomial as follows:
wherein p is 0 ,p 1 ,...,p n For the track parameters (n+1), a parameter vector p= [ p ] is set 0 ,p 1 ,...,p n ] T The trajectory may be written in vector form as: p (t) = [1, t 2 ,...,t n ]·p；
S11-2: the track is divided into k segments according to time, each segment is represented by a polynomial curve, the distance of each segment is calculated according to the path point, the time t is divided equally according to the distance (uniform time distribution),
a time series is obtained:
s11-3: for smooth movement of the inspection robot, the position, speed and acceleration of the starting point and the end point of each section are selected to be limited by the determined values, and the position limitation is used for the middle track point, and the method is expressed as follows:
s11-4: for the example of minimizing the change of acceleration, a polynomial with the order of 5 is constructed, then the polynomial has 6 unknown coefficients, and the mapping relation of mapping the track parameters to the motion quantity, namely the position, the speed and the acceleration is found by conducting derivation on the track for a plurality of times, so that the motion quantity is obtained:
the mapping is as follows: d, d i ＝A i p i ObtainingWherein constraints (position, velocity, acceleration) of the d vector are determined;
s11-5: a QP problem with constraints is further constructed,
and converts it into:
wherein: p represents position, V represents speed, A represents acceleration, t represents time, Q represents orthogonal matrix, P represents position component, V represents speed component, a acceleration component, d comprises position, speed and acceleration;
s11-6: for any time t, the position P (position), velocity V (acceleration), acceleration A (cceleration), jerk (change in angular acceleration), snap (derivative of thrust) of the trajectory is calculated from the parameters, expressed mathematically as having the same derivative value at the same time t in the trajectory segments i and i+1, expressed as:
p i (t i )＝p i+1 (t i )
v i (t i )＝v i+1 (t i )
a i (t i )＝a i+1 (t i )；
s11-7: to maintain the continuity constraint, the repetition variable is eliminated, and the continuity constraint construction mapping matrix M is designed as follows:
yielding d=md';
s11-7: further, the method comprises the steps of,wherein d is F Representing the determined variable in the trajectory, d P Representing a variable that can be freely varied, wherein C is the construction of a permutation matrix;
s11-8: setting k=a -1 MC, obtain
S11-9: further, the constraint condition is converted into a cost function through mathematical introduction, the constraint condition is converted into an unconstrained optimization problem, closed solving is performed, and aiming at the k sections of tracks obtained in the step S11-1, each section of track is a polynomial function of a high order, and the cost function and the following steps are written:
wherein J represents a cost function, which is a scalar;
s11-10: setting r=k T QK＝CA -T QA -1 C T And according to d for R matrix F And d P The following transformations are obtained by blocking the dimensions of (a):
s11-11: the Q matrix is a symmetric matrix, CA -T QA -1 C T ＝(CA -T QA -1 C T ) T The description that R is also a symmetric matrix, results inAnd then (I)>
S11-12: let J to d P Is the derivative of (2)Extreme points are found based on->
Calculation ofObtaining d P According to the formula->Further, p is obtained, and an optimized path is obtained.
The beneficial effects of the invention are as follows: the invention provides the space position information of the complete inspection line based on generation of the conversion-based visible light and radar laser point cloud fusion BEV Map, and completes the training of a target detection model while fusing the BEV Map, thereby being used for inspection of defects and anomalies of a DC station/converter station.
The invention integrates a direct current station/converter station robot path planning strategy and a device of a one-stage bidirectional A-global path planning, a two-stage closed solution path optimization algorithm and a three-stage dynamic path real-time optimization local path planning algorithm based on time space attention coding, thereby greatly reducing the time for calculating the most effective path, ensuring the dynamic obstacle avoidance in the inspection process and effectively improving the inspection efficiency and reliability.
Drawings
Fig. 1 is a schematic block diagram of a converter station robot of the present invention;
FIG. 2 is a functional block diagram of a mechanical structural module;
FIG. 3 is a functional block diagram of a powertrain module;
FIG. 4 is a functional block diagram of a control system module;
FIG. 5 is a functional block diagram of a communication module;
fig. 6 is a flowchart of the whole inspection and path planning algorithm of the inspection robot of the converter station;
fig. 7 is a spatio-temporal attention codec flow chart.
Detailed Description
The technical scheme of the invention is described in detail below with reference to the accompanying drawings and the specific embodiments.
Example 1
Fig. 1 is a schematic block diagram of a converter station robot according to the present embodiment. A path planning system of a converter station inspection robot is used for planning and controlling a walking path of the converter station inspection robot and comprises a mechanical structure module, a power system module for providing motion for the converter station inspection robot, a control system module for operation, path planning and remote control and a communication module for transmitting communication data; the mechanical structure module comprises a chassis serving as a motion carrier, a sensor used for navigating and detecting objects in a station and measuring distance, and an imaging device used for shooting and recording abnormal equipment in the process of inspection, wherein the sensor comprises a laser radar sensor and an infrared sensor, a general inspection robot usually has four or more legs, some wheels or tracks are adopted as a bottom plate to be an actual motion carrier, and in the mechanical structure module, a part of the inspection robot is further provided with a mechanical arm used for executing simple operation tasks such as switch operation or tool use. The inspection robot is arranged on the chassis, a sensor and a camera are arranged on the inspection robot, and the camera is generally used for shooting and recording abnormal equipment in the inspection process; the power system module comprises a battery for supplying power and a motor for supplying power, wherein the motor is used for controlling the leg or wheel of the inspection robot to move; the control system module comprises an operation control unit, a path planning algorithm unit and a remote control unit, wherein the operation control unit is integrated with a processor and the control unit and is used for executing various tasks and navigation, the path planning algorithm is fused with a bidirectional A-global path plan, the reliability of a robot inspection route is ensured based on a closed solving path optimization algorithm and a dynamic path real-time optimization algorithm based on time space attention coding, and the remote control unit is that an operator can control the robot through remote control equipment or preset tasks; the communication module comprises communication equipment and a data transmission unit, wherein the communication equipment is used for communicating with operators or a central monitoring system, transmitting data and receiving instructions, and the robot can periodically send sensor data, including images and measured values, in the communication module for analysis and recording. The functional structure of each functional module is shown in fig. 2-5.
Example 2
The present embodiment provides a path planning method for a converter station inspection robot, using the path planning system for a converter station inspection robot described in embodiment 1 above, including the following steps:
s1, a patrol robot controlled based on a remote control unit surrounds a circle on a patrol line of a converter station, and is familiar with the line;
s2, the inspection robot acquires environment data through a laser radar sensor, an infrared sensor and a camera device, and is used for constructing an environment map, identifying obstacles and determining the current position;
s3, constructing a preliminary point cloud map or grid map based on the acquired sensor data and image data, wherein the preliminary point cloud map or grid map is used for describing obstacles and passable areas in the environment;
s4, preprocessing data, namely preprocessing multi-mode data acquired by a laser radar sensor, an infrared sensor and a camera device, and performing data enhancement or normalization;
s5, data encoding, namely encoding multi-mode data acquired by a laser radar sensor, an infrared sensor and image pickup equipment into a form suitable for input of a transducer model, wherein the encoding comprises discretizing continuous numerical data, converting the discretized data into a vector form and serializing the data;
s6, data feature extraction, namely performing feature extraction on multi-mode data acquired by a laser radar sensor, an infrared sensor and a camera device, wherein cloud data acquired by the laser radar sensor and the infrared sensor are subjected to feature network extraction by adopting a Resnet18, and a visible light image shot by the camera device is subjected to feature extraction by adopting a Darknet53 network;
s7, fusing the multi-mode data acquired by the laser radar sensor, the infrared sensor and the camera equipment into a BEV format;
s8, extracting a feature F (x, y) through a multi-layer self-attention mechanism and position coding for an input multi-channel BEV image B (x, y): f (x, y) =transducer (B (x, y)), where F (x, y) represents a feature map, feature values at coordinates (x, y);
s9, after the self-attention mechanism and the position code extraction feature, forming a BEV Map mapping distribution diagram and target detection of the whole converter station inspection robot;
s10, calculating a path of the inspection robot by adopting a bidirectional A-x algorithm based on the obtained BEV Map mapping distribution diagram;
s11, the path planned by the bidirectional A-algorithm in the previous step is subjected to further path optimization through a polynomial-based closed-form solving algorithm and a transducer-based space-time dynamic fine adjustment;
s12, taking the path track inspected by the history of the robot, the path optimized by a polynomial closed solving algorithm and the current space-time environment state as inputs, sending the inputs to a time space attention coding and decoding network, and predicting the running state of the inspection robot at the next moment;
and S13, sending the sensor data, including the image and measured value information, to a platform for recording and backtracking inquiry based on the communication module at fixed time.
Example 3
The embodiment provides a specific step of fusing the multi-mode data collected by the laser radar sensor, the infrared sensor and the image capturing device into BEV format in step S7 in the method described in embodiment 2:
s7-1: for radar point cloud data acquired by a laser radar sensor and an infrared sensor, projecting a three-dimensional point cloud onto a two-dimensional plane, and then rasterizing the plane to generate a height map;
s7-2: for radar data, converting distance and angle information into Cartesian coordinates, and then rasterizing on a BEV plane;
s7-3: for image data acquired by the image pickup device, projecting the image data onto a BEV plane to generate a color or intensity map;
s7-4: the BEV data are superimposed to form a multi-channel BEV image, let the laser radar BEV height map be H (x, y), the radar BEV distance map be R (x, y), and the camera BEV intensity map be I (x, y), then the multi-channel BEV image can be expressed as: b (x, y) = [ H (x, y), R (x, y), I (x, y) ], where B (x, y) represents the pixel values of the multi-channel BEV image at coordinates (x, y), and [ (j ] represents the channel overlap.
Example 4
The embodiment provides the specific steps of calculating the planned path of the robot inspection by using the bidirectional a-algorithm based on the BEV Map mapping distribution Map in step S10 described in the method in embodiment 2:
s10-1: initializing a starting point and a finishing point, and respectively assigning an initial heuristic function value to the starting point and the finishing point, namely estimating the distance from the starting point to the finishing point;
s10-2: initializing priority open lists of a starting point and a terminal point, wherein the priority open lists are respectively used for storing nodes to be expanded;
s10-3: starting searching from the starting point and the end point at the same time, and selecting the node with the lowest priority each time for expansion;
s10-4: for the currently selected node, calculating and updating a heuristic function value h (n), a cost value g (n) and a total estimated value f (n), wherein f (n) =h (n) +g (n), g (n) is the actual path length from the starting point to the node, and f (n) is the heuristic function value plus the cost value;
s10-5: forward searching is carried out;
s10-6: performing reverse search, wherein the steps are the same as those of forward search;
s10-7: judging the return values of the step S10-6 and the step S10-5, if the return open list is empty and the target node is not reached yet, prompting a searching failure state, and selecting a forward searching path as a final path; if the search completion flag is returned, the full path is displayed.
The forward searching step is as follows:
(1) Taking out the optimal estimated distance node from the open list, if the open list is empty, indicating that all the nodes which can be searched in the forward direction have not reached the target node yet, returning a search failure result, otherwise, performing the step (2);
(2) Updating the optimal path node of forward search, judging whether the optimal node is in the optimal path of reverse search, if so, finding a path, and if not, adding the node into a forward search closing list;
(3) Processing the current optimal node, finding out a connecting point m of the current node, and searching for the next m point if the m node is in the forward closing list; otherwise, judging whether m is in the forward search open list, if not, respectively calculating g (m), h (m), and f (m);
(4) Setting the father node of m as n, and adding m into the forward search open list from small to large according to f value;
(5) Judging whether the distance from n to m is greater than g (m), if so, searching for the next connection point of n, otherwise, recalculating g (m), h (m), f (m),
until the node connected with n is searched.
Example 5
The present embodiment provides the path optimization performed by the polynomial closed-form solving algorithm in step S11 described in embodiment 2, which specifically includes the following steps:
s11-1: the method comprises the steps of representing the number of segments of a path by k in a path planned by a bidirectional A-x algorithm, and representing the number of selected nodes by k+1, wherein the continuous track between two points of each segment is represented by an n-order polynomial as follows:
wherein p is 0 ,p 1 ,...,p n For the track parameters (n+1), a parameter vector p= [ p ] is set 0 ,p 1 ,...,p n ] T The trajectory may be written in vector form as: p (t) = [1, t 2 ,...,t n ]·p；
S11-2: the track is divided into k segments according to time, each segment is represented by a polynomial curve, the distance of each segment is calculated according to the path point, the time t is divided equally according to the distance (uniform time distribution),
a time series is obtained:
s11-3: for smooth movement of the inspection robot, the position, speed and acceleration of the starting point and the end point of each section are selected to be limited by the determined values, and the position limitation is used for the middle track point, and the method is expressed as follows:
s11-4: for the example of minimizing the change of acceleration, a polynomial with the order of 5 is constructed, then the polynomial has 6 unknown coefficients, and the mapping relation of mapping the track parameters to the motion quantity, namely the position, the speed and the acceleration is found by conducting derivation on the track for a plurality of times, so that the motion quantity is obtained:
the mapping is as follows: d, d i ＝A i p i ObtainingWherein constraints (position, velocity, acceleration) of the d vector are determined;
s11-5: a QP problem with constraints is further constructed,
and converts it into:
wherein: p represents position, V represents speed, A represents acceleration, t represents time, Q represents orthogonal matrix, P represents position component, V represents speed component, a acceleration component, d comprises position, speed and acceleration;
s11-6: for any time t, the position P (position), velocity V (acceleration), acceleration A (cceleration), jerk (change in angular acceleration), snap (derivative of thrust), etc. of the trajectory is calculated from the parameters, if we want to minimize the goal of Jerk, we have to ensure that the joint third derivative is continuous, thus ensuring that the derivatives 0, 1, 2 (0 is position, 1 is velocity, and second is acceleration) are smooth, i.e. ensuring that the resulting acceleration function is still smooth, expressed mathematically as the same derivative value at the same time t in the trajectory segments i and i+1, expressed as:
p i (t i )＝p i+1 (t i )
v i (t i )＝v i+1 (t i )
a i (t i )＝a i+1 (t i )；
s11-7: to maintain the continuity constraint, the repetition variable is eliminated, and the continuity constraint construction mapping matrix M is designed as follows:
yielding d=md';
s11-7: further, the method comprises the steps of,wherein d is F Representing the determined variable in the trajectory, d P Representing a variable that can be freely varied, wherein C is the construction of a permutation matrix;
s11-8: setting k=a -1 MC, obtain/>
S11-9: further, the constraint condition is converted into a cost function through mathematical introduction, the constraint condition is converted into an unconstrained optimization problem, closed solving is performed, and aiming at the k sections of tracks obtained in the step S11-1, each section of track is a polynomial function of a high order, and the cost function and the following steps are written:
wherein J represents a cost function, which is a scalar;
s11-10: setting r=k T QK＝CA -T QA -1 C T And according to d for R matrix F And d P The following transformations are obtained by blocking the dimensions of (a):
s11-11: the Q matrix is a symmetric matrix, CA -T QA -1 C T ＝(CA -T QA -1 C T ) T The description that R is also a symmetric matrix, results inAnd then (I)>
S11-12: let J to d P Is the derivative of (2)Extreme points are found based on->
Calculation ofObtaining d P According to the formula->Further, p is obtained, and an optimized path is obtained.
Example 6
The implementation provides the method in the implementation 2, wherein step S12 is to send the path track of the robot after historical inspection and the current space-time environmental state after the path track is optimized by the polynomial closed solving algorithm as input to the time space attention coding and decoding network, and predict the running state of the inspection robot at the next moment, and the specific steps are as follows:
s12-1: through learning historical track data, a transducer can capture a motion mode of the inspection robot;
s12-2: the inspection time in the actual inspection process can be further shortened through learning the optimized inspection path;
s12-3: by learning the motion state of the inspection robot at the current moment and the data of the surrounding space-time environment, the transducer can capture real space-time dynamic information, so that a more accurate prediction result is provided for predicting the actual inspection path;
s12-4: carrying out time-space attention mechanism coding on the data information in the steps S12-1, S12-2 and S12-3, carrying out fusion on space-time path information and semantic information of the robot and the surrounding environment, and extracting key space-time characteristics;
wherein F (s, v, a, P) =TS_Transformer (s 1, v1, a1, s2, v2, a2, P, M)
S1, v1, a1 respectively represent the current position, speed and acceleration of the inspection robot, s2, v2, a2 respectively represent the current position, speed and acceleration of surrounding mobile personnel, P represents path information optimized by a polynomial closed-type solving algorithm, M represents a historical inspection path of the inspection robot, wherein s, v, a of F (s, v, a, P) are the final motion states, and P is the probability of the motion states;
s12-5: decoding the encoded information to generate different motion states and corresponding probabilities P, and selecting the motion state with the largest probability P as the final motion state of the robot;
s12-6: based on the continuous motion state, an actual motion track is generated, and the actual robot inspection operation is completed, wherein when one inspection point is reached, the robot can execute corresponding inspection tasks, such as taking an abnormal photo, and the like, and then can select the next inspection point according to the task list, and then repeat the steps until all the tasks are completed.
Finally, it should be noted that the above-mentioned embodiments are only for illustrating the technical scheme of the present invention and are not limiting; while the invention has been described in detail with reference to the preferred embodiments, those skilled in the art will appreciate that: modifications may be made to the specific embodiments of the present invention or equivalents may be substituted for part of the technical features thereof; without departing from the spirit of the invention, it is intended to cover the scope of the invention as claimed.
Claims (9)
1. The path planning system of the converter station inspection robot is used for planning and controlling the walking path of the converter station inspection robot and is characterized by comprising a mechanical structure module, a power system module for providing motion for the converter station inspection robot, a control system module for operation, path planning and remote control and a communication module for transmitting communication data;
the mechanical structure module comprises a chassis serving as a motion carrier, a sensor used for navigating and detecting objects in a station and measuring distance, and imaging equipment used for shooting and recording abnormal equipment in the process of inspection, wherein the inspection robot is arranged on the chassis, and the sensor and the imaging equipment are arranged on the inspection robot;
the power system module comprises a battery for supplying power and a motor for supplying power;
the control system module comprises an operation control unit, a path planning algorithm unit and a remote control unit;
the communication module comprises communication equipment and a data transmission unit.
2. The system for path planning for a converter station inspection robot of claim 1, wherein the sensor comprises a lidar sensor and an infrared sensor.
3. The route planning system of a converter station inspection robot of claim 1, wherein the arithmetic control unit integrates a processor and a control unit.
4. The system for path planning of a converter station inspection robot according to claim 1, wherein the path planning algorithm incorporates a bidirectional a-global path planning, a closed-form solution path optimization algorithm and a dynamic path real-time optimization algorithm based on time-space attention coding.
5. A method for planning a path of a converter station inspection robot, using the path planning system of the converter station inspection robot according to any one of claims 1 to 4, comprising the steps of:
s1, a patrol robot controlled based on a remote control unit surrounds a circle on a patrol line of a converter station, and is familiar with the line;
s2, the inspection robot acquires environment data through a laser radar sensor, an infrared sensor and a camera device, and is used for constructing an environment map, identifying obstacles and determining the current position;
s3, constructing a preliminary point cloud map or grid map based on the acquired sensor data and image data, wherein the preliminary point cloud map or grid map is used for describing obstacles and passable areas in the environment;
s4, preprocessing data, namely preprocessing multi-mode data acquired by a laser radar sensor, an infrared sensor and a camera device, and performing data enhancement or normalization;
s5, data encoding, namely encoding multi-mode data acquired by a laser radar sensor, an infrared sensor and image pickup equipment into a form suitable for input of a transducer model, wherein the encoding comprises discretizing continuous numerical data, converting the discretized data into a vector form and serializing the data;
s6, data feature extraction, namely performing feature extraction on multi-mode data acquired by a laser radar sensor, an infrared sensor and a camera device, wherein cloud data acquired by the laser radar sensor and the infrared sensor are subjected to feature network extraction by adopting a Resnet18, and a visible light image shot by the camera device is subjected to feature extraction by adopting a Darknet53 network;
s7, fusing the multi-mode data acquired by the laser radar sensor, the infrared sensor and the camera equipment into a BEV format;
s8, extracting a feature F (x, y) through a multi-layer self-attention mechanism and position coding for an input multi-channel BEV image B (x, y): f (x, y) =transducer (B (x, y)), where F (x, y) represents a feature map, feature values at coordinates (x, y);
s9, after the self-attention mechanism and the position code extraction feature, forming a BEV Map mapping distribution diagram and target detection of the whole converter station inspection robot;
s10, calculating a path of the inspection robot by adopting a bidirectional A-x algorithm based on the obtained BEV Map mapping distribution diagram;
s11, the path planned by the bidirectional A-algorithm in the previous step is subjected to further path optimization through a polynomial-based closed-form solving algorithm and a transducer-based space-time dynamic fine adjustment;
s12, taking the path track inspected by the history of the robot, the path optimized by a polynomial closed solving algorithm and the current space-time environment state as inputs, sending the inputs to a time space attention coding and decoding network, and predicting the running state of the inspection robot at the next moment;
and S13, sending the sensor data, including the image and measured value information, to a platform for recording and backtracking inquiry based on the communication module at fixed time.
6. The method for planning a path of a converter station inspection robot according to claim 5, wherein the step S7 of fusing the multi-mode data collected by the laser radar sensor, the infrared sensor and the image capturing device into BEV format comprises the steps of:
s7-1: for radar point cloud data acquired by a laser radar sensor and an infrared sensor, projecting a three-dimensional point cloud onto a two-dimensional plane, and then rasterizing the plane to generate a height map;
s7-2: for radar data, converting distance and angle information into Cartesian coordinates, and then rasterizing on a BEV plane;
s7-3: for image data acquired by the image pickup device, projecting the image data onto a BEV plane to generate a color or intensity map;
s7-4: the BEV data are superimposed to form a multi-channel BEV image, let the laser radar BEV height map be H (x, y), the radar BEV distance map be R (x, y), and the camera BEV intensity map be I (x, y), then the multi-channel BEV image can be expressed as: b (x, y) = [ H (x, y), R (x, y), I (x, y) ], where B (x, y) represents the pixel values of the multi-channel BEV image at coordinates (x, y), and [ (j ] represents the channel overlap.
7. The method for planning a path of a converter station inspection robot according to claim 5, wherein the step S10 of calculating a planned path of the robot inspection by using a bi-directional a-algorithm based on the BEV Map obtained includes the steps of:
s10-1: initializing a starting point and a finishing point, and respectively assigning an initial heuristic function value to the starting point and the finishing point, namely estimating the distance from the starting point to the finishing point;
s10-2: initializing priority open lists of a starting point and a terminal point, wherein the priority open lists are respectively used for storing nodes to be expanded;
s10-3: starting searching from the starting point and the end point at the same time, and selecting the node with the lowest priority each time for expansion;
s10-4: for the currently selected node, calculating and updating a heuristic function value h (n), a cost value g (n) and a total estimated value f (n), wherein f (n) =h (n) +g (n), g (n) is the actual path length from the starting point to the node, and f (n) is the heuristic function value plus the cost value;
s10-5: forward searching is carried out;
s10-6: performing reverse search, wherein the steps are the same as those of forward search;
s10-7: judging the return values of the step S10-6 and the step S10-5, if the return open list is empty and the target node is not reached yet, prompting a searching failure state, and selecting a forward searching path as a final path; if the search completion flag is returned, the full path is displayed.
8. The path planning method of a converter station inspection robot according to claim 7, wherein the step of forward searching is as follows:
(1) Taking out the optimal estimated distance node from the open list, if the open list is empty, indicating that all the nodes which can be searched in the forward direction have not reached the target node yet, returning a search failure result, otherwise, performing the step (2);
(2) Updating the optimal path node of forward search, judging whether the optimal node is in the optimal path of reverse search, if so, finding a path, and if not, adding the node into a forward search closing list;
(3) Processing the current optimal node, finding out a connecting point m of the current node, and searching for the next m point if the m node is in the forward closing list; otherwise, judging whether m is in the forward search open list, if not, respectively calculating g (m), h (m), and f (m);
(4) Setting the father node of m as n, and adding m into the forward search open list from small to large according to f value;
(5) Judging whether the distance from n to m is greater than g (m) or not, if so, searching for the next connection point of n, otherwise, recalculating g (m), h (m), f (m) until the node connected with n is searched.
9. The path planning method of a converter station inspection robot according to claim 7, wherein the path optimization in step S11 is performed by a polynomial closed-form solution algorithm, and specifically comprises the following steps:
s11-1: the method comprises the steps of representing the number of segments of a path by k in a path planned by a bidirectional A-x algorithm, and representing the number of selected nodes by k+1, wherein the continuous track between two points of each segment is represented by an n-order polynomial as follows:
wherein p is 0 ,p 1 ,...,p n For the track parameters (n+1), a parameter vector p= [ p ] is set 0 ,p 1 ,...,p n ] T The trajectory may be written in vector form as: p (t) = [1, t 2 ,…,t n ]·p；
S11-2: the track is divided into k sections according to time, each section is represented by a polynomial curve, the distance of each section is calculated according to the path points, and the time t (uniform time distribution) is halved according to the distance, so that a time sequence is obtained:
s11-3: for smooth movement of the inspection robot, the position, speed and acceleration of the starting point and the end point of each section are selected to be limited by the determined values, and the position limitation is used for the middle track point, and the method is expressed as follows:
s11-4: for the example of minimizing the change of acceleration, a polynomial with the order of 5 is constructed, then the polynomial has 6 unknown coefficients, and the mapping relation of mapping the track parameters to the motion quantity, namely the position, the speed and the acceleration is found by conducting derivation on the track for a plurality of times, so that the motion quantity is obtained:
the mapping is as follows: d, d i ＝A i p i ObtainingWherein constraints (position, velocity, acceleration) of the d vector are determined;
s11-5: a QP problem with constraints is further constructed,
and converts it into:
wherein: p represents position, V represents speed, A represents acceleration, t represents time, Q represents orthogonal matrix, P represents position component, V represents speed component, a acceleration component, d comprises position, speed, acceleration
A degree;
s11-6: for any time t, the position P (position), velocity V (acceleration), acceleration A (cceleration), jerk (change in angular acceleration), snap (derivative of thrust) of the trajectory is calculated from the parameters, expressed mathematically as having the same derivative value at the same time t in the trajectory segments i and i+1, expressed as:
p i (t i )＝p i+1 (t i )
v i (t i )＝v i+1 (t i )
a i (t i )＝a i+1 (t i )；
s11-7: to maintain the continuity constraint, the repetition variable is eliminated, and the continuity constraint construction mapping matrix M is designed as follows:
yielding d=md';
s11-7: further, the method comprises the steps of,wherein d is F Representing the determined variable in the trajectory, d P Representing a variable that can be freely varied, wherein C is the construction of a permutation matrix;
s11-8: setting k=a -1 MC, obtain
S11-9: further, the constraint condition is converted into a cost function through mathematical introduction, the constraint condition is converted into an unconstrained optimization problem, closed solving is performed, and aiming at the k sections of tracks obtained in the step S11-1, each section of track is a polynomial function of a high order, and the cost function and the following steps are written:
wherein J represents a cost function, which is a scalar;
s11-10: setting r=k T QK＝CA -T QA -1 C T And according to d for R matrix F And d P The following transformations are obtained by blocking the dimensions of (a):
s11-11: the Q matrix is a symmetric matrix, CA -T QA -1 C T ＝(CA -T QA -1 C T ) T The description that R is also a symmetric matrix, results inAnd then (I)>
S11-12: let J to d P Is the derivative of (2)Extreme points are found based on->
Calculation ofObtaining d P According to the formula->Further, p is obtained, and an optimized path is obtained.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311616077.7A CN117782084A (en) | 2023-11-29 | 2023-11-29 | Path planning system and method for converter station inspection robot |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311616077.7A CN117782084A (en) | 2023-11-29 | 2023-11-29 | Path planning system and method for converter station inspection robot |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117782084A true CN117782084A (en) | 2024-03-29 |
Family
ID=90390050
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311616077.7A Pending CN117782084A (en) | 2023-11-29 | 2023-11-29 | Path planning system and method for converter station inspection robot |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117782084A (en) |
-
2023
- 2023-11-29 CN CN202311616077.7A patent/CN117782084A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhang et al. | Automated guided vehicles and autonomous mobile robots for recognition and tracking in civil engineering | |
| CN111958591B (en) | Autonomous inspection method and system for semantic intelligent substation inspection robot | |
| CN107328418B (en) | Nuclear radiation detection path autonomous planning method of mobile robot in strange indoor scene | |
| CN113189977B (en) | Intelligent navigation path planning system and method for robot | |
| KR100669250B1 (en) | System and method for real-time calculating location | |
| CN112461227B (en) | Wheel type chassis robot inspection intelligent autonomous navigation method | |
| CN111968262A (en) | Semantic intelligent substation inspection operation robot navigation system and method | |
| US11698640B2 (en) | Method and apparatus for determining turn-round path of vehicle, device and medium | |
| CN112518739A (en) | Intelligent self-navigation method for reconnaissance of tracked chassis robot | |
| Kucuksubasi et al. | Transfer learning-based crack detection by autonomous UAVs | |
| CN113325837A (en) | Control system and method for multi-information fusion acquisition robot | |
| CN111413970A (en) | Ultra-wideband and vision integrated indoor robot positioning and autonomous navigation method | |
| CN111624994A (en) | Robot inspection method based on 5G communication | |
| Chen et al. | Robot navigation with map-based deep reinforcement learning | |
| CN214520204U (en) | Port area intelligent inspection robot based on depth camera and laser radar | |
| CN111158358B (en) | Method and system for self-optimization routing inspection of transformer/converter station based on three-dimensional model | |
| Artuñedo et al. | A decision-making architecture for automated driving without detailed prior maps | |
| CN104953709A (en) | Intelligent patrol robot of transformer substation | |
| CN113433937B (en) | Hierarchical navigation obstacle avoidance system and hierarchical navigation obstacle avoidance method based on heuristic exploration | |
| Zaki et al. | A navigation strategy for an autonomous patrol vehicle based on multi-fusion planning algorithms and multi-paradigm representation schemes | |
| CN114906131A (en) | Call control method, system and readable storage medium for automatic driving vehicle | |
| Garrote et al. | Mobile robot localization with reinforcement learning map update decision aided by an absolute indoor positioning system | |
| CN117260757A (en) | Robot inspection system based on inspection data | |
| US20210382491A1 (en) | Self-propelled inspection robot | |
| CN117782084A (en) | Path planning system and method for converter station inspection robot |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication |