Automatic operation control method and system for RGV (Rail Guided Vehicle)

Abstract

The invention relates to the technical field of bridge engineering, and discloses an RGV rail car automatic operation control method and system, and the method comprises the steps: obtaining a target transportation site of an RGV rail car, and carrying out the initial path planning of the target transportation site based on an optimized algorithm; driving according to a preset driving path, performing real-time environment scanning through a sensor, and analyzing scanned data; based on the real-time environment data, utilizing a TEB time elastic band algorithm to perform dynamic path re-planning on a predetermined driving path, driving according to the obstacle avoidance driving path, obtaining the residual electric quantity of the rail car, utilizing a charging demand decision model to judge the residual electric quantity, and if the residual electric quantity cannot support the rail car to drive to the target transportation place, determining that the rail car is not driven to the target transportation place. If yes, acquiring a nearest charging position, and generating a charging driving path according to the charging position. Various people and obstacles in the running process can be effectively avoided, and the safety and smoothness of the transportation process are guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of RGV rail vehicles, and particularly to an automatic operation control method and system for RGV rail vehicles. Background Art

[0002] In modern logistics warehousing and industrial automation production scenarios, as an important material transportation equipment, the operation efficiency and stability of RGV rail vehicles have a key impact on the entire production process. In traditional RGV rail vehicle operation control methods, path planning is often based on fixed preset routes, making it difficult to flexibly respond to complex and changeable working environments. Once there are people or other obstacles on the track, it is easy to cause problems such as transportation delays and even equipment collisions. Moreover, due to the large size of the warehousing site, the existing RGV rail vehicles manage electricity rather crudely, unable to accurately plan the driving path according to the real-time electricity consumption, often resulting in insufficient electricity and affecting the completion of tasks, reducing the production efficiency and operation cost of the overall logistics transfer. Therefore, how to perform more intelligent and efficient automatic operation control for RGV rail vehicles is a technical problem that urgently needs to be solved at present. Summary of the Invention

[0003] The purpose of the present invention is to solve the above problems and design an automatic operation control method and system for RGV rail vehicles.

[0004] The first aspect of the present invention provides an automatic operation control method for RGV rail vehicles, and the automatic operation control method for RGV rail vehicles includes the following steps:

[0005] Obtain the target transportation location of the RGV rail vehicle, and perform initial path planning on the target transportation location based on the optimized Dijkstra algorithm to obtain a predetermined driving path;

[0006] Drive according to the predetermined driving path, perform real-time environmental scanning through sensors, and analyze the scanned data using the RTAB-Map algorithm to obtain real-time environmental data;

[0007] Based on the real-time environmental data, use the TEB time elastic band algorithm to perform dynamic path replanning on the predetermined driving path to obtain an obstacle avoidance driving path;

[0008] Drive according to the obstacle avoidance driving path, obtain the remaining power of the RGV rail vehicle, and use the charging demand decision model to judge the remaining power. If the remaining power can support the RGV rail vehicle to drive to the target transportation location, continue to drive according to the obstacle avoidance driving path;

[0009] If the remaining power cannot support the RGV rail vehicle to travel to the target transportation location, obtain the nearest charging location, generate a charging travel path based on the charging location, and travel based on the charging travel path.

[0010] Optionally, in the first implementation manner of the first aspect of the present invention, the obtaining of the target transportation location of the RGV rail vehicle and the initial path planning of the target transportation location based on the optimized Dijkstra algorithm to obtain a predetermined travel path includes:

[0011] Obtain the target transportation location of the RGV rail vehicle, establish a coordinate system for the travel workshop of the RGV rail vehicle, with the initial location of the RGV rail vehicle as the origin (0, 0), the track direction as the x-axis, and the direction perpendicular to the track as the y-axis. Then the two-dimensional coordinates of the target transportation location are (x goal , y goal );

[0012] Optimize the Dijkstra algorithm through bidirectional search. Start from the starting point s and the target point t respectively, create two priority queues Q s and Q t , which are respectively used to store the nodes to be expanded; initialize the distance d s (s)=0 of the starting point s, the distance d t (t)=0 of the target point t, and the distances of other nodes are infinite. Create two sets S s and S t , which are respectively used to record the nodes that have been expanded from the starting point and the target point;

[0013] Select the node u s with the smallest distance from Q s , remove it from Q s and add it to S s . If the node u s is already in S s , it means that the two search directions meet, merge the paths and return;

[0014] For all adjacent nodes v s of the node u s , calculate the distance d s from the starting point through the node u s to the adjacent node v new1 =d s (u s )+ω(u s , v s ), where d s (u s ) represents the shortest distance from the starting point to the node u s , and ω(u s , vs ) is the weight of edge (u s , v s ). If d new1 < d s (v s ), then update d s (v s ) = d new1 , and add v s to Q s ;

[0015] Select the node u t with the minimum distance from Q t , remove it from Q t and add it to S t . If u t is already in S t , it means that the two search directions meet. Merge the paths and return. For all adjacent nodes v t of u t , calculate the distance d t from the target point through u t to v new2 = d t (u t ) + ω(u t , v t ), where d t (u t ) represents the shortest distance from the target point to the node, and ω(u t , v t ) is the weight of edge (u t , v t ). If d new2 < d t (v t ), then update d t (v t ) = d new2 , and add v t to Q t ;

[0016] Perform step iteration on Q s and Q t until Q s and Q t are empty. When the two search directions meet, let the meeting node be m, the path from the starting point to m be P s , and the path from the target point to m be P t . Reverse P t and merge it with P s to obtain the predetermined driving path.

[0017] Optionally, in the second implementation manner of the first aspect of the present invention, for traveling according to the predetermined traveling path, real-time environmental scanning is performed through a sensor, and the scanned data is analyzed by using the RTAB-Map algorithm to obtain real-time environmental data, including:

[0018] For traveling according to the predetermined traveling path, a dynamic model of the RGV rail vehicle is established by using the MPC predictive control model, and the linear discrete-time state space equation of the RGV rail vehicle is:

[0019] x k+1 = Ax k + Bu k

[0020] where x k is the system state vector at time k, x k+1 is the system state vector at time k + 1, the system state vector includes at least position, speed, and acceleration, u k is the control input vector at time k, the control input vector includes at least motor torque and motor speed, A is the state transition matrix, and B is the control input matrix;

[0021] Set the objective function J of the MPC predictive control model:

[0022]

[0023] where x ref,k+i|k is the reference state vector at time k + i, x k+i|k is the predicted state vector at time k, Q is the weight matrix of the state error, R is the weight matrix of the control input, N is the prediction horizon length, u k+i|k is the predicted value of the control input vector at time k for time k + i, including at least motor torque and motor speed, is the transpose of u k+i|k and T represents vector transpose.

[0024] Optionally, in the third implementation manner of the first aspect of the present invention, for traveling according to the predetermined traveling path, real-time environmental scanning is performed through a sensor, and the scanned data is analyzed by using the RTAB-Map algorithm to obtain real-time environmental data, further including:

[0025] Performing real-time environmental scanning by using a sensor, the sensor includes at least an infrared sensor, a millimeter-wave radar, an inertial measurement unit, an image sensor, and a lidar; performing real-time environmental scanning by using the sensor to obtain environmental scan image data, collecting the environmental scan image data through SURF features, constructing an image signature, and determining the current positioning point;

[0026] Calculate the approximation between the positioning point and the last positioning point in the STM short-term memory by comparing the number of matching words with the total number of words, and update the weights;

[0027]

[0028] where s(z t , z c ) represents the approximation between the positioning point z t and the last positioning point z c in the STM short-term memory, N p represents the number of matches, and respectively represent the total number of words of z t and z c ; Calculate the closed-loop generation probability of the current positioning point and the positioning point in the WM, evaluate the closed-loop hypothesis, and the posterior probability distribution p(S t |L t ) is calculated as follows:

[0029]

[0030] where L t represents the observation at time t, and the posterior probability distribution p(S t |L t ) represents the posterior probability of the state S t under the condition of the observation sequence L t , θ represents the standard coefficient, p(L t |S t ) is the likelihood probability, representing the probability of observing L t under the condition of the state S t , L t represents the positioning point sequence, t n represents the index of the latest positioning point in the WM in time; p(S t |S t-1 = i) is the state transition probability, representing the probability of transitioning to the state S t at time t under the condition that the state at time t - 1 is i; p(S t-1 = i|L t-1 ) represents the posterior probability that the state at time t - 1 is i under the condition of the observation sequence L t-1 ;

[0031] Calculate the similarity and standard deviation by comparing the new and old positioning points, evaluate the possibility of the new positioning point. If the probability of the closed-loop hypothesis is lower than the preset threshold, select the closed-loop hypothesis with the highest probability, establish a closed-loop link, and update the weights. After detecting the closed-loop, transfer the high-weight positioning points from the LTM back to the WM, update the dictionary, and obtain the real-time environmental data.

[0032] Optionally, in the fourth implementation manner of the first aspect of the present invention, the dynamic path replanning of the predetermined driving path by using the TEB time elastic band algorithm based on the real-time environment data to obtain an obstacle avoidance driving path includes:

[0033] Based on the real-time environment data, obstacle detection is performed. If it is determined that there is an obstacle, the predetermined driving path is converted into discrete-time trajectory points based on the TEB time elastic band algorithm to generate a pose trajectory.

[0034] Let x i represent the i-th pose, and ΔT i represent the time interval between x i and x i+1 . In the space coordinate system, the expression of the pose sequence is: Q = {X i} i i = 0…n, n ∈ N, and the two-pose time series is τ = {ΔT i} i＝0…n-1 ; after the pose sequence and the time sequence are combined, the TEB trajectory information generated is: B := (Q, τ);

[0035] When following the global path and avoiding obstacles, d min，j represents the closest distance between the pose sequence and the obstacle, r pmax represents the maximum distance between the constrained pose point and the global path, r omin represents the minimum distance between the constrained pose point and the obstacle, and ε, S, and n are constant values, which are used as parameters in the penalty function e i to adjust the characteristics of the penalty function. The penalty functions for following the path and avoiding obstacles are as follows:

[0036] f P = e τ (d min，j , r pmax , ε, S, n)

[0037] f ab = e τ (-d min，j , -r omin , ε, S, n)

[0038] f P is used to evaluate the performance during the process of following the global path. The larger the value, the worse the path following effect; f ab is used to evaluate the effect during the obstacle avoidance process. The larger the value, the worse the obstacle avoidance effect;

[0039] The obstacle avoidance constraint represents speed and acceleration. v i , ω i respectively represent the average linear velocity and angular velocity obtained by calculating the time interval between adjacent poses. vmax represents the maximum allowable linear velocity, which limits the maximum linear movement speed of the moving object, ω max represents the maximum allowable angular velocity, which limits the maximum rotational speed of the moving object; at this time, the penalty function is as follows:

[0040]

[0041] Among them, represents the penalty function of the linear velocity, which is a function of v i , v max , ε, S, n, and is used to evaluate whether the linear velocity meets the requirements. The larger the value, the more unreasonable the linear velocity; represents the penalty function of the angular velocity, which is related to ω i , ω max , ε, S, n, and is used to evaluate whether the angular velocity meets the requirements. The larger the value, the more unreasonable the angular velocity; since the local pose state moves in an arc, the cost function related to the heading angle θ i , θ i+1 is:

[0042]

[0043] Among them, θ i and θ i+1 represent the heading angles in the space coordinate system, which are used to describe the direction information of the moving object in different poses; d i represents the distance parameter between adjacent poses, and ΔT i represents the minimum time interval;

[0044] Using the objective function The fastest path is determined through the square of the minimum time interval to obtain the obstacle avoidance driving path.

[0045] Optionally, in the fifth implementation manner of the first aspect of the present invention, driving according to the obstacle avoidance driving path, obtaining the remaining power of the RGV rail vehicle, and using the charging demand decision model to judge the remaining power. If the remaining power can support the RGV rail vehicle to drive to the target transportation location, continue driving according to the obstacle avoidance driving path, including:

[0046] Driving according to the obstacle avoidance driving path, and using the method of weighted fusion to fuse the voltage V, the current integrated power ΔE int , the internal resistance R and the temperature T to estimate the remaining power E remain ;

[0047] E remain = ω1f V (V)+ω2ΔE int +ω3fR (R) + ω4f T (T)

[0048] where ω1 + ω2 + ω3 + ω4 = 1, ω1, ω2, ω3, and ω4 are the weight coefficients of the respective parameters, and f V (V) is the power quantity estimation function based on voltage, and f R (R) is the power quantity correction function based on internal resistance, and is the power quantity correction function based on temperature, to obtain the driving energy consumption of the RGV rail vehicle;

[0049] E consume = (k1m + k2v 2 + k3sinθ + k4) + (1 + α(T - T0) + β(H - H0)) × γ(Z)

[0050] where m is the load weight of the RGV rail vehicle, v is the driving speed, θ is the track gradient, T is the ambient temperature, H is the humidity, k1, k2, k3, and k4 are the basic energy consumption model coefficients, α and β are the influence coefficients of temperature and humidity on energy consumption respectively, T0 and H0 are the standard temperature and humidity, and γ(Z) is the task priority adjustment coefficient. For an urgent task, γ(Z) < 1, indicating that the energy consumption standard can be appropriately reduced to prioritize task completion; for a non-urgent task, γ(Z) ≥ 1;

[0051] Based on reinforcement learning, a charging demand decision-making model is established. Define the state space S = {x, v, E remain , T, H, Z}, where x is the position of the RGV rail vehicle; define the action space A = {charging, continuing to drive}; define the reward function R(s, a),

[0052]

[0053] where E charge is the power required for charging, t charge is the charging time, c1 and c2 are the penalty coefficients related to charging, E consume is the driving energy consumption of the RGV rail vehicle, t delay is the task delay time that may be caused by insufficient power, and c3 and c4 are the reward and penalty coefficients related to continuing to drive;

[0054] By continuously trying different actions a, update the policy according to the reward function R(s, a) to maximize the long-term cumulative reward, and obtain the optimal charging demand decision. If the remaining power can support the RGV rail vehicle to drive to the target transportation location, continue to drive along the obstacle avoidance driving path.

[0055] Optionally, in the fifth implementation manner of the first aspect of the present invention, if the remaining power cannot support the RGV rail vehicle to travel to the target transportation location, the nearest charging location is obtained, a charging travel path is generated according to the charging location, and the vehicle travels based on the charging travel path, including:

[0056] If the remaining power cannot support the RGV rail vehicle to travel to the target transportation location, the nearest charging location is obtained. Let the current position of the RGV rail vehicle be (x rgv , y rgv ), and the coordinates of the i-th charging location be (x p , y p ), and the path distance after track layout and travel restrictions is d i ;

[0057] Use the A algorithm to plan the shortest charging path. The calculation formula for the total cost f(n) of node n is: f(n) = g(n) + h(n), where g(n) is the actual path cost from the starting point to node n, and h(n) is the heuristic estimated cost from node n to the target charging location; h(n) = |x n - x p | + |y n - y p |, and take d i as the target charging location;

[0058] Let the nodes on the path be P0(x0, y0), P1(x1, y1), ···, P n (x n , y n ). The expression of the cubic spline curve in each interval (x p , x p+1 ) is:

[0059] S p (x) = a p + b p (x - x p ) + c p (x - x p ) 2 + d p (x - x p ) 2

[0060] where a p , b p , c p , d p are coefficients to be determined, and need to meet the following conditions:

[0061] The curve passes through the nodes: S p (xp ) = y p , S p (x p+1 ) = y p+1 ;

[0062] The first derivative of the curve is continuous at the node: S' p (x p+1 ) = S' p+1 (x p+1 );

[0063] The second derivative of the curve is continuous at the node: S'' p (x p+1 ) = S'' p+1 (x p+1 );

[0064] By solving the system of equations, the coefficients of the cubic spline curve are obtained, and the smoothed charging path is obtained, that is, the charging driving path is obtained. The charging driving path is input into the RGV rail vehicle and the vehicle travels based on the charging driving path.

[0065] The second aspect of the present invention provides an automatic operation control system for an RGV rail vehicle. The automatic operation control system for an RGV rail vehicle includes a transportation location acquisition module, an environmental data scanning module, a dynamic path planning module, a remaining power judgment module, and a charging path calculation module, wherein:

[0066] The transportation location acquisition module is used to acquire the target transportation location of the RGV rail vehicle, and perform an initial path planning on the target transportation location based on the optimized Dijkstra algorithm to obtain a predetermined driving path;

[0067] The environmental data scanning module is used to travel according to the predetermined driving path, perform real-time environmental scanning through sensors, and analyze the scanned data using the RTAB-Map algorithm to obtain real-time environmental data;

[0068] The dynamic path planning module is used to perform a dynamic path replanning on the predetermined driving path based on the real-time environmental data using the TEB time elastic band algorithm to obtain an obstacle avoidance driving path;

[0069] The remaining power judgment module is used to travel according to the obstacle avoidance driving path, acquire the remaining power of the RGV rail vehicle, and judge the remaining power using the charging demand decision model. If the remaining power can support the RGV rail vehicle to travel to the target transportation location, it continues to travel according to the obstacle avoidance driving path;

[0070] The charging path calculation module is used to, if the remaining power cannot support the RGV rail vehicle to travel to the target transportation location, obtain the nearest charging location, generate a charging travel path based on the charging location, and travel based on the charging travel path.

[0071] Optionally, in the first implementation manner of the second aspect of the present invention, the transportation location acquisition module includes a location acquisition unit, a queue creation unit, a path merging unit, a distance calculation unit, a path update unit, and a path obtaining unit, where:

[0072] The location acquisition unit is used to obtain the target transportation location of the RGV rail vehicle, establish a coordinate system for the travel workshop of the RGV rail vehicle, with the initial location of the RGV rail vehicle as the origin (0, 0), the track direction as the x-axis, and the direction perpendicular to the track as the y-axis, then the two-dimensional coordinates of the target transportation location are (x goal , y goal );

[0073] The queue creation unit is used to optimize the Dijkstra algorithm through bidirectional search. Starting from the starting point s and the target point t respectively, create two priority queues Q s and Q t , which are respectively used to store the nodes to be expanded; initialize the distance d s (s)=0 of the starting point s, the distance d t (t)=0 of the target point t, and the distances of other nodes are infinite. Create two sets S s and S t , which are respectively used to record the nodes that have been expanded from the starting point and the target point;

[0074] The path merging unit is used to select the node u s with the minimum distance from Q s , remove it from Q s and add it to S s . If the node u s is already in S s , it means that the two search directions meet, merge the paths and return;

[0075] The distance calculation unit is used to calculate, for all adjacent nodes V s of the node u s , the distance d s from the starting point through the node u s to the adjacent node v new1 =d s (u s )+ω(u s , v s ), where d s (u s ) represents the distance from the starting point to the node us The shortest distance, ω(u s , v s ) is the weight of the edge (u s , v s ). If d new1 < d s (v s ), then update d s (v s ) = d new1 , and add v s to Q s ;

[0076] Path update unit, used to select the node u t with the minimum distance from Q t , remove it from Q t and add it to S t . If u t is already in S t , it means that the two search directions meet, merge the paths and return. For all adjacent nodes V t of u t , calculate the distance d t from the target point through u t to v new2 = d t (u t ) + ω(u t , v t ), where d t (u t ) represents the shortest distance from the target point to the node, and ω(u t , v t ) is the weight of the edge (u t , v t ). If d new2 < d t (v t ), then update d t (v t ) = d new2 , and add v t to Q t ;

[0077] Path obtaining unit, used to perform step iteration on Q s and Q t until Q s and Q t are empty. When the two search directions meet, let the meeting node be m, the path from the starting point to m be P s , and the path from the target point to m be P t . Reverse P t and combine it with P sMerge to obtain a predetermined driving path.

[0078] Optionally, in the second implementation manner of the second aspect of the present invention, the dynamic path planning module includes:

[0079] An obstacle detection unit, configured to perform obstacle detection based on the real-time environment data. If it is determined that there is an obstacle, the predetermined driving path is converted into discrete-time trajectory points based on the TEB time elastic band algorithm to generate a pose trajectory.

[0080] A trajectory detection unit, configured to set x i to represent the i-th pose, and ΔT i to represent the time interval between x i and x i+1 . In the space coordinate system, the expression of the pose sequence is: Q = {X i} i where i = 0…n, n ∈ N, and the two-pose time series is τ = {ΔT i} i＝0…n-1 ; after merging the pose sequence and the time series, the TEB trajectory information generated is: B := (Q, τ);

[0081] A distance calculation unit, configured to, when following the global path and avoiding obstacles, let d min,j represent the closest distance between the pose sequence and the obstacle, r pmax represent the maximum distance between the constrained pose point and the global path, r omin represent the minimum distance between the constrained pose point and the obstacle, ε, S, and n are constant values, and are used as parameters in the penalty function e τ to adjust the characteristics of the penalty function. The penalty functions for following the path and avoiding obstacles are as follows:

[0082] f P = e τ (d min，j , r pmax , ε, S, n)

[0083] f ab = e τ (-d min，j , -r omin , ε, S, n)

[0084] fP is used to evaluate the performance during the process of following the global path. The larger the value, the worse the path following effect; f ab is used to evaluate the effect during the obstacle avoidance process. The larger the value, the worse the obstacle avoidance effect;

[0085] A parameter calculation unit, configured to calculate the obstacle avoidance constraint representing speed and acceleration, v i , ω irespectively represent the average linear velocity and angular velocity obtained by calculating the time interval between adjacent poses, v max represents the maximum allowable linear velocity, which limits the maximum linear movement speed of the moving object, ω max represents the maximum allowable angular velocity, which limits the maximum rotational speed of the moving object; at this time, the penalty function is as follows:

[0086]

[0087] Among them, represents the penalty function of the linear velocity, which is a function of v i , v max , ε, S, n, and is used to evaluate whether the linear velocity meets the requirements. The larger the value, the more unreasonable the linear velocity; represents the penalty function of the angular velocity, which is a function of ω i , ω max , ε, S, n, and is used to evaluate whether the angular velocity meets the requirements. The larger the value, the more unreasonable the angular velocity; Since the local pose state moves in an arc, the cost function related to the heading angle θ i , θ i+1 is:

[0088]

[0089] Among them, θ i and θ i+1 represent the heading angles in the space coordinate system, which are used to describe the direction information of the moving object in different poses; d i represents the distance parameter between adjacent poses, and ΔT i represents the minimum time interval;

[0090] The path generation unit is used to determine the fastest path by using the objective function squared by the minimum time interval to obtain the obstacle avoidance driving path.

[0091] In the technical solution provided by the present invention, there are the following beneficial effects: 1. Efficient path planning: Through the optimized Dijkstra algorithm for initial path planning, it can quickly and accurately calculate the predetermined driving path for the RGV rail vehicle, greatly improving the planning efficiency in the initial stage of transportation and reducing transportation time loss. 2. Real-time environment adaptation: Use sensors to scan the environment in real time, and analyze the data with the RTAB-Map algorithm to obtain real-time environmental data. Then, based on the TEB time elastic band algorithm, dynamically re-plan the path to generate an obstacle avoidance driving path, enabling the RGV rail vehicle to have the ability to adapt to the environment, effectively avoiding various people and obstacles during operation, ensuring the safety and smoothness of the transportation process, and enhancing the stability of equipment operation. 3. Precise power management: Relying on the charging demand decision model, judge whether it can reach the target transportation location based on the remaining power of the RGV rail vehicle. If the power is insufficient, automatically plan the charging driving path to the nearest charging location, realizing precise power management, avoiding transportation interruption caused by power exhaustion, ensuring the continuity of tasks, and improving the overall production efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0092] By reading the following detailed description of the preferred embodiments, various other advantages and benefits will become clear to those of ordinary skill in the art. The drawings are only for the purpose of showing the preferred embodiments and are not considered to be a limitation of the present invention.

[0093] Figure 1 Schematic diagram of the first embodiment of an automatic operation control method for an RGV rail vehicle provided by an embodiment of the present invention;

[0094] Figure 2 Schematic diagram of a structure of an automatic operation control system for an RGV rail vehicle provided by an embodiment of the present invention;

[0095] Figure 3 Schematic diagram of the second embodiment of an automatic operation control system for an RGV rail vehicle provided by an embodiment of the present invention;

[0096] Figure 4 Schematic diagram of the third embodiment of an automatic operation control system for an RGV rail vehicle provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0097] The terms "first", "second", "third", "fourth", etc. (if any) in the description, claims and the above-mentioned drawings of the present invention are used to distinguish similar objects and do not necessarily describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the term "comprising" or "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, device, product or equipment comprising a series of steps or units does not necessarily limit to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or equipment.

[0098] For ease of understanding, the specific process of the embodiments of the present invention will be described below. Please refer to Figure 1 The first embodiment schematic diagram of an automatic operation control method for an RGV rail vehicle provided by an embodiment of the present invention. The method specifically includes the following steps:

[0099] Step 101: Obtain the target transportation location of the RGV rail vehicle, and perform initial path planning on the target transportation location based on the optimized Dijkstra algorithm to obtain a predetermined driving path;

[0100] Specifically, in this embodiment, the target transportation location of the RGV rail vehicle is obtained, a coordinate system is established for the driving workshop of the RGV rail vehicle. Taking the initial location of the RGV rail vehicle as the origin (0, 0), the track direction as the x-axis, and the direction perpendicular to the track as the y-axis, then the two-dimensional coordinates of the target transportation location are (x goal , y goal );

[0101] Optimize the Dijkstra algorithm through bidirectional search. Starting from the starting point s and the target point t respectively, create two priority queues Q s and Q t , which are respectively used to store the nodes to be expanded; initialize the distance d s (s)=0 of the starting point s, the distance d t (t)=0 of the target point t, and the distances of other nodes are infinite. Create two sets S s and S t , which are respectively used to record the nodes that have been expanded from the starting point and the target point;

[0102] Select the node u s with the smallest distance from Q s , remove it from Q s and add it to S s . If the node u s has already been in S sIt is stated that the two search directions meet, merge the paths, and return.

[0103] For node u s and all its adjacent nodes v s , calculate the distance d s from the starting point, passing through node u s , to the adjacent node v new1 = d s (u s ) + ω(u s , v s ), where d s (u s ) represents the shortest distance from the starting point to node u s , and ω(u s , v s ) is the weight of the edge (u s , v s ). If d new1 < d s (V s ), then update d s (V s ) = d new1 , and add v s to Q s ;

[0104] Select the node u t with the minimum distance from Q t , remove it from Q t and add it to S t . If u t is already in S t , it means that the two search directions meet, merge the paths and return. For all adjacent nodes v t of u t , calculate the distance d t from the target point, passing through u t to v new2 = d t (u t ) + ω(u t , v t ), where d t (u t ) represents the shortest distance from the target point to the node, and ω(u t , v t ) is the weight of the edge (u t , v t ). If d new2 < d t (v t ), then update d t (v t ) = d new2, and add v t to Q t ;

[0105] Perform step iteration on Q s and Q t until Q s and Q t are empty. When the two search directions meet, set the meeting node as m, the path from the starting point to m is P s , and the path from the target point to m is P t . Reverse P t and merge it with P s to obtain a predetermined driving path.

[0106] Step 102: Drive according to the predetermined driving path, perform real-time environment scanning through sensors, and analyze the scanned data using the RTAB-Map algorithm to obtain real-time environment data;

[0107] Specifically, in this embodiment, drive according to the predetermined driving path, and use the MPC predictive control model to establish the dynamic model of the RGV rail vehicle. Then, the linear discrete-time state-space equation of the RGV rail vehicle is:

[0108] x k+1 = Ax k + Bu k

[0109] where x k is the system state vector at time k, x k+1 is the system state vector at time k+1. The system state vector includes at least position, velocity, and acceleration. u k is the control input vector at time k. The control input vector includes at least motor torque and motor speed. A is the state transition matrix, and B is the control input matrix;

[0110] Set the objective function J of the MPC predictive control model:

[0111]

[0112] where x ref,k+i|k is the reference state vector at time k+i, x k+i|k is the predicted state vector at time k. Q is the weight matrix of the state error, R is the weight matrix of the control input, N is the prediction horizon length, u k+i|k is the predicted value of the control input vector at time k for time k+i, including at least motor torque and motor speed, is the transpose of u k+i|k , and T represents vector transpose.

[0113] Perform real-time environmental scanning using sensors, where the sensors at least include an infrared sensor, a millimeter-wave radar, an inertial measurement unit, an image sensor, and a lidar; perform real-time environmental scanning using the sensors to obtain environmental scanning image data, collect the environmental scanning image data through SURF features, construct an image signature, and determine the current positioning point;

[0114] Calculate the approximation degree between the positioning point and the last positioning point in the STM short-term memory by comparing the number of matching words with the total number of words, and update the weight;

[0115]

[0116] Among them, s(z t , z c ) represents the approximation degree between the positioning point z t and the last positioning point z c in the STM short-term memory, N p represents the number of matches, and respectively represent the total number of words of z t and z c ; calculate the closed-loop generation probability of the current positioning point and the positioning point in the WM, evaluate the closed-loop hypothesis, and the posterior probability distribution p(S t |L t ) is calculated as follows:

[0117]

[0118] Among them, L t represents the observation at time t, and the posterior probability distribution p(S t |L t ) represents the posterior probability of the state S t under the condition of the observation sequence L t , θ represents the standard coefficient, p(L t |S t ) is the likelihood probability, which represents the probability of observing L t under the condition of the state S t , L t represents the positioning point sequence, t n represents the index of the latest positioning point in the WM in terms of time; p(S t |S t-1 = i) is the state transition probability, which represents the probability of transitioning to the state S t at time t under the condition that the state at time t-1 is i; p(S t-1 = i|L t-1 ) represents the posterior probability that the state at time t-1 is i under the condition of the observation sequence L t-1 ;

[0119] By comparing the old and new positioning points, calculating the similarity and standard deviation, and evaluating the possibility of the new positioning point. If the probability of the closed-loop hypothesis is lower than the preset threshold, select the closed-loop hypothesis with the highest probability, establish a closed-loop link, and update the weight. After detecting the closed-loop, transmit the high-weight positioning points from the LTM back to the WM, update the dictionary, and obtain the real-time environmental data. WM is the working memory positioning point.

[0120] Step 103: Based on the real-time environmental data, use the TEB (Time Elastic Band) algorithm to dynamically replan the predetermined driving path to obtain an obstacle-avoiding driving path.

[0121] Specifically, in this embodiment, obstacle detection is performed based on the real-time environmental data. If it is determined that there are obstacles, the predetermined driving path is converted into discrete-time trajectory points based on the TEB algorithm to generate a pose trajectory.

[0122] Let x i represent the i-th pose, and ΔT i represent the time interval between x i and x i+1 . In the space coordinate system, the expression of the pose sequence is: Q = {X i} i i = 0…n, n ∈ N, and the two-pose time sequence is τ = {ΔT i} i = 0…n - 1; after merging the pose sequence and the time sequence, the TEB trajectory information generated is: B := (Q, τ).

[0123] When following the global path and avoiding obstacles, d min，j represents the closest distance between the pose sequence and the obstacle, r pmax represents the maximum distance between the constrained pose point and the global path, r omin represents the minimum distance between the constrained pose point and the obstacle, and ε, S, n are constant values, which are used as parameters in the penalty function e i to adjust the characteristics of the penalty function. The penalty functions for following the path and obstacle avoidance constraints are as follows:

[0124] f P = e τ (d min，j , r pmax , ε, S, n)

[0125] f ab = e τ (-d min,j , -r omin , ε, S, n)

[0126] fP is used to evaluate the performance during the process of following the global path. The larger the value, the worse the path following effect; f abIt is used to evaluate the effect during the obstacle avoidance process. The larger the value, the worse the obstacle avoidance effect;

[0127] The obstacle avoidance constraints represent speed and acceleration, v i , ω i respectively represent the average linear velocity and angular velocity obtained by calculating the time interval between adjacent poses. v max represents the maximum allowable linear velocity, which limits the maximum linear movement speed of the moving object. ω max represents the maximum allowable angular velocity, which limits the maximum rotational speed of the moving object; At this time, the penalty function is as follows:

[0128]

[0129] Among them, represents the penalty function of the linear velocity, which is a function of v i , v max , ε, S, and n, and is used to evaluate whether the linear velocity meets the requirements. The larger the value, the more unreasonable the linear velocity; represents the penalty function of the angular velocity, which is related to ω i , ω max , ε, S, and n. The function is used to evaluate whether the angular velocity meets the requirements. The larger the value, the more unreasonable the angular velocity; Since the local pose state moves in an arc, the cost function related to the heading angle θ i , θ i+1 is:

[0130]

[0131] Among them, θ i and θ i+1 represent the heading angles in the space coordinate system, which are used to describe the direction information of the moving object in different poses; d i represents the distance parameter between adjacent poses, and ΔT i represents the minimum time interval;

[0132] Using the objective function The fastest path is determined through the square of the minimum time interval to obtain the obstacle avoidance driving path.

[0133] Step 104: Drive according to the obstacle avoidance driving path, obtain the remaining power of the RGV rail vehicle, and use the charging demand decision model to judge the remaining power. If the remaining power can support the RGV rail vehicle to drive to the target transportation location, continue driving according to the obstacle avoidance driving path;

[0134] Specifically, in this embodiment, when driving according to the obstacle avoidance driving path, the voltage V, the integrated current power ΔE int , the internal resistance R, and the temperature T are fused by a weighted fusion method to estimate the remaining power Eremain ;

[0135] E remain = ω1f V (V)+ω2ΔE int +ω3f R (R)+ω4f T (T)

[0136] where ω1 + ω2 + ω3 + ω4 = 1, ω1, ω2, ω3, and ω4 are the weight coefficients of each parameter, and f V (V) is the power consumption estimation function based on voltage, and f R (R) is the power consumption correction function based on internal resistance, and it is the power consumption correction function based on temperature, to obtain the driving energy consumption of the RGV rail vehicle;

[0137] E consume = (k1m + k2v 2 + k3sinθ + k4)+(1 + α(T - T0)+β(H - H0))×γ(Z)

[0138] where m is the load weight of the RGV rail vehicle, v is the driving speed, θ is the track gradient, T is the ambient temperature, H is the humidity, k1, k2, k3, and k4 are the basic energy consumption model coefficients, α and β are the influence coefficients of temperature and humidity on energy consumption respectively, T0 and H0 are the standard temperature and humidity, and γ(Z) is the task priority adjustment coefficient. For urgent tasks, γ(Z) < 1, indicating that the energy consumption standard can be appropriately reduced to prioritize task completion; for non-urgent tasks, γ(Z) ≥ 1;

[0139] Based on reinforcement learning, a charging demand decision model is established. Define the state space S = {x, v, E remain , T, H, Z}, where x is the position of the RGV rail vehicle; define the action space A = {charge, continue to drive}; define the reward function R(s, a),

[0140]

[0141] where E charge is the power required for charging, t charge is the charging time, c1 and c2 are the penalty coefficients related to charging, E consume is the driving energy consumption of the RGV rail vehicle, t delay is the task delay time that may be caused by insufficient power, and c3 and c4 are the reward and penalty coefficients related to continuing to drive;

[0142] By continuously trying different actions a, the policy is updated according to the reward function R(s, a) to maximize the long-term cumulative reward, and the optimal charging demand decision is obtained. If the remaining power can support the RGV rail vehicle to travel to the target transportation location, it continues to travel along the obstacle avoidance driving path.

[0143] Step 105: If the remaining power cannot support the RGV rail vehicle to travel to the target transportation location, obtain the nearest charging location, generate a charging driving path according to the charging location, and drive based on the charging driving path.

[0144] Specifically, in this embodiment, if the remaining power cannot support the RGV rail vehicle to travel to the target transportation location, obtain the nearest charging location. Let the current position of the RGV rail vehicle be (x rgv , y rgv ), and the coordinates of the i-th charging location be (x p , y p ), and the path distance after track layout and driving restrictions is d i ;

[0145] Use the A algorithm to plan the shortest charging path. The calculation formula for the total cost f(n) of node n is: f(n) = g(n) + h(n), where g(n) is the actual path cost from the starting point to node n, and h(n) is the heuristic estimated cost from node n to the target charging location; h(n) = |x n - x p | + |y n - y p |, and take d i as the target charging location;

[0146] Let the nodes on the path be P0(x0, y0), P1(x1, y1), ···, P n (x n , y n ), and the expression of the cubic spline curve in each interval (x p , x p+1 ) is:

[0147] S p (x) = a p + b p (x - x p ) + c p (x - x p ) 2 + d p (x - x p ) 2

[0148] Among them, a p , b p , cp , d p is a coefficient to be determined and needs to meet the following conditions:

[0149] The curve passes through the node: S p (x p ) = y p , S p (x p+1 ) = y p+1 ;

[0150] The first derivative of the curve is continuous at the node: S' p (x p+1 ) = S' p+1 (x p+1 )

[0151] The second derivative of the curve is continuous at the node: S'' p (x p+1 ) = S'' p+1 (x p+1 )

[0152] By solving the system of equations, the coefficients of the cubic spline curve are obtained, and the smoothed charging path is obtained, that is, the charging driving path is obtained. The charging driving path is input into the RGV rail vehicle and it travels based on the charging driving path.

[0153] In the technical solution provided by the present invention, there are the following beneficial effects: 1. Efficient path planning: Through the optimized Dijkstra algorithm for initial path planning, it can quickly and accurately calculate the predetermined driving path for the RGV rail vehicle, greatly improving the planning efficiency in the initial stage of transportation and reducing transportation time loss. 2. Real-time environment adaptation: The environment is scanned in real time by sensors, and the RTAB-Map algorithm is used to analyze the data to obtain real-time environment data. Then, based on the TEB time elastic band algorithm, the path is dynamically re-planned to generate an obstacle avoidance driving path, enabling the RGV rail vehicle to have the ability to adapt to the environment, effectively avoiding various people and obstacles during operation, ensuring the safety and smoothness of the transportation process, and improving the stability of equipment operation. 3. Precise power management: With the charging demand decision model, it is judged whether it can reach the target transportation location based on the remaining power of the RGV rail vehicle. If the power is insufficient, the charging driving path to the nearest charging location is automatically planned, realizing precise power management, avoiding transportation interruption caused by power exhaustion, ensuring the continuity of tasks, and improving the overall production efficiency.

[0154] Please refer to Figure 2 , a structural schematic diagram of an automatic operation control system for an RGV rail vehicle provided by an embodiment of the present invention. The automatic operation control system for an RGV rail vehicle includes a transportation location acquisition module, an environmental data scanning module, a dynamic path planning module, a remaining power judgment module, and a charging path calculation module, where:

[0155] 201. The transportation location acquisition module is used to acquire the target transportation location of the RGV rail vehicle, and perform initial path planning on the target transportation location based on the optimized Dijkstra algorithm to obtain a predetermined driving path;

[0156] Specifically, this embodiment includes the following units:

[0157] The location acquisition unit is used to acquire the target transportation location of the RGV rail vehicle, establish a coordinate system for the driving workshop of the RGV rail vehicle, with the initial location of the RGV rail vehicle as the origin (0, 0), the track direction as the x-axis, and the direction perpendicular to the track as the y-axis. Then the two-dimensional coordinates of the target transportation location are (x goal , y goal );

[0158] The queue creation unit is used to optimize the Dijkstra algorithm through bidirectional search. Starting from the start point s and the target point t respectively, create two priority queues Q s and Q t , which are respectively used to store the nodes to be expanded; Initialize the distance d s (s)=0, the distance d t (t)=0, and the distances of other nodes are infinite. Create two sets S s and S t , which are respectively used to record the nodes that have been expanded from the start point and the target point;

[0159] The path merging unit is used to select the node u with the minimum distance from Q s , remove it from Q s and add it to S s . If the node u s is already in S s , it means that the two search directions meet, merge the paths and return;

[0160] The distance calculation unit is used to calculate, for all adjacent nodes v s of the node u s , the distance d s from the start point through the node u s to the adjacent node v s as d new1 =d s (u s )+ω(u s , v s ), where d s (u s ) represents the shortest distance from the start point to the node u s , and ω(u s , v s ) is the edge (us , v s )'s weight. If d new1 < d s (V s ), then update d s (v s ) = d new1 , and add v s to Q s ;

[0161] Path update unit, used to select the node u with the minimum distance from Q t , remove it from Q t and add it to S t . If u t is already in S t , it means that the two search directions meet, merge the paths and return. For all adjacent nodes v t of u t , calculate the distance d t from the target point through u t to v t as d new2 = d t (u t ) + ω(u t , v t ), where d t (u t ) represents the shortest distance from the target point to the node, and ω(u t , v t ) is the weight of the edge (u t , vt). If d new2 < d t (v t ), then update d t (v t ) = d new2 , and add v t to Q t ;

[0162] Path obtaining unit, used to perform step iteration on Q s and Q t until Q s and Q t are empty. When the two search directions meet, set the meeting node as m, the path from the starting point to m as P s , the path from the target point to m as P t , reverse P t and merge it with P s to obtain the predetermined driving path.

[0163] 202. Environmental data scanning module, which is used to travel according to a predetermined driving path, perform real-time environmental scanning through sensors, and analyze the scanned data using the RTAB-Map algorithm to obtain real-time environmental data;

[0164] Specifically, this embodiment includes the following units:

[0165] Dynamics model establishment unit, which is used to travel according to a predetermined driving path and establish the dynamics model of the RGV rail vehicle using the MPC predictive control model. The linear discrete-time state space equation of the RGV rail vehicle is:

[0166] x k+1 = Ax k + Bu k

[0167] Among them, x k is the system state vector at time k, x k+1 is the system state vector at time k + 1. The system state vector includes at least position, speed, and acceleration. u k is the control input vector at time k. The control input vector includes at least motor torque and motor speed. A is the state transition matrix, and B is the control input matrix;

[0168] Objective function setting unit, which is used to set the objective function J of the MPC predictive control model:

[0169]

[0170] Among them, x ref,k+i|k is the reference state vector at time k + i, x k+i|k is the predicted state vector at time k. Q is the weight matrix of the state error, R is the weight matrix of the control input, N is the prediction horizon length, u k+i|k is the predicted value of the control input vector at time k for time k + i, including at least motor torque and motor speed, is the transpose of u k+i|k and T represents vector transpose.

[0171] Sensor scanning unit, which is used to perform real-time environmental scanning using sensors. The sensors include at least infrared sensors, millimeter-wave radars, inertial measurement units, image sensors, and lidars; perform real-time environmental scanning using sensors to obtain environmental scanning image data, collect environmental scanning image data through SURF features, construct an image signature, and determine the current positioning point;

[0172] Positioning point calculation unit, which is used to calculate the approximation degree between the positioning point and the last positioning point in the STM short-term memory by comparing the number of matching words with the total number of words and update the weight;

[0173]

[0174] Among them, s(z t , z c ) represents the approximation degree of the positioning point z t and the last positioning point z c in the STM short-term memory. N p represents the number of matches. and respectively represent the total number of words of z t and z c ; Calculate the closed-loop generation probability of the current positioning point and the positioning point in the WM, evaluate the closed-loop hypothesis, and the posterior probability distribution p(S t |L t ) is calculated as follows:

[0175]

[0176] Among them, L t represents the observation at time t, and the posterior probability distribution p(S t |L t ) represents the posterior probability of the state S t under the condition of the observation sequence L t . θ represents the standard coefficient, and p(L t |S t ) is the likelihood probability, which represents the probability of observing L t under the condition of the state S t . L t represents the positioning point sequence, and t n represents the index of the latest positioning point in the WM in terms of time; p(S t |S t-1 = i) is the state transition probability, which represents the probability of transitioning to the state S t at time t under the condition that the state at time t - 1 is i; p(S t-1 = i|L t-1 ) represents the posterior probability that the state at time t - 1 is i under the condition of the observation sequence L t-1 ;

[0177] The positioning point evaluation unit is used to calculate the similarity and standard deviation by comparing the new and old positioning points, evaluate the possibility of the new positioning point. If the probability of the closed-loop hypothesis is lower than the preset threshold, select the closed-loop hypothesis with the highest probability, establish a closed-loop link, and update the weight. After detecting the closed-loop, transmit the high-weight positioning point from the LTM back to the WM, update the dictionary, and obtain the real-time environmental data. WM is the working memory positioning point.

[0178] 203. The dynamic path planning module is used to dynamically re-plan the predetermined driving path based on the real-time environmental data by using the TEB (Time Elastic Band) algorithm to obtain an obstacle-avoiding driving path.

[0179] Specifically, this embodiment includes the following units:

[0180] The obstacle detection unit is used to detect obstacles based on the real-time environmental data. If it is determined that there are obstacles, the predetermined driving path is converted into discrete-time trajectory points based on the TEB algorithm, and a pose trajectory and a path generation unit are generated, where:

[0181] The trajectory detection unit is used to set x i to represent the i-th pose, and ΔT i to represent the time interval between x i and x i+1 In the space coordinate system, the expression of the pose sequence is: Q = {X i} i i = 0…n, n ∈ N, and the two-pose time sequence is τ = {ΔT i} i＝0…n-1 ; After the pose sequence and the time sequence are combined, the TEB trajectory information generated is: B := (Q, τ);

[0182] The distance calculation unit is used to, when following the global path and avoiding obstacles, let d min,j represent the closest distance between the pose sequence and the obstacle, r pmax represent the maximum distance between the constrained pose point and the global path, r omin represent the minimum distance between the constrained pose point and the obstacle, and ε, S, n are constant values, which are used as parameters in the penalty function e τ to adjust the characteristics of the penalty function. The penalty functions for following the path and avoiding obstacles are as follows:

[0183] f P = e τ (d min，j , r pmax , ε, S, n)

[0184] f ab = e τ (-d min，j , -r omin , ε, S, n)

[0185] fP is used to evaluate the performance during the process of following the global path. The larger the value, the worse the path following effect; f ab is used to evaluate the effect during the obstacle avoidance process. The larger the value, the worse the obstacle avoidance effect;

[0186] The parameter calculation unit is used to calculate the obstacle avoidance constraint representing speed and acceleration, vi and ω i respectively represent the average linear velocity and angular velocity obtained by calculating the time interval between adjacent poses, and v max represents the maximum allowable linear velocity, which limits the maximum linear movement speed of the moving object, and ω max represents the maximum allowable angular velocity, which limits the maximum rotational speed of the moving object; at this time, the penalty function is as follows:

[0187]

[0188] wherein, represents the penalty function of the linear velocity, which is a function of v i , v max , ε, S, and n, and is used to evaluate whether the linear velocity meets the requirements. The larger the value, the more unreasonable the linear velocity; represents the penalty function of the angular velocity, which is related to ω i , ω max , ε, S, and n, and is used to evaluate whether the angular velocity meets the requirements. The larger the value, the more unreasonable the angular velocity; since the local pose state is in an arc motion, the cost function related to the heading angle θ i , θ i+1 is as follows:

[0189]

[0190] wherein, θ i and θ i+1 represent the heading angles in the space coordinate system, which are used to describe the direction information of the moving object in different poses; d i represents the distance parameter between adjacent poses, and ΔT i represents the minimum time interval;

[0191] The path generation unit is used to determine the fastest path by using the objective function squared by the minimum time interval to obtain the obstacle avoidance driving path.

[0192] 204. The remaining power judgment module is used to drive according to the obstacle avoidance driving path, obtain the remaining power of the RGV rail vehicle, and judge the remaining power by using the charging demand decision model. If the remaining power can support the RGV rail vehicle to drive to the target transportation location, it will continue to drive according to the obstacle avoidance driving path;

[0193] Specifically, the present embodiment includes the following units:

[0194] The power estimation unit is used to drive according to the obstacle avoidance driving path, and fuse the voltage V, the current integral power ΔE int , the internal resistance R, and the temperature T by using a weighted fusion method to estimate the remaining power Eremain ;

[0195] E remain = ω1f V (V)+ω2ΔE int +ω3f R (R)+ω4f T (T)

[0196] where ω1 + ω2 + ω3 + ω4 = 1, ω1, ω2, ω3, and ω4 are the weight coefficients of each parameter, and f V (V) is the power consumption estimation function based on voltage, and f R (R) is the power consumption correction function based on internal resistance, and is the power consumption correction function based on temperature, to obtain the driving energy consumption of the RGV rail vehicle;

[0197] E consume = (k1m + k2v 2 + k3sinθ + k4)+(1 + α(T - T0)+β(H - H0))×γ(Z)

[0198] where m is the load weight of the RGV rail vehicle, v is the driving speed, θ is the track gradient, T is the ambient temperature, H is the humidity, k1, k2, k3, and k4 are the basic energy consumption model coefficients, α and β are the influence coefficients of temperature and humidity on energy consumption respectively, T0 and H0 are the standard temperature and humidity, and γ(Z) is the task priority adjustment coefficient. For urgent tasks, γ(Z) < 1, indicating that the energy consumption standard can be appropriately reduced to give priority to task completion; for non-urgent tasks, γ(Z) ≥ 1;

[0199] The model establishment unit is used to establish a charging demand decision model based on reinforcement learning, define the state space S = {x, v, E remain , T, H, Z}, where x is the position of the RGV rail vehicle; define the action space A = {charge, continue to drive}; define the reward function R(s, a),

[0200]

[0201] where E charge is the power required for charging, t charge is the charging time, c1 and c2 are the penalty coefficients related to charging, E consume is the driving energy consumption of the RGV rail vehicle, t delay is the task delay time that may be caused by insufficient power, and c3 and c4 are the reward and penalty coefficients related to continuing to drive;

[0202] The charging demand decision-making unit is used to continuously try different actions a, update the policy according to the reward function R(s, a) to maximize the long-term cumulative reward, and obtain the optimal charging demand decision. If the remaining power can support the RGV rail vehicle to travel to the target transportation location, it will continue to travel along the obstacle avoidance driving path.

[0203] 205. The charging path calculation module is used to, if the remaining power cannot support the RGV rail vehicle to travel to the target transportation location, obtain the nearest charging position, generate a charging driving path based on the charging position, and travel based on the charging driving path.

[0204] Specifically, the present embodiment includes the following units:

[0205] The remaining power judgment unit is used to judge that if the remaining power cannot support the RGV rail vehicle to travel to the target transportation location, obtain the nearest charging position. Let the current position of the RGV rail vehicle be (x rgv , y rgv ), the coordinates of the i-th charging position be (x p , y p ), and the path distance after track layout and driving restrictions be d i ;

[0206] The path planning unit is used to plan the shortest charging path using the A algorithm. The calculation formula for the total cost f(n) of node n is: f(n) = g(n) + h(n), where g(n) is the actual path cost from the starting point to node n, and h(n) is the heuristic estimated cost from node n to the target charging position; h(n) = |x n - x p | + |y n - y p |, and take d i as the target charging position;

[0207] Let the nodes on the path be P0(x0, y0), P1(x1, y1), ···, P n (x n , y n ). The expression of the cubic spline curve in each interval (x p , x p+1 ) is:

[0208] S p (x) = a p + b p (x - x p ) + c p (x - x p ) 2 + d p (x - x p )2

[0209] Among them, a p , b p , c p , d p are coefficients to be determined and need to satisfy the following conditions:

[0210] The curve passes through the nodes: S p (x p ) = y p , S p (x p+1 ) = y p+1 ;

[0211] The first derivative of the curve is continuous at the nodes: S' p (x p+1 ) = S' p+1 (x p+1 );

[0212] The second derivative of the curve is continuous at the nodes: S'' p (x p+1 ) = S'' p+1 (x p+1 );

[0213] The path obtaining unit is used to obtain the coefficients of the cubic spline curve by solving the system of equations, obtain the smoothed charging path, that is, obtain the charging driving path, input the charging driving path into the RGV rail vehicle, and drive based on the charging driving path.

[0214] In the technical solution provided by the present invention, through the transportation location acquisition module, the target transportation location of the RGV rail vehicle is acquired, and based on the optimized Dijkstra algorithm, initial path planning is performed on the target transportation location to obtain a predetermined driving path; the environmental data scanning module travels according to the predetermined driving path, performs real-time environmental scanning through sensors, and analyzes the scanned data using the RTAB-Map algorithm to obtain real-time environmental data; the dynamic path planning module re-plans the predetermined driving path dynamically using the TEB time elastic band algorithm based on the real-time environmental data to obtain an obstacle avoidance driving path; the remaining power judgment module is used to travel according to the obstacle avoidance driving path, acquire the remaining power of the RGV rail vehicle, and judge the remaining power using the charging demand decision model. If the remaining power can support the RGV rail vehicle to travel to the target transportation location, continue to travel according to the obstacle avoidance driving path; the charging path calculation module is used to, if the remaining power cannot support the RGV rail vehicle to travel to the target transportation location, acquire the nearest charging location, generate a charging driving path according to the charging location, and travel based on the charging driving path. 1. It can quickly and accurately calculate a predetermined driving path for the RGV rail vehicle, greatly improving the planning efficiency in the initial stage of transportation and reducing transportation time loss. 2. It can effectively avoid various people and obstacles during operation, ensure the safety and smoothness of the transportation process, and improve the stability of equipment operation. 3. It realizes precise management of power, avoids transportation interruption caused by power exhaustion, ensures the continuity of tasks, and improves overall production efficiency.

[0215] Please refer to Figure 3 , the schematic diagram of the second embodiment of an automatic operation control system for an RGV rail vehicle provided by an embodiment of the present invention. The transportation location acquisition module includes a location acquisition unit, a queue creation unit, a path merging unit, a distance calculation unit, a path update unit, and a path obtaining unit, where:

[0216] 2011. The location acquisition unit is used to acquire the target transportation location of the RGV rail vehicle, establish a coordinate system for the driving workshop of the RGV rail vehicle, take the initial location of the RGV rail vehicle as the origin (0, 0), the track direction as the x-axis, and the direction perpendicular to the track as the y-axis. Then the two-dimensional coordinates of the target transportation location are (x goal , y goal );

[0217] 2012. The queue creation unit is used to optimize the Dijkstra algorithm through bidirectional search. Starting from the starting point s and the target point t respectively, create two priority queues Q s and Q t , which are respectively used to store nodes to be expanded; initialize the distance d s (s) = 0, and the distance d of the target point tt (t) = 0, the distances of other nodes are infinite, and two sets S are created. s and S t , used to record the nodes that have been expanded from the starting point and the target point respectively;

[0218] 2013, Path merging unit, used to s Select the node u with the smallest distance s , and change it from Q s Remove and add S s , if node u s Already in S s In the middle, it means that the two search directions meet, the paths are merged and returned;

[0219] 2014, distance calculation unit, used for node u s All adjacent nodes v s , calculate the time from the starting point through node u s To the adjacent node v s The distance d new1 =d s (u s )+ω(u s , v s ), where d s (u s ) represents the distance from the starting point to node u s The shortest distance, ω(u s , v s ) is the edge (u s , v s ) weight, if d new1 <d s (v s ), then update d s (v s ) = d new1 , and v s Join Q s middle;

[0220] 2015, Path Update Unit, for t Select the node u with the smallest distance t , and change it from Q t Remove and add S t In, if u t Already in S t In the example, the two search directions meet, the paths are merged and returned, and for u t All adjacent nodes V t , calculate the distance from the target point through u t to v t The distance d new2 =dt (u t ) + ω(u t ,v t ), where d t (u t ) represents the shortest distance from the target point to the node, and ω(u t ,v t ) is the weight of the edge (u t ,v t ). If d new2 < d t (v t ), then update d t (v t ) = d new2 , and add v t to Q t ;

[0221] 2016. The path obtaining unit is used to perform step iteration on Q s and Q t until Q s and Q t are empty. When the two search directions meet, let the meeting node be m, the path from the starting point to m be P s , and the path from the target point to m be P t . Reverse P t and merge it with P s to obtain the predetermined driving path.

[0222] The beneficial effects are as follows: 1. Efficient path search: By optimizing the Dijkstra algorithm through bidirectional search, priority queues are created and searched simultaneously from the starting point and the target point, greatly reducing the search space and time complexity. Compared with the traditional unidirectional Dijkstra algorithm, it can find the predetermined driving path faster, significantly improving the efficiency of RGV rail vehicle path planning. It is especially suitable for large and complex driving workshop environments, effectively shortening the task execution cycle and improving the overall production efficiency. 2. Precise path determination: By establishing a coordinate system for the driving workshop, the two-dimensional coordinates of the target transportation location are accurately determined. Combining accurate calculation and update mechanisms for node distances, the high accuracy of path planning is ensured. Whether in a complex track layout or in the presence of multiple target transportation locations, the optimal driving path can be planned for the RGV rail vehicle, avoiding transportation errors or resource waste caused by path deviations. 3. Optimized resource utilization: During the path search process, sets are reasonably used to record the expanded nodes, avoiding repeated searches and effectively utilizing computing resources. At the same time, efficient path planning enables the RGV rail vehicle to travel along the shortest path, reducing energy consumption and equipment wear, extending the service life of the equipment and reducing operating costs, achieving optimized utilization of resources. 4. Flexible path merging: When the two search directions meet, the paths can be quickly merged and returned. This flexible path merging mechanism ensures that in a complex environment, even in the face of various interference factors, the RGV rail vehicle can still quickly obtain a complete and effective predetermined driving path, enhancing the adaptability and robustness of the system.

[0223] Please refer to Figure 4 , the schematic diagram of the third embodiment of an automatic operation control system for an RGV rail vehicle provided by the embodiment of the present invention. The dynamic path planning module includes an obstacle detection unit, a trajectory detection unit, a distance calculation unit, a speed calculation unit,

[0224] 2031. Obstacle detection unit, used to detect obstacles based on real-time environmental data. If it is determined that there are obstacles, the predetermined driving path is converted into discrete-time trajectory points based on the TEB time elastic band algorithm, and a pose trajectory and a path generation unit are generated, where:

[0225] 2032. Trajectory detection unit, used to set x i to represent the i-th pose, and ΔT i to represent the time interval between x i and x i+1 . In the space coordinate system, the expression of the pose sequence is: Q = {X i} i = 0…n, n ∈ N, and the two-pose time sequence is τ = {ΔT i} i＝0…n-1 ; after merging the pose sequence and the time sequence, the TEB trajectory information is generated as: B := (Q, τ);

[0226] 2033. The distance calculation unit is used to make d when following the global path and avoiding obstacles min，j represent the closest distance between the pose sequence and the obstacle, and r pmax represent the maximum distance between the constrained pose point and the global path, and r omin represent the minimum distance between the constrained pose point and the obstacle. ε, S, and n are constant values and are used as parameters in the penalty function e τ to adjust the characteristics of the penalty function. The penalty functions for following the path and obstacle avoidance constraints are as follows:

[0227] f P = e τ (d min，j , r pmax , ε, S, n)

[0228] f ab = e τ (-d min,j , -r omin , ε, S, n)

[0229] fP is used to evaluate the performance during the process of following the global path. The larger the value, the worse the path following effect; f ab is used to evaluate the effect during the obstacle avoidance process. The larger the value, the worse the obstacle avoidance effect;

[0230] 2034. The parameter calculation unit is used to calculate the speed and acceleration represented by the obstacle avoidance constraint, where v i , ω i respectively represent the average linear velocity and angular velocity obtained by calculating the time interval between adjacent poses. v max represents the maximum allowable linear velocity, which limits the maximum linear movement speed of the moving object. ω max represents the maximum allowable angular velocity, which limits the maximum rotational speed of the moving object; At this time, the penalty function is as follows:

[0231]

[0232] Among them, represents the penalty function of the linear velocity and is a function of v i , v max , ε, S, n, and is used to evaluate whether the linear velocity meets the requirements. The larger the value, the more unreasonable the linear velocity; represents the penalty function of the angular velocity and is a function of ω i , ω max , ε, S, n, and is used to evaluate whether the angular velocity meets the requirements. The larger the value, the more unreasonable the angular velocity; Since the local pose state moves in an arc and is related to the heading angle θ i , θ i+1The relevant cost function is as follows:

[0233]

[0234] Among them, θ i and θ i+1 represent the heading angles in the spatial coordinate system, which are used to describe the direction information of the moving object in different poses; d i represents the distance parameter between adjacent poses, and ΔT i represents the minimum time interval;

[0235] 2035. The path generation unit is used to determine the fastest path by using the objective function multiplied by the square of the minimum time interval, and obtain the obstacle avoidance driving path.

[0236] Its beneficial effects are as follows: 1. Timely and efficient obstacle response: Through the obstacle detection unit, accurate detection is carried out based on real-time environmental data. Once an obstacle is detected, the predetermined driving path can be quickly converted into discrete-time trajectory points by means of the TEB time elastic band algorithm, generating pose trajectories and paths, realizing a rapid response to sudden obstacle situations, avoiding collision accidents, and ensuring the safe and smooth transportation process of the RGV rail vehicle. 2. High-precision path planning: Utilize the trajectory detection unit to accurately express the pose sequence and time sequence, combine with the distance calculation unit to accurately calculate multiple dimensions such as the distance between the path and the obstacle and the global path distance, and carefully evaluate the path following and obstacle avoidance effects through the penalty function, so that the generated obstacle avoidance driving path can meet the obstacle avoidance requirements while fitting the global path planning to the greatest extent, improving the accuracy of path planning, and reducing the additional driving distance and time consumption caused by path deviation. 3. Precise motion state control: The speed calculation unit can accurately calculate the linear velocity and angular velocity during obstacle avoidance, and strictly evaluate whether the speed is reasonable through the penalty function. At the same time, combined with the cost function related to the heading angle, comprehensively control the motion state of the RGV rail vehicle. Ensure that the vehicle can meet the speed limit requirements during obstacle avoidance driving and maintain a reasonable driving direction, avoiding unstable operation caused by excessive speed or direction deviation, and improving the stability and reliability of equipment operation.

[0237] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and the descriptions in the specification are only preferred examples of the present invention, and are not used to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. An automatic operation control method for an RGV rail vehicle, characterized in that, The automatic operation control method of the RGV rail vehicle includes the following steps: Obtain the target transportation location of the RGV rail vehicle, and perform initial path planning on the target transportation location based on the optimized Dijkstra algorithm to obtain a predetermined driving path; Drive according to the predetermined driving path, perform real-time environment scanning through sensors, and analyze the scanned data using the RTAB-Map algorithm to obtain real-time environment data; Based on the real-time environment data, use the TEB time elastic band algorithm to perform dynamic path replanning on the predetermined driving path to obtain an obstacle avoidance driving path; Drive according to the obstacle avoidance driving path, obtain the remaining power of the RGV rail vehicle, and use the charging demand decision model to judge the remaining power. If the remaining power can support the RGV rail vehicle to drive to the target transportation location, continue driving according to the obstacle avoidance driving path; If the remaining power cannot support the RGV rail vehicle to drive to the target transportation location, obtain the nearest charging location, generate a charging driving path according to the charging location, and drive based on the charging driving path.

2. The automatic operation control method of an RGV rail vehicle according to claim 1, wherein The obtaining of the target transportation location of the RGV rail vehicle, performing initial path planning on the target transportation location based on the optimized Dijkstra algorithm to obtain a predetermined driving path includes: Obtain the target transportation location of the RGV rail vehicle, establish a coordinate system for the driving workshop of the RGV rail vehicle, with the initial location of the RGV rail vehicle as the origin (0, 0), the direction along the track as the x-axis, and the direction perpendicular to the track as the y-axis. Then the two-dimensional coordinates of the target transportation location are (x goal , y goal ); Optimize the Dijkstra algorithm through bidirectional search. Start from the starting point s and the target point t respectively, and create two priority queues Q s and Q t , which are used to store the nodes to be expanded respectively; Initialize the distance d s (s)=0 of the starting point s, the distance d t (t)=0 of the target point t, and the distances of other nodes are infinite. Create two sets S s and S t , which are used to record the nodes that have been expanded from the starting point and the target point respectively; Select the node u with the smallest distance from Q s and remove it from Q s and add it to S s If the node u s is already in S s it means that the two search directions meet, merge the paths and return; s ​ For node u s For all adjacent nodes v s of it, calculate the distance d s from the starting point through node u s to the adjacent node v new1 = d s (u s ) + ω(u s , v s ), where d s (u s ) represents the shortest distance from the starting point to node u s , and ω(u s , v s ) is the weight of the edge (u s , v s ). If d new1 < d s (v s ), then update d s (v s ) = d new1 , and add v s to Q s ; Select the node u with the minimum distance from Q t , remove it from Q t and add it to S t . If u t is already in S t , it means that the two search directions meet, merge the paths and return. For all adjacent nodes v t of u t , calculate the distance d t from the target point through u t to v t : d new2 = d t (u t ) + ω(u t , v t ), where d t (u t ) represents the shortest distance from the target point to the node, and ω(u t , v t ) is the weight of the edge (u t , v t ). If d new2 < d t (v t ), then update d t (v t ) = d new2 , and add v t to Q t ; For Q s and Q t Perform step iteration until Q s and Q t is empty. When the two search directions meet, set the meeting node as m, the path from the starting point to m is P s , and the path from the target point to m is P t . Reverse P t and merge it with P s to obtain the predetermined driving path.

3. The automatic operation control method of an RGV rail vehicle according to claim 1, characterized in that The driving according to the predetermined driving path, performing real-time environment scanning through sensors, and analyzing the scanned data using the RTAB-Map algorithm to obtain real-time environment data includes: Drive according to the predetermined driving path, and use the MPC predictive control model to establish the dynamic model of the RGV rail vehicle. Then the linear discrete-time state space equation of the RGV rail vehicle is: x k+1 = Ax k + Bu k where x k is the system state vector at time k, and x k+1 is the system state vector at time k + 1. The system state vector includes at least position, velocity, and acceleration. u k is the control input vector at time k. The control input vector includes at least motor torque and motor speed. A is the state transition matrix, and B is the control input matrix; Set the objective function J of the MPC predictive control model: where, x ref,k+i|k is the reference state vector at time k + i, x k+i|k is the predicted state vector at time k, Q is the weight matrix of the state error, R is the weight matrix of the control input, N is the prediction horizon length, u k+i|k is the predicted value of the control input vector at time k for time k + i, and at least includes the motor torque and the motor speed, is the transpose of u k+i|k and T represents the vector transpose.

4. The automatic operation control method of an RGV rail vehicle according to claim 1, characterized in that, The driving according to the predetermined driving path, performing real-time environment scanning through sensors, and analyzing the scanned data using the RTAB-Map algorithm to obtain real-time environment data further includes: Perform real-time environment scanning using sensors. The sensors at least include infrared sensors, millimeter wave radars, inertial measurement units, image sensors, and lidar; perform real-time environment scanning using sensors to obtain environmental scan image data, collect the environmental scan image data through SURF features, construct an image signature, and determine the current positioning point; Calculate the approximation degree between the positioning point and the last positioning point in the STM short-term memory by comparing the number of matching words with the total number of words, and update the weight; Among them, s(z t , z c ) represents the approximation degree of the positioning point z t and the last positioning point z in the STM short-term memory c , N p represents the number of matches, and respectively represent the total number of words of z t and z c ; calculate the closed-loop generation probability of the current positioning point and the positioning point in the WM, evaluate the closed-loop hypothesis, and the posterior probability distribution p(S t |L t ) is calculated as follows: Among them, L t represents the observation at time t, and the posterior probability distribution p(S t |L t ) represents the posterior probability of the state S t under the condition of the observation sequence L t . represents the standard coefficient, and p(L t |S t ) is the likelihood probability, which represents the probability of observing L t under the condition of the state S t . L t represents the sequence of positioning points, and t n represents the index of the latest positioning point in the WM in terms of time; p(S t |S t-1 = i) is the state transition probability, which represents the probability of transitioning to the state S t at time t under the condition that the state at time t - 1 is i; p(S t-1 = i|L t-1 ) represents the posterior probability that the state at time t - 1 is i under the condition of the observation sequence L t-1 . Calculate the similarity and standard deviation by comparing the old and new positioning points, evaluate the possibility of the new positioning point. If the probability of the closed-loop hypothesis is lower than the preset threshold, select the closed-loop hypothesis with the highest probability, establish a closed-loop link, and update the weight. After detecting the closed loop, transfer the high-weight positioning points from the LTM back to the WM, update the dictionary, and obtain the real-time environment data.

5. The automatic operation control method of an RGV rail vehicle according to claim 1, characterized in that, The performing of dynamic path replanning on the predetermined driving path based on the real-time environment data using the TEB time elastic band algorithm to obtain an obstacle avoidance driving path includes: Based on the real-time environmental data, obstacle detection is performed. If an obstacle is judged, based on the TEB (Time Elastic Band) algorithm, the predetermined driving path is converted into discrete-time trajectory points to generate a pose trajectory. Let x i represent the i-th pose, and ΔT i represent the time interval between i x i+1 and i x i . In the space coordinate system, the expression of the pose sequence is: Q = {X i} i＝0···n-1 where i = 0···n, n ∈ N. The two-pose time sequence is τ = {ΔT i} i＝0···n-1 . After merging the pose sequence and the time sequence, the TEB trajectory information generated is: B ∶= (Q, τ); When following the global path and avoiding obstacles, d min,j represents the closest distance between the pose sequence and the obstacle, r pmax represents the maximum distance between the constrained pose point and the global path, r omin represents the minimum distance between the constrained pose point and the obstacle. ε, S, and n are constant values, which are used as parameters in the penalty function e τ to adjust the characteristics of the penalty function. The penalty functions for following the path and obstacle avoidance constraints are as follows: f P = e τ (d min,j , r pmax , ε, S, n) f ab = e τ (-d min,j , -r omin , ε, S, n) f P Used to evaluate the performance during following the global path. The larger the value, the worse the path following effect; f ab Used to evaluate the effect during obstacle avoidance. The larger the value, the worse the obstacle avoidance effect; The obstacle avoidance constraint represents speed and acceleration, v i , ω i respectively represent the average linear velocity and angular velocity obtained by calculating the time interval between adjacent poses, v max represents the maximum allowable linear velocity, which limits the maximum linear movement speed of the moving object, ω max represents the maximum allowable angular velocity, which limits the maximum rotational speed of the moving object; at this time, the penalty function is as follows: Among them, The penalty function representing the linear velocity is a function of v i , v max , ε, S, and n, and is used to evaluate whether the linear velocity meets the requirements. The larger the value, the more unreasonable the linear velocity; The penalty function representing the angular velocity is a function of ω i , ω max , ε, S, and n, and is used to evaluate whether the angular velocity meets the requirements. The larger the value, the more unreasonable the angular velocity; Since the local pose state moves in an arc, the cost function related to the heading angle θ i , θ i+1 is as follows: Among them, θ i and θ i+1 represent the heading angles in the spatial coordinate system, which are used to describe the direction information of the moving object in different poses; d i represents the distance parameter between adjacent poses, and ΔT i represents the minimum time interval; Using the objective function Determine the fastest path through the square of the minimum time interval to obtain the obstacle avoidance driving path.

6. The automatic operation control method of an RGV rail vehicle according to claim 1, characterized in that, Drive according to the obstacle avoidance driving path, obtain the remaining power of the RGV rail vehicle, and use the charging demand decision-making model to judge the remaining power. If the remaining power can support the RGV rail vehicle to drive to the target transportation location, continue driving according to the obstacle avoidance driving path, including: Drive according to the obstacle avoidance driving path, and use the weighted fusion method to fuse the voltage V, the integrated current power ΔE int , the internal resistance R and the temperature T to estimate the remaining power E remain ; E remain = ω1f V (V) + ω2ΔE int + ω3f R (R) + ω4f T (T) where ω1 + ω2 + ω3 + ω4 = 1, ω1, ω2, ω3, and ω4 are the weight coefficients of each parameter, and f V (V) is the power quantity estimation function based on voltage, and f R (R) is the power quantity correction function based on internal resistance and is the power quantity correction function based on temperature, and the driving energy consumption of the RGV rail vehicle is obtained; E consume = (k1m + k2v 2 + k3sinθ + k4) + (1 + α(T - T0) + β(H - H0)) × γ(Z) Where m is the load weight of the RGV rail vehicle, v is the driving speed, θ is the track gradient, T is the environmental temperature, H is the humidity, k1, k2, k3, and k4 are basic energy consumption model coefficients, α and β are the influence coefficients of temperature and humidity on energy consumption respectively, T0 and H0 are the standard temperature and humidity, and γ(Z) is the task priority adjustment coefficient. For urgent tasks, γ(Z) < 1, indicating that the energy consumption standard can be appropriately reduced to give priority to task completion; for non-urgent tasks, γ(Z) ≥ 1. Based on reinforcement learning, a charging demand decision-making model is established, and the state space S = {x, v, E remain , T, H, Z} is defined, where x is the position of the RGV rail vehicle; the action space A = {charge, continue to drive} is defined; the reward function R(s, a) is defined, Among them, E charge is the amount of electricity required for charging, t charge is the charging time, c1 and c2 are penalty coefficients related to charging, E consume is the driving energy consumption of the RGV rail vehicle, t delay is the task delay time that may be caused by insufficient power, and c3 and c4 are reward and penalty coefficients related to continuing to drive; By continuously trying different actions a, update the policy according to the reward function R(s, a) to maximize the long-term cumulative reward, obtain the optimal charging demand decision, and if the remaining power can support the RGV rail vehicle to drive to the target transportation location, continue driving according to the obstacle avoidance driving path.

7. The automatic operation control method of an RGV rail vehicle according to claim 1, characterized in that If the remaining power cannot support the RGV rail vehicle to drive to the target transportation location, obtain the nearest charging location, generate a charging driving path according to the charging location, and drive based on the charging driving path, including: If the remaining power cannot support the RGV rail vehicle to travel to the target transportation location, obtain the nearest charging location. Let the current position of the RGV rail vehicle be (x rgv , y rgv ), and the coordinates of the i-th charging location be (x p , y p ), and the path distance after the track layout and driving restrictions is d i ; Using the A algorithm to plan the shortest charging path, the calculation formula for the total cost f(n) of node n is: f(n) = g(n) + h(n), where g(n) is the actual path cost from the starting point to node n, and h(n) is the heuristic estimated cost from node n to the target charging position; h(n) = |x n - x p | + |y n - y p |, taking d i as the target charging position; Let the nodes on the path be P0(x0, y0), P1(x1, y1), ···, P n (x n , y n ). The expression of the cubic spline curve on each interval (x p , x p+1 ) is: S p f(x) = a p + b p (x - x p ) + c p (x - x p ) 2 + d p (x - x p ) 2 Among them, a p , b p , c p , d p are coefficients to be determined and need to satisfy the following conditions: Curve passes through node: S p (x p ) = y p , S p (x p+1 ) = y p+1 ; The first derivative of the curve is continuous at the node: S' p (x p+1 ) = S' p+1 (x p+1 ); The second derivative of the curve is continuous at the node: S″ p (x p+1 ) = S″ p+1 (x p+1 ); Obtain the coefficients of the cubic spline curve by solving the system of equations, obtain the smoothed charging path, that is, obtain the charging driving path, input the charging driving path into the RGV rail vehicle, and drive based on the charging driving path.

8. An automatic operation control system for an RGV rail vehicle, characterized in that, The automatic operation control system of the RGV rail vehicle includes a transportation location acquisition module, an environmental data scanning module, a dynamic path planning module, a remaining power judgment module, and a charging path calculation module, where: The transportation location acquisition module is used to acquire the target transportation location of the RGV rail vehicle, and perform initial path planning on the target transportation location based on the optimized Dijkstra algorithm to obtain a predetermined driving path. The environmental data scanning module is used to drive according to the predetermined driving path, perform real-time environmental scanning through sensors, and analyze the scanned data using the RTAB-Map algorithm to obtain real-time environmental data. The dynamic path planning module is used to perform dynamic path replanning on the predetermined driving path based on the real-time environmental data using the TEB (Time Elastic Band) algorithm to obtain an obstacle avoidance driving path. The remaining power judgment module is used to drive according to the obstacle avoidance driving path, obtain the remaining power of the RGV rail vehicle, and use the charging demand decision-making model to judge the remaining power. If the remaining power can support the RGV rail vehicle to drive to the target transportation location, continue driving according to the obstacle avoidance driving path. The charging path calculation module is used to obtain the nearest charging location if the remaining power cannot support the RGV rail vehicle to travel to the target transportation location, generate a charging driving path according to the charging location, and drive based on the charging driving path.

9. The automatic operation control method of an RGV rail vehicle according to claim 8, characterized in that, The dynamic path planning module includes a location acquisition unit, a queue creation unit, a path merging unit, a distance calculation unit, a path update unit, and a path obtaining unit, where: A location acquisition unit is used to acquire the target transportation location of the RGV rail vehicle, establish a coordinate system for the driving workshop of the RGV rail vehicle, take the initial location of the RGV rail vehicle as the origin (0, 0), the track direction as the x-axis, and the direction perpendicular to the track as the y-axis. Then the two-dimensional coordinates of the target transportation location are (x goal , y goal ); A queue creation unit, which uses a bidirectional search to optimize the Dijkstra algorithm. Starting from the starting point s and the target point t respectively, two priority queues Q s and Q t are created, which are used to store the nodes to be expanded respectively; initialize the distance d s (s) = 0, the distance d t (t) = 0, and the distances of other nodes are infinite. Create two sets S s and S t to record the nodes that have been expanded from the starting point and the target point respectively; Path merging unit, used to s Select the node u with the smallest distance s , and change it from Q s Remove and add S s , if node u s Already in S s In the middle, it means that the two search directions meet, the paths are merged and returned; A distance calculation unit, which is used for node u s for all adjacent nodes v s , calculates the distance d s from the starting point through node u s to the adjacent node v new1 = d s (u s ) + ω(u s , v s ), where d s (u s ) represents the shortest distance from the starting point to node u s , and ω(u s , v s ) is the weight of the edge (u s , v s ). If d new1 < d s (v s ), then update d s (v s ) = d new1 , and add v s to Q s ; A path update unit, used to select the node u with the minimum distance from Q t , remove it from Q t and add it to S t . If u t is already in S t , it means that the two search directions meet, merge the paths and return. For all adjacent nodes v t of u t , calculate the distance d t = d t from the target point through u t to v new2 = d t (u t ) + ω(u t , v t ), where d t (u t ) represents the shortest distance from the target point to the node, and ω(u t , v t ) is the weight of the edge (u t , v t ). If d new2 < d t (v t ), then update d t (v t ) = d new2 , and add v t to Q t ; A path obtaining unit for Q s and Q t Execute step iteration until Q s and Q t is empty. When the two search directions meet, set the meeting node as m, and the path from the starting point to m is P s , and the path from the target point to m is P t . Reverse P t and merge it with P s to obtain a predetermined driving path.

10. The automatic operation control method of an RGV rail vehicle according to claim 8, characterized in that, The motion dynamic path planning module includes an obstacle detection unit, a trajectory detection unit, a distance calculation unit, a speed calculation unit, and a path generation unit, where: The obstacle detection unit is used to detect obstacles based on the real-time environmental data. If it is determined that there are obstacles, the predetermined driving path is converted into discrete-time trajectory points based on the TEB time elastic band algorithm to generate a pose trajectory and a path generation unit, where: A trajectory detection unit for setting x i denotes the i-th pose, and ΔT i denotes x i and x i+1 The time interval between them. In the space coordinate system, the expression of the pose sequence is: Q = {X i} i = 0···n, n ∈ N. The two-pose time series is τ = {ΔT i} i＝0···n-1 ; After merging the pose sequence and the time series, the TEB trajectory information generated is: B := (Q, τ); A distance calculation unit, which is used to make d when following the global path and avoiding obstacles min,j represent the closest distance between the pose sequence and the obstacle, r pmax represent the maximum distance between the constrained pose point and the global path, r omin represent the minimum distance between the constrained pose point and the obstacle, ε, S, n are constant values, and are used as parameters in the penalty function e τ to adjust the characteristics of the penalty function. The penalty functions for following the path and obstacle avoidance constraints are as follows: f P = e τ (d min,j , r pmax , ε, S, n) f ab = e τ (-d min,j , -r omin , ε, S, n) f P Used to evaluate the performance during following the global path. The larger the value, the worse the path following effect; f ab Used to evaluate the effect during obstacle avoidance. The larger the value, the worse the obstacle avoidance effect; A parameter calculation unit for calculating the obstacle avoidance constraint representation of speed and acceleration, v i , ω i respectively represent the average linear velocity and angular velocity obtained by calculating the time interval between adjacent poses, v max represents the maximum allowable linear velocity, which limits the maximum linear movement speed of the moving object, ω max represents the maximum allowable angular velocity, which limits the maximum rotational speed of the moving object; at this time, the penalty function is as follows: Among them, The penalty function representing the linear velocity is a function of v i , v max , ε, S, and n, which is used to evaluate whether the linear velocity meets the requirements. The larger the value, the more unreasonable the linear velocity; The penalty function representing the angular velocity is a function of ω i , ω max , ε, S, and n, which is used to evaluate whether the angular velocity meets the requirements. The larger the value, the more unreasonable the angular velocity; Since the local pose state moves in an arc, the cost function related to the heading angle θ in the space coordinate system is: i , θ i+1 is as follows: Among them, θ i and θ i+1 represent the heading angles in the spatial coordinate system, which are used to describe the direction information of the moving object in different poses; d i represents the distance parameter between adjacent poses, and ΔT i represents the minimum time interval; A path generation unit for determining the fastest path by using the square of the minimum time interval and obtaining an obstacle avoidance driving path. After passing through the square of the minimum time interval to determine the fastest path, an obstacle avoidance driving path is obtained.

Images