Large bulk cargo equipment path planning method based on risk boundary

Abstract

The invention discloses a large-scale bulk cargo equipment path planning method based on a risk boundary, and relates to the fields of machine learning, path planning and the like. The method comprises the following steps: firstly, acquiring an obstacle in an environment by using a deterministic point cloud, calculating a boundary of a safety area under given risk tolerance to define a risk boundary of the obstacle, and training a neural network sampler through an expert demonstration learning decision strategy; then, a neural cost estimator is constructed to select an optimal information state from samples, a neural network is used to recursively and bidirectionally advance to a starting point and a target, and then the collision risk of an intermediate connection is verified based on quadratic sum optimization. Compared with other methods, the method not only can be applied to path planning of large-scale bulk cargo equipment, but also can be expanded to other fields and scenes, such as unmanned vehicles, unmanned aerial vehicle navigation, environment monitoring and the like.

Description

Technical Field

[0001] The present invention relates to the fields of machine learning, path planning, and in particular to a path planning method for large bulk material equipment based on risk boundaries. Background Art

[0002] Large bulk material equipment is a heavy-duty bulk material handling and transportation machinery system with continuous processing capabilities. It is widely used in port loading and unloading, power plants, mineral resource development, and equipment manufacturing. With the development of intelligent transformation and digital transformation in various industries in my country, smart ports, smart power plants, smart mines, and smart factories have become an inevitable trend in the development of the industrial field. In this context, unattended operation of large bulk material equipment is an important part of improving the automation of the production process and promoting intelligent transformation and digital transformation. In a complex environment such as a smart port involving dense multiple factors such as people, vehicles, objects, and ships, how to enable the system to accurately locate and plan the optimal safe path during cantilever operation, and realize safe and efficient automatic stacking and reclaiming operations, has become a core challenge in the development of unattended systems.

[0003] At present, the path planning scheme for the cantilever of large bulk material equipment is still not mature enough, and most of them use manual experience and semi-automatic methods to locate the material point and plan the material path. Some researchers simplify the bucket wheel machine operating environment model into a convex shape and assume that the uncertainty of the path follows a Gaussian distribution to achieve path planning for autonomous operation. This method makes the path planning problem easier to handle by simplifying the complexity of the environment. However, this simplification may ignore the non-convexity and non-Gaussian properties in actual scenarios such as ports and mines, thereby increasing the possibility of the system encountering safety risks such as collisions during operation. These methods have achieved certain results in closed and single material yard environments. However, in complex and changeable actual application scenarios such as ports and mines, ignoring the non-convexity and non-Gaussian properties of the environment may cause the system to encounter safety risks such as collisions during operation.

[0004] In view of the above problems and practical needs, the present invention proposes a path planning method for large bulk material equipment based on risk boundaries, which aims to plan a near-optimal path with an acceptable collision probability and enable large bulk material equipment to autonomously learn risk-bounded path planning. The method first uses deterministic point cloud maps to identify potential hazards such as mobile machinery, unstable ground, and dynamically changing material storage areas for ports, mines, or other bulk material processing sites, and defines risk boundaries to clarify the risk levels of different areas and their acceptable operating ranges. The Neural Network Sampler (NNS) is trained through expert demonstration to predict the state distribution in the potential subspace of large bulk material equipment, such as the flow of materials, the movement of equipment, etc. Next, a Neural Cost Estimator (NCE) is constructed to filter out the best information state from the sample set, which helps large bulk material equipment select the path with the lowest cost and highest safety among multiple possible paths, while taking into account the efficiency of the operation. By using a neural network recursive algorithm, advancing in both directions from the starting point and the target point, and verifying the collision risk of the intermediate path based on the Sum-of-Squares Optimization (SOS) method, the safety and efficiency of the path are considered more comprehensively, especially in the face of a constantly changing operating environment. Through this process, the present invention can effectively reduce operational risks, optimize equipment maintenance strategies, and improve environmental protection performance, thereby reducing various types of losses while ensuring production safety. In addition, the method performs path planning in an uncertain environment and can provide an effective path solution. Summary of the invention

[0005] The present invention proposes a large bulk material equipment path planning method based on risk boundary, aiming to efficiently plan a path with guaranteed risk tolerance for large bulk material equipment in an uncertain environment.

[0006] The specific implementation steps are as follows:

[0007] Step 1: In an uncertain environment, first use Gazebo to build a bulk material equipment environment, and then define the path planning problem, including the safe space, uncertain dangerous environment, safe initial state, target state, and acceptable risk level.

[0008] First, suppose , Represents the working space for the operation of large bulk material equipment. , represents the static risk area in the workspace. The danger zone is described as a polynomial in coordinate positions as :

[0009] ,

[0010] in, is a polynomial function of obstacles and can describe risk areas of uncertain size, location or geometry.

[0011] set up It is the configuration space of bulk material equipment, where There are obstacles and risk bounded Configuration space. Let and As safe starting configuration and target configuration respectively. The goal of bulk equipment is to find the following risk-bounded path:

[0012] ,

[0013] in Indicates that it contains Configuration and connection in and risky path.

[0014] Based on the above assumptions, the location of the path Given The risk-bounded optimal path planning problem is defined as:

[0015] ,

[0016] Constraints ,

[0017] The cost function is the path length measured by Euclidean distance , Represents the set of all feasible paths.

[0018] The probability constraint representing the collision risk is:

[0019] ,

[0020] Step 2: Risk Contours Map: Create a risk contour map to describe the probability information of the uncertain dangerous environment and divide the environment into safe areas, dangerous areas, and risk areas. Use a neural network to encode the point cloud map and convert it into a potential embedding representation, so that the neural network can learn and understand the environment.

[0021] Will Risk Profile Defined as the following set of states:

[0022] ,

[0023] The deterministic constraints on the safe state set are approximated as:

[0024] ,

[0025] in Indicates that obstacles are The mean of the positions, Indicates that obstacles are The variance under position,

[0026] gather is the original risk profile An inner approximation based on rational polynomials. In the risk bounded state set The resulting path consisting of these states Guaranteed not more than The risk profile is as follows:

[0027] ,

[0028] , is the target state,

[0029] ,

[0030] Risk areas and safe areas are defined by risk contours in the point cloud. The risk area (larger blue sphere) represents the space where the probability of collision with the uncertain dangerous environment is greater than a certain threshold, while the collision probability in the safe area (outside the area) does not exceed the threshold.

[0031] Step 3: Build and train neural networks. First, use RRT-SOS in the environment around bulk equipment to obtain near-optimal solutions given a large number of start and target state pairs. Based on this, two deep neural networks, namely NNS and NCE, are constructed to simulate the behavior of RRT-SOS and generate critical states with information. Through training, NNS is able to extract latent embedding features from point cloud maps and predict the state distribution in the latent subspace. NNS is trained using expert demonstration data to generate informative states to guide path planning. NCE is used to evaluate the cost from the current state to the target state and select the best state from it. The inputs of NNS and NCE are point cloud maps, respectively. , Current Status ,Target and specify risk tolerance NNS outputs the next promising state, and NCE outputs the cost estimate from the current state to the goal. The neural network is trained using mean square error loss as the loss function, and point cloud data is used to represent the three-dimensional environment to provide input information for the neural network.

[0032] Step 3.1: NNS consists of a barrier encoder network (PCnet), and an inference network (Inet). PCnet has four key building blocks for extracting the latent embedding Z, including the Transformer Network (T-Net), CNN, Max Pooling Layer, and Multilayer Perceptron (MLP). T-Net is a mini network that performs certain convolution and fully connected operations on the original sample data.

[0033] Step 3.2: NCE consists of the same encoder and the encoder for cost prediction. Considering the randomness of the neural network system, let the neural network system process a batch of current states , and generate several promising states Then, the NCE forecast reaches The cost required to improve the online planning performance target. Note that the tree will be pruned by eliminating redundant nodes. This method attempts to connect two nodes that are far apart. If a node has a high target cost value, there may be some nodes in the area between the target and itself, resulting in a high probability of failure of the long-distance connection. The node with the lowest cost is searched in iterations, the critical state is selected, and planning time is saved.

[0034] Step 4: Adopt a bidirectional search strategy, combining NNS, NCE, and a risk assessor (RA) bidirectional search path planning method, namely 3D NR-RRT. Start from the starting point and the target, and gradually move toward the middle area. It is used to achieve collision avoidance and find a risk-bounded path in an uncertain environment. The algorithm searches for a rough global path bidirectionally through NNS and NCE. NNS expands the information of the search tree, while NCE helps to find a low-cost tree.

[0035] Step 4.1: RA is used to verify the path The collision risk of a given configuration node Location , if all risk profiles If the inequality in does not hold, the nodes in the risk area are deleted. Then, the collision risk of each edge between two adjacent nodes is verified by the SOS condition. There are three steps to verify the SOS condition using the spotless toolbox:

[0036] Step 1: Export linear trajectory , given the positions of two nodes as the endpoints of the line segment;

[0037] Step 2: List the time Semi-algebraic constraints , and created multiple multiplier variables ,in is a free variable, Polynomial coefficients for constructing and verifying path security;

[0038] Step 3: Mosek solver is used to minimize the objective function , which helps determine whether the state of the expression is primal and dual feasible. If it does not exist, the line segment does not have SOS form and the probability of a collision along it is greater than .

[0039] Step 4.2: Use the 3D NR-RRT algorithm with the input being the starting and target configuration pairs { , }, Risk Tolerance , and the risk profile And its corresponding point cloud First, use PCnet to Encode to Encode In an environment with Initialized to empty. Create a forward node set containing the start state and the set of backward nodes with the target state . and The updating process of is the same. For each iteration, a batch of samples is generated by NNS. As the local starting point, NCE predicts Choose the one with the lowest cost. and add it to If and The connection of the top node in the sequence satisfies the SOS condition, then we consider that we have found a node containing and Rough global path of the sample . If not, returns an empty path.

[0040] In getting Then, use Lazy States Contraction (LSC) to remove The redundant nodes in the The non-adjacent nodes of the two ends are used to form a shortcut segment. If the shortcut segment is located in the safe area, the nodes originally located between the two ends are removed. The remaining nodes may become part of the final solution. Then, the RA method is used to evaluate the path And get the success mark , risk edge index , and the global path of the secure node .

[0041] If the path succeeds and the risk is bounded, return the path Otherwise, use Replan to replan the risk segment and use LSC and RA methods to update the path. If a feasible path is found, return the updated path. .

[0042] This method designs a path planning method for large bulk material equipment based on risk boundaries, which is used to solve the path planning of large bulk material equipment under uncertain environment. Compared with the prior art, the technical solution of the present invention has the following beneficial effects:

[0043] 1. The present invention builds a simulation environment for large-scale bulk material equipment, obtains obstacles in the environment through point cloud data, calculates the boundaries of the safe area under a given risk tolerance, and finally defines the risk boundaries of the obstacles. It can perform reasonable path planning under the premise of ensuring safety, overcomes the limitations of traditional methods, and thus realizes safe and collision-free autonomous operation of large-scale bulk material equipment in uncertain environments, reducing the risk of safety accidents.

[0044] 2. Through the expert demonstration learning decision-making strategy of deep neural network, it can imitate the choice and response behavior of experts in different environments and situations, guide path planning in new and unknown situations, and handle obstacles of various shapes and sizes through the 3D NR-RRT algorithm, improve the efficiency and quality of path planning in complex and dynamic environments, enhance generalization, robustness and adaptability, so as to better adapt to complex and changeable practical application scenarios, especially in unpredictable and highly changeable environments such as ports and mines.

[0045] 3. A path planning method for large bulk material equipment based on risk boundaries mainly uses machine learning and neural networks to plan safe and efficient paths. This method is able to handle the uncertainty and complexity of the environment and can be extended to applications in areas such as unmanned vehicles, robot navigation, and abnormal environment monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is an overall flow chart of an embodiment of the present invention. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0048] See attached Figure 1,This embodiment provides a path planning method for large bulk material equipment facing risk boundaries in an uncertain environment. Taking a bucket wheel stacker reclaimer robot arm as an example, it specifically includes the following processes: building a bucket wheel stacker reclaimer robot arm in a dock environment, creating a risk contour map to describe the probability information of an uncertain dangerous environment, building a neural network sampler and a neural cost estimator to simulate RRT-SOS and generate an informed critical state, and using 3D NR-RRT to find a risk-bounded path.

[0049] The specific facts are as follows:

[0050] Implementation step 1: In a deep learning system, the interaction between the agent and the environment constitutes the learning process. In this embodiment, the environment is set as a medium or large port, and the agent is the bucket wheel stacker cantilever. The environment is determined and the action space of the bucket wheel cantilever is defined as follows.

[0051] Step 1.1: Build the bucket wheel stacker cantilever environment and set , Represents the working space of bucket wheel stacker reclaimer boom operation. , represents the static risk area in the workspace. The danger zone is described as a polynomial in coordinate positions as :

[0052] ,

[0053] in, is a given polynomial that can describe risk areas of uncertain size, location or geometry.

[0054] Step 1.2: Define the path planning problem. It is the cantilever configuration space of bucket wheel stacker reclaimer, where There are obstacles and risk bounded Configuration space. Let and They are respectively used as safe startup configuration and target configuration. The goal is to find the following risk-bounded path:

[0055] ,

[0056] in Indicates that it contains Configuration and connection in and According to the above assumptions, the location of the path Given The risk-bounded optimal path planning problem is defined as:

[0057] ,

[0058] Constraints

[0059] The cost function is the path length measured by Euclidean distance , Represents the set of all feasible paths.

[0060] The probability constraint representing the collision risk is:

[0061] ,

[0062] Implementation step 2: Create a risk contour map to describe the probability information of the uncertain dangerous environment of the port, and divide the environment into safe area, dangerous area and risk area. Among them, the safe area is the area where the bucket wheel machine can operate safely, the dangerous area is the area where entry is prohibited, and the risk area is the area where there is a certain safety risk but it may be necessary to enter. Use a neural network to encode the point cloud map and convert it into a potential embedding representation, so that the neural network can learn and understand the environment.

[0063] Will Risk Profile Defined as the following set of states:

[0064] ,

[0065] Then, the deterministic constraint on the set of safe states is approximated as:

[0066] ,

[0067] in Indicates that obstacles are The mean of the positions, Indicates that obstacles are The variance under position,

[0068] gather is the original risk profile An inner approximation based on rational polynomials. In the risk bounded state set The resulting path consisting of these states Guaranteed not more than The risk profile is as follows:

[0069] ,

[0070] , is the target state,

[0071] ,

[0072] Risk areas and safe areas are defined by risk contours in the point cloud. The risk area (larger blue sphere) represents the space where the probability of collision with the uncertain dangerous environment is greater than a certain threshold, while the collision probability in the safe area (outside the area) does not exceed the threshold.

[0073] Implementation step 3: First, use RRT-SOS in the environment surrounding the bulk material equipment to obtain a near-optimal solution given a large number of start and target state pairs. Based on this, two deep neural networks, namely NNS and NCE, are constructed to simulate the behavior of RRT-SOS and generate critical states with information. Through training, NNS is able to extract latent embedding features from the point cloud map and predict the state distribution in the latent subspace. NNS is trained using expert demonstration data to generate informative states to guide path planning. NCE is used to evaluate the cost from the current state to the target state and select the best state from it. The inputs of NNS and NCE are the point cloud map, respectively. , Current Status ,Target and specify risk tolerance NNS outputs the next promising state, and NCE outputs the cost estimate from the current state to the goal. The neural network is trained using mean square error loss as the loss function, and point cloud data is used to represent the three-dimensional environment to provide input information for the neural network.

[0074] Step 3.1: NNS consists of a barrier encoder network (PCnet), and an inference network (Inet). PCnet has four key building blocks for extracting the latent embedding Z, including the Transformer Network (T-Net), CNN, Max Pooling Layer, and Multilayer Perceptron (MLP). T-Net is a mini network that performs certain convolution and fully connected operations on the original sample data.

[0075] Step 3.2: NCE consists of the same encoder and the encoder for cost prediction. Considering the randomness of the neural network system, let the neural network system process a batch of current states , and generate several promising states Then, the NCE forecast reaches The cost required to improve the online planning performance target. The tree will be pruned by eliminating redundant nodes. This method attempts to connect two nodes that are far apart. If a node has a high target cost value, there may be some nodes in the area between the target and itself, resulting in a high probability of failure of the long-distance connection. The node with the lowest cost is searched in iterations, the critical state is selected, and planning time is saved.

[0076] Implementation step 4: Adopt a bidirectional search strategy, combining NNS, NCE, and Risk Assessor (RA) bidirectional search path planning method, namely 3D NR-RRT. Start from the starting point and the goal, and gradually move towards the middle area. It is used to achieve collision avoidance and find a risk-bounded path in an uncertain environment. The algorithm searches for a rough global path bidirectionally through NNS and NCE. NNS expands the search tree, while NCE helps to find a low-cost tree.

[0077] Step 4.1: RA is used to verify the path The collision risk of a given configuration node Location , if all risk profiles If the inequality in does not hold, the nodes in the risk area are deleted. Then, the collision risk of each edge between two adjacent nodes is verified by the SOS condition. There are three steps to verify the SOS condition using the spotless toolbox:

[0078] Step 1: Export linear trajectory , given the positions of two nodes as the endpoints of the line segment;

[0079] Step 2: List the time Semi-algebraic constraints , and created multiple multiplier variables ,in is a free variable, Polynomial coefficients for constructing and verifying path security;

[0080] Step 3: Mosek solver is used to minimize , which helps determine whether the state of the expression is primal and dual feasible. If it does not exist, the line segment does not have SOS form and the probability of a collision along it is greater than .

[0081] Step 4.2: Use the 3D NR-RRT algorithm with the input being the starting and target configuration pairs { , }, Risk Tolerance , and the risk profile And its corresponding point cloud First, use PCnet to Encode to Encode In an environment with Initialized to empty. Create a forward node set containing the start state and the set of backward nodes with the target state . and The updating process of is the same. For each iteration, a batch of samples is generated by NNS. As the local starting point, NCE predicts Choose the one with the lowest cost. and add it to If and The connection of the top node in the sequence satisfies the SOS condition, then we consider that we have found a node containing and Rough global path of the sample . If not, returns an empty path.

[0082] In getting Afterwards, use LSC to remove The redundant nodes in the The non-adjacent nodes of the two ends are used to form a shortcut segment. If the shortcut segment is located in the safe area, the nodes originally located between the two ends are removed. The remaining nodes may become part of the final solution. Then, the RA method is used to evaluate the path And get the success mark , risk edge index , and the global path of the secure node .

[0083] If the path succeeds and the risk is bounded, return the path Otherwise, use Replan to replan the risk segment and use LSC and RA methods to update the path. If a feasible path is found, return the updated path. .

Claims

1. A large bulk material equipment path planning method based on risk boundary, characterized in that: The following steps are involved: Step 1: In an uncertain environment, first use Gazebo to build a bulk material equipment environment, and then define the path planning problem, including the safe space, uncertain dangerous environment, safe initial state, target state, and acceptable risk level; Step 2: Create a risk contour map to describe the probability information of the uncertain dangerous environment, divide the environment into safe area, dangerous area and risk area, use a neural network to encode the point cloud map, and convert the point cloud map into a potential embedding representation; Step 3: Build and train a neural network. NNS can extract potential embedding features from point cloud images through training and predict the state distribution in the potential subspace. Use NCE to build a neural network to evaluate the cost from the current state to the target state and select the best state. Step 4: Adopt a bidirectional search strategy and combine the bidirectional search path planning method of NNS, NCE, and RA, namely 3D NR-RRT, starting from the starting point and the target and gradually moving toward the middle area, to achieve collision avoidance and find a risk-bounded path in an uncertain environment.

2. A large bulk material equipment path planning method based on risk boundary according to claim 1, characterized in that: In the uncertain environment described in step 1, first use Gazebo to build the environment where the bulk material equipment is located, and then define the path planning problem, including the safe space, uncertain dangerous environment, safe initial state, target state, and acceptable risk level. The specific implementation is as follows: set up , Represents the working space for the operation of large bulk material equipment; The danger zone is described as a polynomial in coordinates as : ， in, , represents the static risk area in the workspace, is a given polynomial describing a risk region of uncertain size, location or geometry; set up It is the configuration space of bulk material equipment, where ; Obstacles and risk bounded Configuration space, let and As secure startup configuration and target configuration respectively; The goal of bulk equipment is to find the following risk-bounded path: ， in Indicates that it contains Configuration and connection in and risky paths; The location of the path Given , the risk-bounded optimal path planning problem is defined as: ， Constraints , The cost function is the path length measured by Euclidean distance , Represents the set of all feasible paths; The probability constraint representing the collision risk is: ， in Represents a sequence of positions on a path.

3. The large bulk material equipment path planning method based on risk boundary according to claim 1 is characterized in that: The risk contour map described in step 2 is created to describe the probability information of the uncertain dangerous environment, and the environment is divided into safe area, dangerous area and risk area. The point cloud map is encoded using a neural network and converted into a potential embedding representation. The specific implementation is as follows: For each obstacle, an associated risk profile is defined, representing the safe area around the obstacle at that risk level. Risk Profile Defined as the following set of states: ， The deterministic constraints on the safe state set are approximated as: ， in Indicates that obstacles are The mean of the positions, Indicates that obstacles are Variance under location; gather is the original risk profile An inner approximation based on rational polynomials, in the risk-bounded state set The aspect has a deterministic constraint, and the resulting path consisting of these states Guaranteed not more than The chance of a collision is as follows: ， , is the target state, 。 4. The large bulk material equipment path planning method based on risk boundary according to claim 1 is characterized in that: The bidirectional search strategy described in step 4 combines the bidirectional search path planning method of NNS, NCE, and RA, namely 3DNR-RRT, starting from the starting point and the target and gradually moving toward the middle area, is used to achieve collision avoidance and find a risk-bounded path in an uncertain environment. The specific implementation is as follows: RA is used to verify the path The collision risk of a given configuration node Location , the SOS condition is used to verify the collision risk of each edge between two critical points. There are three steps to verify the SOS condition using the spotless toolbox: Step 1: Substitution - Given the positions of the two nodes as the endpoints of the line segment, substitute the trajectory into For each item, get the time Uncertain polynomial , and derive the linear trajectory: ； Step 2: List the time Semi-algebraic constraints , and create multiple multiplier variables: ， in is a free variable, Polynomial coefficients for constructing and verifying path security; Step 3: Use Mosek solver for minimization ; Using the 3D NR-RRT algorithm, the input is the starting and target configuration pairs { , }, Risk Tolerance , and the risk profile And its corresponding point cloud All interior approximations of First, use PCnet to Encode to Encode In an environment with Initialize to empty, create a forward node set containing the start state and the set of backward nodes with the target state ; and The updating process is the same; for each iteration, a batch of samples is generated by NNS As the local starting point, NCE predicts cost; choose the one with the lowest cost and add it to In; if and The connection of the top node in the sequence satisfies the SOS condition, then we consider that we have found a node containing and Rough global path of the sample ; if not, returns an empty path, In getting Afterwards, use LSC to remove The redundant nodes in the The non-adjacent nodes of are used to form a shortcut segment. If the shortcut segment is located in the safe area, the nodes originally located between the two ends are removed, and the remaining nodes become part of the final solution. The RA method is used to evaluate the path And get the success mark , risk edge index , and the global path of the secure node , If the path succeeds and the risk is bounded, return the path Otherwise, use Replan to replan the risk segment and use LSC and RA methods to update the path. If a feasible path is found, return the updated path .

Images