Adaptive path planning and plume tracking method for underway monitoring

Abstract

The application discloses an adaptive path planning and pollution plume tracking method for underway monitoring, and comprises the following steps: a space model of a monitoring area is established by acquiring pose information, motion constraints and a sensor sampling period of a mobile measurement platform; pollution concentration and position information are collected in real time during platform travel to form observation data; a level set evolution method is used to construct and update a boundary representation of a pollution plume based on the observation data to depict the spatial variation of the plume; in combination with the boundary variation information, a rolling time domain optimization algorithm is used to generate a measurement path in a rolling update time range to guide adaptive adjustment of a travel route of the platform; the underway monitoring is completed by executing the path, and boundary tracking results of the pollution plume and corresponding spatial position determination results of pollution sources are output. The application is suitable for mobile monitoring and positioning tasks in complex environments.

Description

Technical Field

[0001] This invention relates to the field of measurement and navigation technology, and in particular to an adaptive path planning and pollution plume tracking method for mobile monitoring. Background Technology

[0002] With the increasing demand for environmental monitoring and ecological governance, mobile monitoring, as an important means of obtaining spatial distribution information of pollutants by relying on mobile measurement platforms, has been widely used in scenarios such as air pollution inspection, industrial emission supervision, and emergency monitoring of sudden pollution incidents. Existing mobile monitoring technologies typically guide mobile measurement platforms to collect pollutant concentration and location information within the monitoring area through preset or regularized routes, and analyze and assess the pollution distribution based on this information.

[0003] However, existing mobile monitoring technologies still have significant shortcomings in addressing the dynamic changes in the spatial morphology of pollution plumes. On the one hand, most path planning methods rely on fixed paths or static planning strategies based on historical data, making it difficult to reflect real-time changes in pollution plumes during diffusion, contraction, or drift. This results in insufficient coverage of the effective sampling area and low monitoring efficiency. On the other hand, existing technologies typically use discrete sampling points or simple threshold methods to describe pollution boundaries, making it difficult to accurately characterize the irregular and continuously evolving boundary morphology of pollution plumes. This causes a lag in the response of path adjustment and pollution source localization to plume changes, thereby affecting the accuracy of pollution plume boundary tracking and the determination of pollution source spatial location.

[0004] Therefore, how to provide adaptive path planning and pollution plume tracking methods for mobile monitoring is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] One objective of this invention is to propose an adaptive path planning and pollution plume tracking method for mobile monitoring. This invention employs a level set evolution method to continuously and implicitly model and update the pollution plume boundary, and combines it with a rolling time-domain optimization algorithm to dynamically generate and adjust the measurement path. By controlling a mobile measurement platform to complete mobile monitoring along the path, the invention outputs the boundary tracking results of the pollution plume and the corresponding spatial location determination results of the pollution source. This method can adapt to the dynamic changes in the spatial morphology of the pollution plume and has the advantages of timely path adjustment, continuous boundary characterization, and high efficiency in pollution source localization.

[0006] The adaptive path planning and pollution plume tracking method for mobile monitoring according to embodiments of the present invention includes the following steps: The initial pose, kinematic constraint parameters, and sampling period of the pollutant sensor of the mobile measurement platform are obtained. A spatial coordinate system of the monitoring area is established, and an environmental constraint model including traversable and prohibited boundaries is constructed. Within the passable area defined by the environmental constraint model, pollutant concentration data and corresponding pose data are collected according to the sampling period, and time synchronization and spatial mapping are performed to generate a pollutant concentration observation set. The level set function of the pollution plume is initialized based on the pollutant concentration observation set, an implicit representation is established, and the corresponding evolution parameters are set. The level set evolution method is adopted to update the level set function using the pollutant concentration observation set and evolution parameters, thereby generating an updated implicit representation of the pollution plume boundary and boundary propagation information derived from the level set function; Based on a preset rolling time domain length, a path planning optimization problem is constructed. A rolling time domain optimization algorithm is adopted, taking the implicit representation of the pollution plume boundary and the boundary propagation information as the planning input, and the kinematic constraint parameters and environmental constraint model as the constraint conditions. The optimal control sequence in the rolling time domain is obtained and the measurement path segment is generated. Based on the measurement path segment, the mobile measurement platform is controlled to complete the mobile monitoring task, and outputs the boundary tracking results of the pollution plume and the spatial location determination results of the pollution source.

[0007] Optionally, the construction of the environmental constraint model includes: The initial pose of the mobile measurement platform is obtained, which includes the initial position coordinates, the initial heading angle and the initial timestamp, and the initial pose is used as the reference pose for establishing the coordinate system of the monitoring area. Obtain the kinematic constraint parameters of the mobile measurement platform, including the maximum travel speed, maximum angular velocity, maximum acceleration, and minimum turning radius; The sampling period of the pollutant sensor is obtained, and the consistency configuration of the pose sampling period and the pollutant concentration sampling period is determined based on the sampling period to form a time discrete reference. A monitoring area plane coordinate system is established with the position corresponding to the initial pose as the origin, and the axis of the area plane coordinate system is determined with the initial heading angle, so that the pose data collected subsequently can be directly converted into a sequence of trajectory points under the monitoring area coordinate system. In the plane coordinate system of the monitoring area, a set of restricted areas is generated based on the pre-acquired road boundary, building outline and obstacle distribution data, and the restricted boundary is extracted from the set of restricted areas to constrain the feasible movement space of the mobile measurement platform; In the coordinate system of the monitoring area, based on the set of prohibited areas, kinematic constraint parameters and time discrete reference, the accessibility of the passable space is determined and a set of passable areas is generated. The passable boundaries are extracted from the set of passable areas, and an environmental constraint model containing the passable boundaries and prohibited boundaries is output.

[0008] Optionally, the generation of the pollutant concentration observation set includes: Read the passable and prohibited boundaries in the environmental constraint model, determine the passable area based on the passable boundaries, and generate an area identifier for motion constraint determination within the passable area; The mobile measurement platform is controlled to move within the passable area, and the current pose data of the platform is acquired in each control cycle. Based on the area identifier, it is determined whether the spatial position corresponding to the current pose data falls into the restricted area. When the sampling period of the pollutant sensor is obtained, the pollutant concentration is collected, the pollutant concentration data and its sampling timestamp are recorded, and the pose data corresponding to the sampling timestamp are read. The pose data includes position coordinates, heading angle and pose timestamp, forming a pair of concentration data and pose data with timestamp. When the timestamp of the concentration data is inconsistent with the timestamp of the pose data, the pose data is time-aligned using the adjacent timestamps before and after, so that each piece of concentration data corresponds to a time-aligned pose data, and the time-aligned pose data is written back to the concentration data and pose data pair. In the plane coordinate system of the monitoring area, the time-aligned pose data is converted into corresponding spatial location identifiers, and the concentration data is bound to the spatial location identifiers to generate concentration observation points; Based on the environmental constraint model, the concentration observation points are screened for effectiveness. Concentration observation points whose spatial location markers fall into restricted areas are removed, while concentration observation points that fall into accessible areas are retained and aggregated in chronological order to generate a pollutant concentration observation set.

[0009] Optionally, the establishment and setting of the implicit representation and the corresponding evolution parameters include: Read the pollutant concentration observation set, extract the spatial location identifier, concentration value and corresponding time information of each concentration observation point, and determine the horizontal set calculation domain based on the spatial range under the plane coordinate system of the monitoring area. A spatial discrete grid is established within the computational domain of the level set, and the spatial location identifiers in the pollutant concentration observation set are mapped to the grid cells of the spatial discrete grid to obtain the concentration assignment results corresponding to the grid cells; The initial boundary seed region of the plume is determined based on the concentration assignment result. The initial boundary seed region of the plume is composed of grid cells that meet the preset initialization conditions, and the plume interior region and the plume exterior region are generated from the initial boundary seed region of the plume. Based on the internal and external regions of the plume, a level set function is initialized on the spatial discrete grid. The level set function takes a first sign in the internal region of the plume and a second sign in the external region of the plume, and the first sign is opposite to the second sign. The boundary of the contaminated plume is represented by the set of zero values ​​of the level set function, forming an implicit representation of the contaminated plume. Evolutionary parameters for updating the level set function are set, including time step parameters, grid spatial resolution parameters, and driving parameters determined by the pollutant concentration observation set, wherein the driving parameters are generated by the concentration assignment results and stored corresponding to the spatial discrete grid.

[0010] Optionally, the generation of the updated implicit representation of the contamination plume boundary and the boundary propagation information derived from the level set function includes: It receives the level set computational domain, spatial discrete grid, initialization level set function, and evolution parameters, and reads the pollutant concentration observation set; The spatial location identifiers of the concentration observation points in the pollutant concentration observation set are mapped to the grid cells of the spatial discrete grid. The pollutant concentration observation values ​​are written into the corresponding grid cells to form concentration assignment results. Based on the concentration assignment results, evolution driving information corresponding to the grid cells is generated. Based on evolution-driven information and evolution parameters, the update rule of the level set function is determined on the spatial discrete grid, and the update rule is applied to the initial level set function to form the update state to be advanced; Based on the time step parameter in the evolution parameters, a time advancement operation is performed on the updated state to be advanced to obtain the updated level set function, and the updated level set function is written back to the spatial discrete grid as the input for subsequent processing. Numerical consistency processing is performed on the updated level set function to obtain the consistent level set function, and the consistent level set function is used as the input for boundary extraction and propagation information derivation; Extract the zero-value set from the level set function after consistency processing, determine the zero-value set as the implicit representation of the updated pollution plume boundary, and map the zero-value set to the plane coordinate system of the monitoring area to generate the spatial location of the plume boundary; At the spatial location of the plume boundary, the boundary propagation direction information is calculated based on the level set function after consistency processing, and the boundary propagation rate information is determined based on evolution-driven information and time-progression calculation. The updated implicit representation of the pollution plume boundary, as well as the boundary propagation direction information and the boundary propagation rate information, are output.

[0011] Optionally, the generation of the measurement path segment includes: The system receives the current pose of the mobile measurement platform, kinematic constraint parameters, environmental constraint model, implicit representation of the pollution plume boundary, boundary propagation direction information, and boundary propagation rate information, and uses the current pose as the initial state input for path planning optimization. Read the preset rolling time domain length, generate a discrete time sequence within the rolling time domain based on the rolling time domain length, and use the discrete time sequence to determine the prediction steps and the decision variable dimension of the control sequence within the rolling time domain for the path planning optimization problem; The rolling time domain is divided into at least two continuous sub-time domains based on discrete time-time sequences, and time index intervals corresponding to each sub-time domain are generated. Within the time index interval of each sub-time domain, a corresponding plume boundary prediction representation is generated based on the implicit representation of the pollution plume boundary, as well as the boundary propagation direction information and boundary propagation rate information. The plume boundary prediction representation is then converted into the corresponding sub-time domain constraint structure or sub-time domain target structure, and the piecewise optimization input is output. Based on the piecewise optimization input, candidate measurement corridors are generated in the rolling time domain, and the candidate measurement corridors are decomposed into a set of corridor segments according to the discrete time sequence. The set of corridor segments is used as the spatial constraint input for the construction of the path skeleton. Within the corridor segment set, candidate path skeletons corresponding to discrete time sequences are generated to obtain skeleton point sequences, and these skeleton point sequences are used as input objects for geometric constraint generation. Based on the skeleton point sequence, trajectory following constraints and deviation constraints are constructed, and the trajectory following constraints and deviation constraints are bound to the corresponding corridor segment set according to the time index interval, forming a piecewise geometric constraint set corresponding to the sub-time domain. The piecewise geometric constraint set is output as the path planning optimization constraint input. Using the rolling time domain control sequence as the decision variable for the path planning optimization problem, the state sequence and trajectory point sequence are generated from the rolling time domain control sequence based on the mobile measurement platform. The trajectory point sequence is then substituted into the piecewise geometric constraint set, environmental constraint model and kinematic constraint parameters to form the constraint conditions of the path planning optimization problem. The objective function of the path planning optimization problem is constructed by combining the corridor guidance term, control change term and trajectory smoothing term corresponding to the candidate measurement corridor into the objective function, and the objective function and the constraints together constitute the path planning optimization problem based on the rolling time domain length. An initial control sequence is set for the path planning optimization problem. The optimal control sequence of the previous rolling time domain is time-shifted and aligned to the discrete time sequence to obtain the initial control sequence. The optimal control sequence in the current rolling time domain is obtained by iterative solution. A measurement path segment is generated based on the optimal control sequence, and the feasibility of the measurement path segment is verified based on the environmental constraint model and kinematic constraint parameters.

[0012] Optionally, the output of the boundary tracking results of the pollution plume and the spatial location determination results of the pollution source includes: Receive the measurement path segment, parse the measurement path segment into a sequence of path points arranged in chronological order, and assign a corresponding time index to the sequence of path points to form a path execution sequence; A motion control command sequence is generated based on the path execution sequence, and the motion control command sequence is sent to the mobile measurement platform to control the mobile measurement platform to move sequentially along the path execution sequence; During the execution of the path execution sequence on the mobile measurement platform, the execution pose data of the mobile measurement platform is acquired in real time, and the execution pose data is aligned with the time index of the path execution sequence to form a path execution record; The pollutant concentration is collected according to the sampling cycle of the pollutant sensor, and the pollutant concentration data is obtained. The pollutant concentration data is then correlated with the time-aligned execution pose data to generate a mobile monitoring and observation set. The mobile monitoring observation set is input into the level set evolution process to obtain the updated implicit representation of the pollution plume boundary, and the boundary tracking results of the pollution plume are extracted based on the implicit representation of the pollution plume boundary. Based on the boundary tracking results of the pollution plume, the spatial location of the pollution source corresponding to the boundary tracking results is determined in the plane coordinate system of the monitoring area, and the boundary tracking results of the pollution plume and the spatial location determination results of the pollution source are output.

[0013] The beneficial effects of this invention are: This invention introduces the level set evolution method and rolling time-domain optimization algorithm into the path planning and navigation process of mobile monitoring, realizing continuous modeling, dynamic updating and forward prediction of the spatial boundary of pollution plumes. This enables the mobile measurement platform to adaptively adjust the measurement path according to the evolution trend of pollution plumes, thereby effectively overcoming the problems of fixed path planning and delayed response to changes in the position of pollution plumes in the prior art, and significantly improving the ability of mobile monitoring to track complex and irregular pollution plume morphologies.

[0014] This invention achieves a tight coupling between the spatial evolution of pollution plumes and the generation of measurement paths by directly using the implicit representation of the pollution plume boundary and its propagation information as path planning input in the rolling time domain, and by unifying the kinematic constraint parameters and environmental constraint models into the optimization constraints. This enables the measurement paths to continuously point to information-dense areas while meeting safety and feasibility requirements, reducing invalid sampling and improving effective monitoring coverage and path utilization efficiency.

[0015] This invention guides a mobile measurement platform to perform mobile monitoring along the changing direction of the pollution plume boundary and continuously updates the relationship between the plume boundary and spatial location during the monitoring process. This achieves gradual approach and stable positioning of the pollution source area. Without increasing the number of sensors or hardware complexity, it improves the continuity of pollution plume boundary tracking and the accuracy of determining the spatial location of the pollution source, and enhances the practicality and engineering adaptability of the mobile monitoring system in complex environments. Attached Figure Description

[0016] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of the adaptive path planning and pollution plume tracking method for mobile monitoring proposed in this invention; Figure 2 This is a schematic diagram of the measurement path segment generation and update process based on the rolling time-domain optimization algorithm in this invention. Detailed Implementation

[0017] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0018] refer to Figures 1-2 An adaptive path planning and pollution plume tracking method for mobile monitoring includes the following steps: The initial pose, kinematic constraint parameters, and sampling period of the pollutant sensor of the mobile measurement platform are obtained. A spatial coordinate system of the monitoring area is established, and an environmental constraint model including traversable and prohibited boundaries is constructed. Within the passable area defined by the environmental constraint model, pollutant concentration data and corresponding pose data are collected according to the sampling period, and time synchronization and spatial mapping are performed to generate a pollutant concentration observation set. The level set function of the pollution plume is initialized based on the pollutant concentration observation set, an implicit representation is established, and the corresponding evolution parameters are set. The level set evolution method is adopted to update the level set function using the pollutant concentration observation set and evolution parameters, thereby generating an updated implicit representation of the pollution plume boundary and boundary propagation information derived from the level set function; Based on a preset rolling time domain length, a path planning optimization problem is constructed. A rolling time domain optimization algorithm is adopted, taking the implicit representation of the pollution plume boundary and the boundary propagation information as the planning input, and the kinematic constraint parameters and environmental constraint model as the constraint conditions. The optimal control sequence in the rolling time domain is obtained and the measurement path segment is generated. Based on the measurement path segment, the mobile measurement platform is controlled to complete the mobile monitoring task, and outputs the boundary tracking results of the pollution plume and the spatial location determination results of the pollution source.

[0019] In this embodiment, the construction of the environmental constraint model includes: The initial pose of the mobile measurement platform is obtained, which includes the initial position coordinates, the initial heading angle and the initial timestamp, and the initial pose is used as the reference pose for establishing the coordinate system of the monitoring area. A mobile measurement platform refers to a mobile carrier equipped with pollutant sensors and capable of autonomous or semi-autonomous movement. It can acquire its own position information and collect pollutant concentration data simultaneously within the monitoring area according to control commands. Obtain the kinematic constraint parameters of the mobile measurement platform, including the maximum travel speed, maximum angular velocity, maximum acceleration, and minimum turning radius; The sampling period of the pollutant sensor is obtained, and the consistency configuration of the pose sampling period and the pollutant concentration sampling period is determined based on the sampling period to form a time discrete reference. The consistency configuration of pose sampling period and pollutant concentration sampling period refers to using the sampling period of pollutant sensor as a unified time reference, so that the pose data of mobile measurement platform are collected at the same time interval or aligned by interpolation, ensuring that each set of pollutant concentration data corresponds to a unique spatial pose. A monitoring area plane coordinate system is established with the position corresponding to the initial pose as the origin, and the axis of the area plane coordinate system is determined with the initial heading angle, so that the pose data collected subsequently can be directly converted into a sequence of trajectory points under the monitoring area coordinate system. In the plane coordinate system of the monitoring area, a set of restricted areas is generated based on the pre-acquired road boundary, building outline and obstacle distribution data, and the restricted boundary is extracted from the set of restricted areas to constrain the feasible movement space of the mobile measurement platform; Pre-acquired data on road boundaries, building outlines, and obstacle distribution refers to environmental spatial information obtained through electronic maps, geographic information system data, or on-site surveying methods, which is used to determine the spatial extent and boundary location of impassable areas within the monitoring area. In the coordinate system of the monitoring area, based on the set of prohibited areas, kinematic constraint parameters and time discrete reference, the accessibility of the passable space is determined and a set of passable areas is generated. The passable boundary is extracted from the set of passable areas, and an environmental constraint model containing the passable boundary and the prohibited boundary is output. Accessibility determination for passable space refers to determining whether a mobile measurement platform can reach the target space area from its current position without violating speed, steering, or acceleration limits, given a known restricted area and kinematic constraints.

[0020] In this embodiment, the generation of the pollutant concentration observation set includes: Read the passable and prohibited boundaries in the environmental constraint model, determine the passable area based on the passable boundaries, and generate an area identifier for motion constraint determination within the passable area; The mobile measurement platform is controlled to move within the passable area, and the current pose data of the platform is acquired in each control cycle. Based on the area identifier, it is determined whether the spatial position corresponding to the current pose data falls into the restricted area. When the sampling period of the pollutant sensor is obtained, the pollutant concentration is collected, the pollutant concentration data and its sampling timestamp are recorded, and the pose data corresponding to the sampling timestamp are read. The pose data includes position coordinates, heading angle and pose timestamp, forming a pair of concentration data and pose data with timestamp. When the timestamp of the concentration data is inconsistent with the timestamp of the pose data, the pose data is time-aligned using the adjacent timestamps before and after, so that each piece of concentration data corresponds to a time-aligned pose data, and the time-aligned pose data is written back to the concentration data and pose data pair. When the concentration data timestamp is inconsistent with the pose timestamp, it means that the sampling time of the pollutant sensor and the pose measurement unit are different, resulting in the sampling time corresponding to a certain concentration measurement being between two adjacent pose recording times or not having the same timestamp recorded. Time alignment of pose data refers to selecting or interpolating the corresponding pose between adjacent pose timestamps before and after the sampling timestamp of pollutant concentration data, so that the concentration data and pose data correspond under the same time reference. In the plane coordinate system of the monitoring area, the time-aligned pose data is converted into corresponding spatial location identifiers, and the concentration data is bound to the spatial location identifiers to generate concentration observation points; To transform discrete concentration observations into a spatially continuous concentration distribution that can be used for subsequent level set modeling, pollutant concentrations are assigned values ​​on a spatially discrete grid. Let the i-th level set in the spatially discrete grid be... Each grid cell is Its central position is , No. The locations of the pollutant concentration monitoring points are as follows: The corresponding pollutant concentration value Then the grid cell At any moment The concentration assignment result is expressed as follows: ; in, Indicates time Lower grid cell The concentration assignment results Indicates the first A spatial discrete grid cell, Represents grid cells The center position coordinates, Indicates the first Spatial coordinates of each pollutant concentration observation point Indicates the first Concentration measurements at each pollutant concentration observation point This represents the total number of pollutant concentration observation points participating in the current grid concentration assignment calculation. Indicates time Next The observation point for the first The assigned weight of each grid cell; The weight The determination is based on the spatial distance and temporal difference between the observation point and the grid cell, and is expressed as: ; in, Indicates the location of the observation point relative to the center position of the grid The square of the spatial distance between them Indicates the first The sampling timestamps corresponding to each pollutant concentration observation point This indicates the target time for assigning grid concentration values. This represents the time difference between the sampling time at the observation point and the target time. This represents the spatial decay parameter, used to adjust the degree to which spatial distance affects the weights. This represents the time decay parameter, used to adjust the degree of influence of the time difference on the weights. Represents an exponential function; The initial source of this formula is the weighted average formula in mathematics and the weighted fusion method in spatial data interpolation, used to calculate a comprehensive estimate when multiple samples contribute differently to the target quantity. In the fields of environmental monitoring and spatial interpolation, this formula is often extended to calculate the estimated value of the target location based on the relationship between the observation point and the target location. This application, based on this classic form, transforms it from a general sample mean calculation into a pollutant concentration assignment formula for mobile monitoring scenarios, oriented towards spatial discrete grid cells. The estimated object is replaced by the "grid cell" instead of the ordinary scalar mean. At any moment "Concentration estimates", sample values ​​are from Replace with pollutant concentration observations The weights have been expanded from constant weights to dynamic weights that are related to spatial location and temporal state. The derivation process is as follows: First, the pollutant concentration field is regarded as being formed by the joint contribution of discrete observation points to the target grid cell. Then, different weights are set according to the relative influence of each observation point on the grid cell. Finally, the concentration assignment result of the grid cell is obtained by normalized weighted summation, so that the discrete mobile observation data can be transformed into a continuous spatial concentration distribution that can be used for subsequent level set modeling. Based on the environmental constraint model, the concentration observation points are screened for effectiveness. Concentration observation points whose spatial location markers fall into restricted areas are removed, while concentration observation points that fall into accessible areas are retained and aggregated in chronological order to generate a pollutant concentration observation set. Effectiveness screening refers to determining whether the spatial location of each concentration observation point is within a passable area based on the environmental constraint model, and removing observation points located in prohibited areas or lacking corresponding pose information, retaining only the observation data that meet the spatial constraint conditions.

[0021] In this embodiment, the establishment and setting of the implicit representation and the corresponding evolution parameters include: Read the pollutant concentration observation set, extract the spatial location identifier, concentration value and corresponding time information of each concentration observation point, and determine the horizontal set calculation domain based on the spatial range under the plane coordinate system of the monitoring area. The spatial extent in the monitoring area plane coordinate system refers to the overall spatial area used for level set calculation, which is jointly determined by the passable area and the restricted area defined by the environmental constraint model in the established monitoring area plane coordinate system. A spatial discrete grid is established within the computational domain of the level set, and the spatial location identifiers in the pollutant concentration observation set are mapped to the grid cells of the spatial discrete grid to obtain the concentration assignment results corresponding to the grid cells; Establishing a spatial discrete grid refers to dividing a continuous space into several regular or irregular discrete units within a defined computational space according to a preset spatial resolution, for storing and updating level set function values; The initial boundary seed region of the plume is determined based on the concentration assignment result. The initial boundary seed region of the plume is composed of grid cells that meet the preset initialization conditions, and the plume interior region and the plume exterior region are generated from the initial boundary seed region of the plume. The preset initialization conditions refer to the rules used to determine whether a grid cell is the initial region of a plume. These rules are set based on the concentration values ​​or their relative magnitudes in the pollutant concentration observation set. Based on the internal and external regions of the plume, a level set function is initialized on the spatial discrete grid. The level set function takes a first sign in the internal region of the plume and a second sign in the external region of the plume, and the first sign is opposite to the second sign. The boundary of the contaminated plume is represented by the set of zero values ​​of the level set function, forming an implicit representation of the contaminated plume. Initializing the level set function refers to assigning a corresponding initial function value to each grid cell on a spatial discrete grid, based on the determined internal and external regions of the plume, to form an initial implicit function distribution representing the location of the plume boundary. The first symbol being opposite to the second symbol means that during the initialization of the level set function, the function values ​​of the internal region of the plume are uniformly set to positive or negative values, while the function values ​​of the external region of the plume are set to the opposite sign, in order to distinguish the region affiliation; The set of zero values ​​for the level set function represents the boundary of a pollution plume. It refers to the formation of one or more curves at all positions in space where the level set function takes the value of zero, which is used to represent the boundary between the internal and external regions of the pollution plume. Evolutionary parameters for updating the level set function are set, including time step parameters, grid spatial resolution parameters, and driving parameters determined by the pollutant concentration observation set, wherein the driving parameters are generated by the concentration assignment results and stored corresponding to the spatial discrete grid.

[0022] In this embodiment, the generation of the updated implicit representation of the contamination plume boundary and the boundary propagation information derived from the level set function includes: It receives the level set computational domain, spatial discrete grid, initialization level set function, and evolution parameters, and reads the pollutant concentration observation set; The spatial location identifiers of the concentration observation points in the pollutant concentration observation set are mapped to the grid cells of the spatial discrete grid. The pollutant concentration observation values ​​are written into the corresponding grid cells to form concentration assignment results. Based on the concentration assignment results, evolution driving information corresponding to the grid cells is generated. Based on evolution-driven information and evolution parameters, the update rule of the level set function is determined on the spatial discrete grid, and the update rule is applied to the initial level set function to form the update state to be advanced; The update rule for the level set function is implemented by constructing an evolutionary driving function. Let time... The level set function is Then in the time step The updated level set function is expressed as follows: ; in, Indicates time The level set function value, Indicates time The level set function value, Represents the position within a discrete spatial grid. The time step parameter represents the time increment between two level set function updates. Indicates time Lower spatial position Evolution driving function at the location, Level set function In spatial location gradient at, This represents the magnitude of the gradient; The evolution driving function Based on the concentration distribution, boundary curvature, and the consistency between the concentration gradient and the boundary normal, it is constructed as follows: ; in, Indicates spatial location ,time The evolution driving function under, This represents the concentration difference-driven weighting parameter, used to adjust the degree of influence of the concentration difference term on boundary evolution. Indicates time Lower spatial position The estimated concentration at that location, Indicates background concentration. This indicates the deviation of the current concentration from the background concentration. This represents the curvature suppression weight parameter, used to adjust the degree of influence of the boundary curvature term on the evolution process. Indicates time Lower spatial position The curvature at the boundary, This represents the concentration gradient-driven weighting parameter, used to adjust the degree of influence of the concentration gradient direction term on the evolution process. Indicates the spatial location of the concentration field gradient at, Represents the spatial location of the level set function gradient at, Indicates the direction of the boundary normal; The initial source of this formula comes from threshold determination expressions in set theory and background concentration difference discrimination methods in environmental monitoring. Classical threshold determination usually means that elements that meet certain conditions constitute a set; in the field of pollution monitoring, the criterion for identifying abnormal or polluted areas is often "the observed concentration minus the background concentration exceeds a certain threshold". This application expands the determination object from a single sampling point to "spatial discrete grid cells" and uses the threshold determination results for the initialization of subsequent level set functions. The derivation process is as follows: first, the concentration assignment result of each grid cell is obtained using Formula 1; then, the background concentration is introduced to eliminate the influence of the overall environmental baseline, and the concentration deviation of each grid cell relative to the background is obtained; a threshold is set to filter out grid cells whose concentration deviation reaches the preset condition; and the spatial locations corresponding to these grid cells form the initial boundary seed region of the plume, which serves as the basis for subsequent implicit boundary construction. Determining the update rule of the level set function refers to clarifying the numerical update direction and update magnitude of the level set function at each spatial discrete grid cell based on evolution driving information and evolution parameters, in order to constrain the evolution of the level set function over time. Based on the time step parameter in the evolution parameters, a time advancement operation is performed on the updated state to be advanced to obtain the updated level set function, and the updated level set function is written back to the spatial discrete grid as the input for subsequent processing. The time step parameter in the evolution parameters refers to the size of the time increment between two adjacent evolution updates of the level set function, so as to control the discrete time scale of the numerical update of the level set function. Executing time-progression operations refers to updating the level set function values ​​of each cell on a spatial discrete grid step by step according to a predetermined update rule, with the time step parameter as the interval, to obtain the level set function numerical distribution at the next moment; Numerical consistency processing is performed on the updated level set function to obtain the consistent level set function, and the consistent level set function is used as the input for boundary extraction and propagation information derivation; Performing numerical consistency processing refers to adjusting the function values ​​on the spatial discrete grid after the level set function is updated, so that the sign distribution of the inner region of the plume is consistent with that of the outer region of the plume, and to avoid discontinuities between adjacent grid cells. Extract the zero-value set from the level set function after consistency processing, determine the zero-value set as the implicit representation of the updated pollution plume boundary, and map the zero-value set to the plane coordinate system of the monitoring area to generate the spatial location of the plume boundary; Extracting the zero-value set refers to searching for locations on a spatial discrete grid where the level set function takes the value of zero or where the sign changes, and combining these locations to form a set of points or lines used to represent the boundary of the pollution plume; At the spatial location of the plume boundary, the boundary propagation direction information is calculated based on the level set function after consistency processing, and the boundary propagation rate information is determined based on evolution-driven information and time-progression calculation. The updated implicit representation of the pollution plume boundary, as well as the boundary propagation direction information and the boundary propagation rate information, are output. Calculating boundary propagation direction information means determining the normal or principal direction of change of the boundary at each location based on the local variation trend of the horizontal set function on the spatial discrete grid. Determining the boundary propagation rate information means calculating the rate of change of the boundary along its propagation direction at the location of the plume boundary, based on the magnitude of the change of the horizontal set function before and after time progression and the corresponding evolution driving information.

[0023] In this embodiment, the generation of the measurement path segment includes: The system receives the current pose of the mobile measurement platform, kinematic constraint parameters, environmental constraint model, implicit representation of the pollution plume boundary, boundary propagation direction information, and boundary propagation rate information, and uses the current pose as the initial state input for path planning optimization. Read the preset rolling time domain length, generate a discrete time sequence within the rolling time domain based on the rolling time domain length, and use the discrete time sequence to determine the prediction steps and the decision variable dimension of the control sequence within the rolling time domain for the path planning optimization problem; The preset rolling time domain length refers to the future time range set in advance in path planning optimization, which is used to determine the predicted time interval covered by each rolling optimization and the number of discrete moments corresponding to it. Determining the number of prediction steps and the decision variable dimension of the control sequence in the rolling time domain for a path planning optimization problem refers to determining the number of discrete control steps and the corresponding number of control variables in the optimization based on the length of the rolling time domain and the time discrete interval. The rolling time domain is divided into at least two continuous sub-time domains based on the discrete time sequence, and a time index interval corresponding to each sub-time domain is generated. The time index interval is used as the index basis for subsequent segmentation constraint construction and binding. Within the time index interval of each sub-time domain, a corresponding plume boundary prediction representation is generated based on the implicit representation of the pollution plume boundary, as well as the boundary propagation direction information and boundary propagation rate information. The plume boundary prediction representation is then converted into the corresponding sub-time domain constraint structure or sub-time domain target structure, and the piecewise optimization input is output. Based on the piecewise optimization input, candidate measurement corridors are generated in the rolling time domain, and the candidate measurement corridors are decomposed into a set of corridor segments according to the discrete time sequence. The set of corridor segments is used as the spatial constraint input for the construction of the path skeleton. Decomposing a discrete time sequence into a set of corridor segments means dividing the candidate measurement corridor according to the time intervals corresponding to each discrete time in the rolling time domain, so that each time interval corresponds to a spatial corridor region. Within the corridor segment set, candidate path skeletons corresponding to discrete time sequences are generated to obtain skeleton point sequences, and these skeleton point sequences are used as input objects for geometric constraint generation. Based on the skeleton point sequence, trajectory following constraints and deviation constraints are constructed, and the trajectory following constraints and deviation constraints are bound to the corresponding corridor segment set according to the time index interval, forming a piecewise geometric constraint set corresponding to the sub-time domain. The piecewise geometric constraint set is output as the path planning optimization constraint input. Constructing trajectory following constraints and deviation limit constraints refers to determining the desired trajectory position range based on the path skeleton and setting allowable deviation limits for the generated trajectory points to restrict the path from moving near the skeleton; Using the rolling time domain control sequence as the decision variable for the path planning optimization problem, the state sequence and trajectory point sequence are generated from the rolling time domain control sequence based on the mobile measurement platform. The trajectory point sequence is then substituted into the piecewise geometric constraint set, environmental constraint model and kinematic constraint parameters to form the constraint conditions of the path planning optimization problem. The objective function of the path planning optimization problem is constructed by combining the corridor guidance term, control change term and trajectory smoothing term corresponding to the candidate measurement corridor into the objective function, and the objective function and the constraints together constitute the path planning optimization problem based on the rolling time domain length. Constructing the objective function for the path planning optimization problem involves combining the cost terms corresponding to path guidance requirements, control change degree, and trajectory smoothing requirements to form a unified optimization objective for evaluating the quality of the control sequence. The corridor guidance term is a cost term used to constrain the path to be located within the candidate measurement corridor; the control variation term is used to limit the variation between adjacent control variables; and the trajectory smoothing term is used to constrain the continuity and smoothness between trajectory points. An initial control sequence is set for the path planning optimization problem. The optimal control sequence of the previous rolling time domain is time-shifted and aligned to the discrete time sequence to obtain the initial control sequence. The optimal control sequence in the current rolling time domain is obtained by iterative solution. Time-shifting the optimal control sequence of the previous rolling time domain means removing the first segment of control that has been executed and shifting the remaining control quantities forward in chronological order to align them with the discrete time sequence of the current rolling time domain, as a new initial control sequence. Iterative solution refers to repeatedly updating the control sequence and calculating the corresponding objective function value and constraint satisfaction based on a given initial control sequence until the preset iteration termination condition is met. A measurement path segment is generated based on the optimal control sequence, and the feasibility of the measurement path segment is verified based on the environmental constraint model and kinematic constraint parameters. Feasibility verification involves checking each point of the generated measurement path segment to see if it meets the requirements of the passable area defined by the environmental constraint model and the motion conditions defined by the kinematic constraint parameters, and eliminating or correcting path parts that do not meet the constraints.

[0024] In this embodiment, the output of the boundary tracking results of the pollution plume and the spatial location determination results of the pollution source includes: Receive the measurement path segment, parse the measurement path segment into a sequence of path points arranged in chronological order, and assign a corresponding time index to the sequence of path points to form a path execution sequence; A motion control command sequence is generated based on the path execution sequence, and the motion control command sequence is sent to the mobile measurement platform to control the mobile measurement platform to move sequentially along the path execution sequence; Controlling the mobile measurement platform to move sequentially along the path execution sequence means generating corresponding speed and steering control commands based on the path point sequence and sending them to the mobile measurement platform in time index order, so that the platform arrives at each path point in sequence to complete the path execution; During the execution of the path execution sequence on the mobile measurement platform, the execution pose data of the mobile measurement platform is acquired in real time, and the execution pose data is aligned with the time index of the path execution sequence to form a path execution record; Aligning the execution pose data with the time index of the path execution sequence means matching the collected execution pose data to the corresponding time position based on each time index in the path execution sequence, so that each execution pose corresponds to a unique path point index; The pollutant concentration is collected according to the sampling cycle of the pollutant sensor, and the pollutant concentration data is obtained. The pollutant concentration data is then correlated with the time-aligned execution pose data to generate a mobile monitoring and observation set. Triggering pollutant concentration acquisition means that, according to the sampling period or sampling time set by the pollutant sensor, the control system sends a sampling command to the sensor to obtain the pollutant concentration measurement data at the corresponding time. The mobile monitoring observation set is input into the level set evolution process to obtain the updated implicit representation of the pollution plume boundary, and the boundary tracking results of the pollution plume are extracted based on the implicit representation of the pollution plume boundary. The input level set evolution process refers to passing the generated mobile monitoring observation set to the level set evolution module, which serves as the input of observation data for updating the level set function and the implicit representation of the pollution plume boundary; Based on the boundary tracking results of the pollution plume, the spatial location of the pollution source corresponding to the boundary tracking results is determined in the plane coordinate system of the monitoring area, and the boundary tracking results of the pollution plume and the spatial location determination results of the pollution source are output. The spatial location of the pollution source is estimated based on the concentration gradient information at each boundary point in the pollution plume boundary tracing results; let the number of boundary points be... , No. The location of the boundary points is Its corresponding concentration gradient is The spatial location of the pollution source Represented as: ; in, This indicates the results of determining the spatial location of the pollution source. This indicates the number of boundary points involved in the pollution source location estimation. Indicates the first The spatial location of each boundary point in the plane coordinate system of the monitoring area Represents boundary points Concentration gradient at that location, The magnitude of the concentration gradient is used to reflect the intensity of concentration change at the corresponding boundary point; The initial source of this formula belongs to the weighted centroid formula and the gradient weighted estimation method. The weighted centroid formula is used to calculate the comprehensive location estimate based on the importance of different sample points. In source localization and spatial distribution analysis, gradient intensity, energy density or confidence level are often used as weights to perform weighted summation on candidate points. Based on this, this application uses the boundary point positions in the boundary tracking results as candidate positions and the corresponding concentration gradient magnitude as weights to form a pollution source spatial location estimate. The derivation process is as follows: first, obtain the boundary point set based on the level set boundary tracking results; then, consider that the larger the concentration gradient at the boundary point, the more likely the position is to be located in a key position where the concentration change points to the source region. Therefore, the concentration gradient magnitude is selected as the weighting coefficient. Finally, the overall pollution source spatial location estimate result is obtained through weighted centroid calculation. Example 1

[0025] To verify the feasibility of this invention in practice, it was applied to an open monitoring area with complex environmental characteristics as a mobile monitoring application scenario. This area contains multiple roads, buildings of varying heights, and various fixed and temporary obstacles. The sources of pollutants within the area are complex, and the pollution plume exhibits significant irregular diffusion, localized aggregation, and overall drift characteristics under the influence of environmental wind fields and topography. Traditional mobile monitoring methods typically employ pre-set cruise routes or regular grid scanning paths. In actual monitoring, it was found that fixed paths struggle to cover areas with high pollutant concentrations in a timely manner, especially when the pollution plume boundary changes rapidly. The mobile measurement platform often requires a considerable amount of time to re-reach the critical area, resulting in delayed pollution boundary location, a low effective sampling ratio, and unstable results in determining the spatial location of pollution sources.

[0026] In this scenario, a mobile measurement platform is deployed, equipped with pollutant concentration sensors, a pose acquisition module, and a motion control unit. Before entering the monitoring area, the platform first acquires its initial pose information, including initial position coordinates, heading angle, and time reference, while simultaneously loading an environmental constraint model of the monitoring area. This environmental constraint model consists of pre-organized data on road boundaries, building outlines, and obstacle distribution, clearly distinguishing between passable and restricted areas, providing a spatial constraint basis for subsequent path planning. The platform then initiates its mobile monitoring task within the passable area defined by the environmental constraint model.

[0027] During platform movement, pollutant sensors continuously collect pollutant concentration data according to a set sampling period, and the platform pose module synchronously outputs corresponding pose data. Through time alignment processing, each pollutant concentration data is bound to its corresponding spatial pose, forming continuous mobile monitoring observation data. Unlike traditional methods that rely solely on discrete sampling points, this embodiment continuously accumulates and updates the collected observation data to construct a spatial distribution description of the pollution plume.

[0028] Based on the collected pollutant concentration observation set, a spatial discrete grid is established in the plane coordinate system of the monitoring area, and the observation data is mapped to the corresponding grid cells. By processing the distribution relationship of concentration values ​​in the grid cells, the implicit boundary representation of the pollution plume is initialized. In this embodiment, a level set function is used to model the pollution plume boundary, distinguishing the internal and external regions of the plume by function sign; the plume boundary is represented by the set of zero values ​​of the function. As new observation data is continuously input, the level set function is updated according to set evolution parameters, thereby reflecting the spatial diffusion, contraction, and drift process of the pollution plume.

[0029] After each level set evolution update, the system obtains the updated implicit representation of the contamination plume boundary, along with the corresponding boundary propagation direction and rate information. This information is no longer used solely for post-analysis but directly serves as crucial input for path planning. Based on a preset rolling time domain length, the system constructs a path planning optimization problem, dividing the path planning task over a future period into multiple continuous sub-time domains. Within each sub-time domain, a corresponding boundary prediction representation is generated based on the evolution trend of the contamination plume boundary, and this representation is transformed into a sub-time domain constraint structure or a sub-time domain target structure.

[0030] Within the rolling time domain, the system generates candidate measurement corridors based on the aforementioned segmented constraint structure. These measurement corridors dynamically adjust to changes in the contamination plume boundary, no longer being fixed-width or fixed-direction scanning areas, but rather expanding around the contamination plume boundary and its propagation direction. A path skeleton is further constructed within the candidate measurement corridors. This path skeleton consists of a series of spatial points corresponding to time indices, used to constrain the overall shape of subsequent path generation.

[0031] In the path planning optimization process, the control sequence within the rolling time domain is used as the decision variable, and the corresponding state sequence and trajectory point sequence are generated by combining the kinematic model of the mobile measurement platform. The trajectory point sequence is simultaneously constrained by the piecewise geometric constraint set, the environmental constraint model, and the kinematic constraint parameters, ensuring that the generated path can both approach the boundary of the pollution plume and avoid entering restricted areas or exceeding the platform's mobility range. By iteratively solving the path planning optimization problem, the optimal control sequence within the current rolling time domain is obtained, and the measurement path segment is generated from it.

[0032] During actual mobile survey execution, the mobile measurement platform moves according to the generated measurement path segments. The platform control system parses the path segments into continuous motion control commands, enabling the platform to sequentially reach the path points. While executing the path, the platform continuously collects pollutant concentration and pose data. New observation data is input into the level set evolution process in real time to update the pollution plume boundary representation, thus forming a closed-loop process of path planning and plume tracking.

[0033] Comparison of actual operational data from this embodiment reveals that, compared to traditional mobile monitoring methods using fixed cruise paths, the method of this invention significantly increases the proportion of time the mobile measurement platform spends in areas with high pollutant concentrations. Under the same monitoring duration, the number of effective sampling points increases by approximately 45%, with the proportion of sampling points located near the pollution plume boundary increasing from less than 30% in the original method to nearly 60%. Regarding the accuracy of pollution plume boundary tracking, comparison of multiple evolutionary update results shows a significant reduction in spatial deviation at the boundary position, and a significant enhancement in boundary continuity and stability.

[0034] Regarding the spatial location determination of pollution sources, the mobile measurement platform was guided to gradually approach the pollution source area based on the contraction and propagation characteristics of the plume boundary over time. Through the cumulative analysis of continuous mobile monitoring data, the spatial range of the candidate pollution source area was continuously reduced, and the final determined pollution source location had a deviation of approximately 40% compared to the location results of traditional methods. In multiple repeated experiments, the results of the pollution source spatial location determination showed good stability, and the range of location fluctuations was significantly reduced.

[0035] In summary, this embodiment fully demonstrates the practical application effect of the technical solution proposed by this invention to address the problems of fixed path planning and delayed response to changes in pollution plumes in existing mobile monitoring systems. By deeply integrating the level set evolution method with the rolling temporal optimization algorithm into the mobile monitoring path planning and navigation control process, continuous tracking of pollution plume boundaries and adaptive adjustment of measurement paths are achieved. This effectively improves the efficiency of mobile monitoring, the accuracy of pollution plume boundary tracking, and the reliability of determining the spatial location of pollution sources in complex environments, verifying the feasibility and practical value of the method in practical engineering applications.

[0036] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An adaptive path planning and pollution plume tracking method for mobile monitoring, characterized in that, Includes the following steps: The initial pose, kinematic constraint parameters, and sampling period of the pollutant sensor of the mobile measurement platform are obtained. A spatial coordinate system of the monitoring area is established, and an environmental constraint model including traversable and prohibited boundaries is constructed. Within the passable area defined by the environmental constraint model, pollutant concentration data and corresponding pose data are collected according to the sampling period, and time synchronization and spatial mapping are performed to generate a pollutant concentration observation set. The level set function of the pollution plume is initialized based on the pollutant concentration observation set, an implicit representation is established, and the corresponding evolution parameters are set. The level set evolution method is adopted to update the level set function using the pollutant concentration observation set and evolution parameters, thereby generating an updated implicit representation of the pollution plume boundary and boundary propagation information derived from the level set function; Based on a preset rolling time domain length, a path planning optimization problem is constructed. A rolling time domain optimization algorithm is adopted, taking the implicit representation of the pollution plume boundary and the boundary propagation information as the planning input, and the kinematic constraint parameters and environmental constraint model as the constraint conditions. The optimal control sequence in the rolling time domain is obtained and the measurement path segment is generated. Based on the measurement path segment, the mobile measurement platform is controlled to complete the mobile monitoring task and output the boundary tracking results of the pollution plume and the spatial location determination results of the pollution source; The construction of the environmental constraint model includes: The initial pose of the mobile measurement platform is obtained, which includes the initial position coordinates, the initial heading angle and the initial timestamp, and the initial pose is used as the reference pose for establishing the coordinate system of the monitoring area. Obtain the kinematic constraint parameters of the mobile measurement platform, including the maximum travel speed, maximum angular velocity, maximum acceleration, and minimum turning radius; The sampling period of the pollutant sensor is obtained, and the consistency configuration of the pose sampling period and the pollutant concentration sampling period is determined based on the sampling period to form a time discrete reference. A monitoring area plane coordinate system is established with the position corresponding to the initial pose as the origin, and the axis of the area plane coordinate system is determined with the initial heading angle. In the plane coordinate system of the monitoring area, a set of restricted areas is generated based on the pre-acquired road boundary, building outline and obstacle distribution data, and the restricted boundary is extracted from the set of restricted areas to constrain the feasible movement space of the mobile measurement platform; In the coordinate system of the monitoring area, based on the set of prohibited areas, kinematic constraint parameters and time discrete reference, the accessibility of the passable space is determined and a set of passable areas is generated. The passable boundary is extracted from the set of passable areas, and an environmental constraint model containing the passable boundary and the prohibited boundary is output. The generation of the pollutant concentration observation set includes: Read the passable and prohibited boundaries in the environmental constraint model, determine the passable area based on the passable boundaries, and generate an area identifier for motion constraint determination within the passable area; The mobile measurement platform is controlled to move within the passable area, and the current pose data of the platform is acquired in each control cycle. Based on the area identifier, it is determined whether the spatial position corresponding to the pose falls into the restricted area. When the sampling period of the pollutant sensor is obtained, the pollutant concentration is triggered to collect data, and the pollutant concentration data and its sampling timestamp are recorded. At the same time, the pose data corresponding to the sampling timestamp is read to form a pair of concentration data and pose data with timestamps. When the timestamp of the concentration data is inconsistent with the timestamp of the pose data, the pose data is time-aligned using the adjacent timestamps, and the time-aligned pose data is written back to the concentration data and pose data pair. In the plane coordinate system of the monitoring area, the time-aligned pose data is converted into corresponding spatial location identifiers, and the concentration data is bound to the spatial location identifiers to generate concentration observation points; Based on the environmental constraint model, the concentration observation points are screened for effectiveness. Concentration observation points whose spatial location markers fall into restricted areas are removed, while concentration observation points that fall into accessible areas are retained and aggregated in chronological order to generate a pollutant concentration observation set. The establishment and setting of the implicit representation and the corresponding evolution parameters include: Read the pollutant concentration observation set, extract the spatial location identifier, concentration value and corresponding time information of each concentration observation point, and determine the horizontal set calculation domain based on the spatial range under the plane coordinate system of the monitoring area. A spatial discrete grid is established within the computational domain of the level set, and the spatial location identifiers in the pollutant concentration observation set are mapped to the grid cells of the spatial discrete grid to obtain the concentration assignment results corresponding to the grid cells; The initial boundary seed region of the plume is determined based on the concentration assignment result. The initial boundary seed region of the plume is composed of grid cells that meet the preset initialization conditions, and the plume interior region and the plume exterior region are generated from the initial boundary seed region of the plume. Based on the internal and external regions of the plume, a level set function is initialized on the spatial discrete grid. The level set function takes a first sign in the internal region of the plume and a second sign in the external region of the plume, and the first sign is opposite to the second sign. The boundary of the contaminated plume is represented by the set of zero values ​​of the level set function, forming an implicit representation of the contaminated plume. The evolution parameters for updating the level set function are set, including time step parameters, grid spatial resolution parameters, and driving parameters determined by the pollutant concentration observation set; The updated implicit representation of the contamination plume boundary and the generation of boundary propagation information derived from the level set function include: It receives the level set computational domain, spatial discrete grid, initialization level set function, and evolution parameters, and reads the pollutant concentration observation set; The spatial location identifiers of the concentration observation points in the pollutant concentration observation set are mapped to the grid cells of the spatial discrete grid. The pollutant concentration observation values ​​are written into the corresponding grid cells to form concentration assignment results. Based on the concentration assignment results, evolution driving information corresponding to the grid cells is generated. Based on evolution-driven information and evolution parameters, the update rule of the level set function is determined on the spatial discrete grid, and the update rule is applied to the initial level set function to form the update state to be advanced; Based on the time step parameter in the evolution parameters, perform time advancement operation on the updated state to be advanced to obtain the updated level set function; Perform numerical consistency processing on the updated level set function to obtain the consistent level set function. Extract the zero-value set from the level set function after consistency processing, determine the zero-value set as the implicit representation of the updated pollution plume boundary, and map the zero-value set to the plane coordinate system of the monitoring area to generate the spatial location of the plume boundary; At the spatial location of the plume boundary, the boundary propagation direction information is calculated based on the level set function after consistency processing, and the boundary propagation rate information is determined based on evolution-driven information and time-progression calculation. The updated implicit representation of the pollution plume boundary, as well as the boundary propagation direction information and the boundary propagation rate information, are output. The generation of the measurement path segment includes: The system receives the current pose of the mobile measurement platform, kinematic constraint parameters, environmental constraint model, implicit representation of the pollution plume boundary, boundary propagation direction information, and boundary propagation rate information, and uses the current pose as the initial state input for path planning optimization. Read the preset rolling time domain length, generate a discrete time sequence within the rolling time domain based on the rolling time domain length, and use the discrete time sequence to determine the prediction steps and the decision variable dimension of the control sequence within the rolling time domain for the path planning optimization problem; The rolling time domain is divided into at least two continuous sub-time domains based on discrete time-time sequences, and time index intervals corresponding to each sub-time domain are generated. Within the time index interval of each sub-time domain, a corresponding plume boundary prediction representation is generated based on the implicit representation of the pollution plume boundary, as well as the boundary propagation direction information and boundary propagation rate information. The plume boundary prediction representation is then converted into the corresponding sub-time domain constraint structure or sub-time domain target structure, and the piecewise optimization input is output. Based on the piecewise optimization input, candidate measurement corridors are generated in the rolling time domain, and the candidate measurement corridors are decomposed into a set of corridor segments according to the discrete time sequence. The set of corridor segments is used as the spatial constraint input for the construction of the path skeleton. Within the corridor segment set, candidate path skeletons corresponding to discrete time sequences are generated to obtain skeleton point sequences. Based on the skeleton point sequence, trajectory following constraints and deviation constraints are constructed, and the trajectory following constraints and deviation constraints are bound to the corresponding corridor segment set according to the time index interval, forming a piecewise geometric constraint set corresponding to the sub-time domain. The piecewise geometric constraint set is output as the path planning optimization constraint input. Using the rolling time domain control sequence as the decision variable for the path planning optimization problem, the state sequence and trajectory point sequence are generated from the rolling time domain control sequence based on the mobile measurement platform. The trajectory point sequence is then substituted into the piecewise geometric constraint set, environmental constraint model and kinematic constraint parameters to form the constraint conditions of the path planning optimization problem. The objective function of the path planning optimization problem is constructed by combining the corridor guidance term, control change term and trajectory smoothing term corresponding to the candidate measurement corridor into the objective function, and the objective function and the constraints together constitute the path planning optimization problem based on the rolling time domain length. An initial control sequence is set for the path planning optimization problem. The optimal control sequence of the previous rolling time domain is time-shifted and aligned to the discrete time sequence to obtain the initial control sequence. The optimal control sequence in the current rolling time domain is obtained by iterative solution. A measurement path segment is generated based on the optimal control sequence, and the feasibility of the measurement path segment is verified based on the environmental constraint model and kinematic constraint parameters. The output of the boundary tracking results of the pollution plume and the spatial location determination results of the pollution source include: Receive the measurement path segment, parse the measurement path segment into a sequence of path points arranged in chronological order, and assign a corresponding time index to the sequence of path points to form a path execution sequence; A motion control command sequence is generated based on the path execution sequence, and the motion control command sequence is sent to the mobile measurement platform to control the mobile measurement platform to move sequentially along the path execution sequence; During the execution of the path execution sequence on the mobile measurement platform, the execution pose data of the mobile measurement platform is acquired in real time, and the execution pose data is aligned with the time index of the path execution sequence to form a path execution record; The pollutant concentration is collected according to the sampling cycle of the pollutant sensor, and the pollutant concentration data is obtained. The pollutant concentration data is then correlated with the time-aligned execution pose data to generate a mobile monitoring and observation set. The mobile monitoring observation set is input into the level set evolution process to obtain the updated implicit representation of the pollution plume boundary, and the boundary tracking results of the pollution plume are extracted based on the implicit representation of the pollution plume boundary. Based on the boundary tracking results of the pollution plume, the spatial location of the pollution source corresponding to the boundary tracking results is determined in the plane coordinate system of the monitoring area, and the boundary tracking results of the pollution plume and the spatial location determination results of the pollution source are output.

Images