A method for intelligent obstacle avoidance control of tiling robots in the construction process

Abstract

This invention relates to the field of intelligent obstacle avoidance control technology, and more particularly to an intelligent obstacle avoidance control method for a paving robot in the construction process, comprising the following steps: S1, acquiring chassis kinematic parameters including rated linear velocity and rated angular acceleration, as well as ambient temperature and relative humidity parameters; S2, inputting the ambient temperature and relative humidity parameters into the physical state calculation model of the auxiliary material, and outputting the dynamic viscosity and dynamic remaining effective time of the auxiliary material. In this invention, a spatial cost map containing obstacles and their extended areas is constructed as the obstacle avoidance environment benchmark, and a dynamic velocity sampling space is constructed using the adjusted boundary, enabling the chassis to adapt to changes in road surface slip characteristics when avoiding obstacles according to the spatial cost map, effectively reducing the risk of sideslip of the chassis on the auxiliary material spreading surface, and improving the obstacle avoidance control safety and driving stability of the robot under variable friction conditions.

Description

Technical Field

[0001] This invention relates to the field of intelligent obstacle avoidance control technology, and in particular to an intelligent obstacle avoidance control method for a paving robot in the construction process. Background Technology

[0002] With the development of building construction automation technology, paving robots have been widely used in construction operations. When performing paving tasks, these robots usually need to use various environmental sensor data to construct a spatial cost map containing the distribution of obstacles such as surrounding personnel and materials, and use this as an environmental benchmark to perform real-time intelligent obstacle avoidance path planning and chassis motion control.

[0003] However, existing paving robots often use fixed centripetal acceleration and angular velocity boundaries to generate bottom layer speed commands when performing obstacle avoidance control. They do not take into account the actual impact of dynamic viscosity changes of paving auxiliary materials such as cement slurry or adhesives on the slip characteristics of the chassis and tires at the construction site. As a result, when the robot performs obstacle avoidance and steering actions on the variable friction construction surface, it is very easy for the generated control speed to exceed the physical adhesion limit of the actual road surface, which may lead to the risk of sideslip and loss of control. Summary of the Invention

[0004] To overcome the above shortcomings, this invention provides an intelligent obstacle avoidance control method for a paving robot in the construction process, aiming to improve the problem that existing paving robots often use fixed centripetal acceleration and angular velocity boundaries to generate bottom-layer velocity commands when performing obstacle avoidance control.

[0005] This invention provides the following technical solution: an intelligent obstacle avoidance control method for a tiling robot during the construction process, comprising the following steps:

[0006] S1. Obtain the chassis kinematic parameters, including the rated linear velocity and rated angular acceleration, as well as the ambient temperature and relative humidity parameters;

[0007] S2. Input the ambient temperature and relative humidity parameters into the auxiliary material physical state calculation model, and output the dynamic viscosity and dynamic remaining effective time of the auxiliary material.

[0008] S3. Adjust the surface roughness penalty weight according to the dynamic viscosity of the auxiliary material, adjust the obstacle expansion radius parameter according to the dynamic remaining effective time, and generate a spatial cost map based on the adjusted surface roughness penalty weight and the obstacle expansion radius parameter.

[0009] S4. Based on the spatial cost map, execute the path planning algorithm to generate a global avoidance path, and combine the chassis kinematic parameters to calculate the predicted total time along the global avoidance path;

[0010] S5. If the predicted total time is less than the dynamic remaining effective time, the global avoidance path is confirmed as an executable route; otherwise, a termination instruction is generated.

[0011] S6. Generate a local trajectory based on the executable route, and adjust the centripetal acceleration boundary and angular velocity boundary according to the dynamic viscosity of the auxiliary material to generate control commands that match the local trajectory.

[0012] Preferably, in step S2, the step of outputting the dynamic viscosity and dynamic remaining effective time of the auxiliary material specifically includes:

[0013] The ambient temperature and relative humidity parameters are input into the physical state calculation model of the auxiliary materials, and the initial ratio viscosity, preparation time and failure viscosity threshold of the auxiliary materials are read.

[0014] Based on the ambient temperature and relative humidity parameters, the initial viscosity ratio and the time difference between the current time and the preparation time, the dynamic viscosity and dynamic curing rate of the auxiliary material are calculated.

[0015] Based on the dynamic viscosity of the auxiliary material, the dynamic curing rate, and the failure viscosity threshold, the remaining time required for the dynamic viscosity of the auxiliary material to reach the failure viscosity threshold is calculated, and the dynamic remaining effective time is output.

[0016] The physical state calculation model for the auxiliary materials is a model trained based on historical ambient temperature, historical relative humidity, historical initial viscosity, duration of time, and historical measured viscosity.

[0017] Preferably, in step S3, the steps of adjusting the surface roughness penalty weight and adjusting the obstacle expansion radius parameter specifically include:

[0018] Obtain the basic penalty weight, calculate the weight gain based on the dynamic viscosity of the auxiliary material, the weight gain is negatively correlated with the value of the dynamic viscosity of the auxiliary material, and calculate the basic penalty weight and the weight gain by superimposing them to output the surface roughness penalty weight.

[0019] Obtain the physical contour radius of the chassis, calculate the expansion gain based on the dynamic remaining effective time, the expansion gain is positively correlated with the value of the dynamic remaining effective time, and calculate the obstacle expansion radius parameter by superimposing the physical contour radius of the chassis and the expansion gain.

[0020] Preferably, in step S3, the step of generating the spatial cost map specifically includes:

[0021] Obtain the static obstacle map and surface roughness layer of the current work area;

[0022] An obstacle expansion cost layer is generated based on the static obstacle map and the obstacle expansion radius parameter;

[0023] The surface roughness layer and the surface roughness penalty weight are weighted and calculated to generate a surface roughness penalty cost layer.

[0024] The static obstacle map, the obstacle expansion cost layer, and the working surface roughness penalty cost layer are fused together to output a spatial cost map.

[0025] Preferably, the step of performing a fusion calculation on the static obstacle map, the obstacle expansion cost layer, and the working surface roughness penalty cost layer to output a spatial cost map specifically includes:

[0026] Extract the coordinates of the restricted boundary from the static obstacle map, and assign the grid value corresponding to the restricted boundary coordinates as a restricted constant;

[0027] For the coordinates of the passable area other than the restricted boundary coordinates, extract the expansion cost value in the obstacle expansion cost layer and the roughness cost value in the working surface roughness penalty cost layer.

[0028] Calculate the roughness gradient value of the passable area coordinates in a preset spatial neighborhood, and multiply the roughness gradient value with the expansion cost value to generate a coupling penalty cost value that characterizes the risk of slippage.

[0029] The expansion cost, the roughness cost, and the coupling penalty cost are input into the exponential fusion function for smoothing calculation to obtain the feasible fusion cost.

[0030] Based on the distribution sequence of the original spatial grid coordinates, the prohibition constant and the passable fusion cost matrix are reconstructed to output a spatial cost map.

[0031] Preferably, in step S4, the step of generating a global avoidance path based on the spatial cost map using a path planning algorithm, and calculating the predicted total time along the global avoidance path in conjunction with the chassis kinematic parameters specifically includes:

[0032] Obtain the current starting coordinates and target coordinates of the chassis, execute the path planning algorithm in the spatial cost map, and generate a global avoidance path containing discrete path nodes;

[0033] Analyze the chassis kinematic parameters and extract the rated linear velocity and rated angular acceleration;

[0034] The basic straight-line driving time is calculated by combining the spatial distance between the discrete path nodes and the rated linear speed. The corner nodes in the discrete path nodes are extracted, and the turning acceleration and deceleration time is calculated based on the angle difference of the corner nodes and the rated angular acceleration.

[0035] Add the basic straight-line driving time to the turning acceleration and deceleration time to output the predicted total time.

[0036] Preferably, in S5, the step of confirming the global evasion path as an executable route specifically includes:

[0037] Extract the original coordinate sequence corresponding to the global avoidance path;

[0038] The original coordinate sequence is smoothed by calling a curve fitting algorithm to generate a smooth space curve.

[0039] Extract the discrete coordinate data of the smoothed space curve, encapsulate the discrete coordinate data into a control task data packet containing the discrete coordinate data, velocity constraint parameters and task identifier, push the control task data packet into the control buffer queue, and mark the global avoidance path status as an executable route.

[0040] Preferably, in step S6, the step of generating control instructions that match the local trajectory specifically includes:

[0041] Extract the look-ahead path curvature within a preset look-ahead distance range in the local trajectory, and construct a dynamic velocity sampling space with upper and lower limits of linear velocity and angular velocity by taking the current linear velocity and current angular velocity of the chassis as a reference, and combining the acceleration and deceleration constraint parameters of the chassis and the adjusted centripetal acceleration boundary and angular velocity boundary.

[0042] Multiple sets of sampled velocity pairs are generated within the dynamic velocity sampling space. An evaluation function is constructed that includes trajectory matching cost, obstacle distance cost, and velocity smoothing cost. The evaluation function is used to score the multiple sets of sampled velocity pairs. The sampled velocity pairs whose evaluation values ​​satisfy the preset extreme value conditions are selected as the target linear velocity and target angular velocity for matching the local trajectory.

[0043] The target linear velocity and the target angular velocity are input into the chassis inverse kinematics model to calculate the target speed sequence of the chassis drive motor and generate control commands.

[0044] The present invention has the following beneficial effects:

[0045] 1. This invention constructs a spatial cost map containing obstacles and their extended areas as an obstacle avoidance environment benchmark, and dynamically adjusts the centripetal acceleration and angular velocity boundaries of the chassis by combining the real-time acquired dynamic viscosity of the road surface material. Then, it uses the adjusted boundaries to construct a dynamic velocity sampling space, enabling the chassis to adapt to changes in road surface slip characteristics when avoiding obstacles according to the spatial cost map. This effectively reduces the risk of sideslip of the chassis on the road surface where the material is spread, and improves the obstacle avoidance control safety and driving stability of the robot under variable friction conditions.

[0046] 2. This invention introduces a predicted time node for four-dimensional state space search during the global path planning stage to avoid dynamic obstacles, and calculates obstacle distance cost and performs collision elimination judgment in real time during local trajectory evaluation. This achieves coordination between global time constraints and local environmental changes, improving the obstacle avoidance control accuracy of the chassis in complex dynamic environments while ensuring that the task meets timeliness requirements.

[0047] 3. This invention uses a cubic spline curve fitting algorithm to smooth the discrete node sequence of the global avoidance path, and calculates the allowable linear velocity constraints point by point based on the absolute value of the curve curvature at each sampling point and the maximum allowable lateral acceleration of the system. This improves the speed change problem caused by the path polyline, and reduces the wear of the underlying mechanical transmission mechanism while ensuring that the chassis travels smoothly along the desired trajectory. Attached Figure Description

[0048] Figure 1 The flowchart shows an intelligent obstacle avoidance control method for a paving robot in the construction process proposed in this invention. Detailed Implementation

[0049] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0050] In a first embodiment of the present invention, the present invention provides an intelligent obstacle avoidance control method for a tiling robot in the construction process, such as... Figure 1 As shown, it includes the following steps:

[0051] S1. Obtain the chassis kinematic parameters, including the rated linear velocity and rated angular acceleration, as well as the ambient temperature and relative humidity parameters;

[0052] Specifically, the main control unit of the tiling robot reads the configuration firmware in the underlying driver via the field controller local area network bus. It reads the speed register data and steering control register data from the driver, converts the maximum permissible pulse frequency per unit time of the drive wheels into the maximum straight-line speed allowed by the chassis by combining the acquired maximum permissible pulse frequency per unit time with the physical wheel diameter, and confirms this as the rated linear speed. It extracts the maximum angular velocity limit parameter from the steering control register, and calculates the maximum permissible angular velocity change per unit time by combining it with the preset control cycle duration at the underlying level, confirming this as the rated angular acceleration. The acquired rated linear speed and rated angular acceleration are written to the system's global shared memory for direct use by the subsequent path planning and total time prediction modules.

[0053] A temperature and humidity probe mounted on the top of the device, away from the motor's heat source, continuously collects raw temperature and relative humidity sequences at a preset sampling period. Due to airflow disturbances and local temperature differences at the construction site, the raw acquired signals are prone to high-frequency fluctuations. A time-domain weighted smoothing filter algorithm is applied to the extracted raw time series to obtain stable physical environmental variables after eliminating transient fluctuations.

[0054] Let the current sampling time be t. The system extracts the original temperature and relative humidity sets from the N most recent consecutive historical sampling periods. The effective ambient temperature and effective relative humidity at the current time are calculated using a time-domain weighted smoothing filter model. The temperature filtering calculation formula is as follows: The formula for humidity filtering is as follows: In the above formula, This indicates the effective ambient temperature of the calculated output. This represents the calculated effective relative humidity. This represents the total number of historical continuous sampling points involved in the calculation. greater than zero and less than or equal to natural numbers, Indicates the first Raw temperature data collected at historical moments. Indicates the first Raw relative humidity data collected at historical moments. Indicates the first The time decay weighting coefficients corresponding to each historical sampled data point. The system-defined weighting coefficient constraints are as follows: Furthermore, the closer the sampling point is to the current time t in the time dimension, the larger the weight coefficient is assigned.

[0055] The effective ambient temperature and relative humidity output after filtering are pushed to the data stream bus as the final acquired ambient temperature and relative humidity parameters, serving as direct input data to trigger the operation of the auxiliary material physical state calculation model. This step quantifies the chassis's motion limit state and eliminates random errors in the environmental sensing data, providing accurate basic parameter support for subsequent auxiliary material viscosity derivation and global time consumption calculation.

[0056] S2. Input the ambient temperature and relative humidity parameters into the physical state calculation model of the auxiliary material, and output the dynamic viscosity and dynamic remaining effective time of the auxiliary material.

[0057] Furthermore, in S2, the steps for outputting the dynamic viscosity and dynamic remaining effective time of the auxiliary material specifically include:

[0058] Input the ambient temperature and relative humidity parameters into the physical state calculation model of the auxiliary materials, and read the initial ratio viscosity, preparation time and failure viscosity threshold of the auxiliary materials.

[0059] Based on ambient temperature and relative humidity parameters, initial viscosity ratio and the time difference between the current time and the preparation time, calculate the dynamic viscosity and dynamic curing rate of the auxiliary materials;

[0060] Based on the dynamic viscosity of the auxiliary material, the dynamic curing rate, and the failure viscosity threshold, the remaining time required for the dynamic viscosity of the auxiliary material to reach the failure viscosity threshold is calculated, and the dynamic remaining effective time is output.

[0061] The physical state calculation model for auxiliary materials is a model trained based on historical ambient temperature, historical relative humidity, historical initial viscosity, duration of time, and historical measured viscosity.

[0062] Specifically, after the system acquires and outputs the effective ambient temperature and effective relative humidity, the main control unit of the laying robot reads these two parameters in real time from the data stream bus and assigns them as the current ambient temperature input value and the current relative humidity input value, respectively. Simultaneously, it directly reads the process parameters of the currently carried auxiliary material from the system's preset engineering task configuration data package, extracting the initial viscosity ratio of the auxiliary material, the time of completion of auxiliary material preparation, and the failure viscosity threshold. In this scheme, the initial viscosity ratio of the auxiliary material is preferably 200 Pascal-seconds, and the failure viscosity threshold is preferably 600 Pascal-seconds. The current system time is acquired, and the time difference between the current system time and the time of completion of auxiliary material preparation is calculated. This time difference is confirmed as the elapsed time from the current time to the preparation time.

[0063] The current ambient temperature, current relative humidity, initial viscosity, and duration are input into the auxiliary material physical state calculation model. This model includes built-in formulas for calculating the dynamic viscosity and dynamic curing rate of the auxiliary material. The dynamic viscosity calculation formula is expressed as follows: The dynamic curing rate is obtained by differentiating the dynamic viscosity with respect to time, and the specific formula is expressed as follows: In the two formulas above, This represents the calculated dynamic viscosity of the auxiliary material. This represents the calculated dynamic curing rate. This represents the current ambient temperature input value. This indicates the current relative humidity input value. Indicates the initial viscosity of the auxiliary materials. Indicates the duration of time. to This represents the model parameter coefficients to be obtained through training with historical data. It should be noted that, considering that paving robots typically operate under normal construction temperatures and the effective usage cycle after a single mixing of auxiliary materials is relatively short, at this specific spatiotemporal scale, the polymer cross-linking reaction within the auxiliary materials has not yet entered a drastic nonlinear abrupt change stage. Therefore, using the aforementioned quadratic polynomial as an approximate empirical fitting model significantly reduces the computational power consumption of the underlying control system while its fitting accuracy is sufficient to meet the subsequent adjustment requirements of the chassis dynamic boundaries. Furthermore, in determining the dynamic curing rate at the current moment... At the same time, it is assumed that the ambient temperature and relative humidity remain constant at the current collected values ​​within a short remaining prediction window to ensure the reliability of the curing rate prediction.

[0064] To obtain accurate model parameter coefficients, the physical state calculation model for auxiliary materials was calibrated in an offline laboratory environment before deployment. Specifically, a constant temperature and humidity test chamber and a rotational rheometer were used in the laboratory to rigorously simulate temperature ranges and relative humidity conditions covering different construction seasons. Curves of the actual curing viscosity of multiple batches of auxiliary materials under corresponding conditions were collected as historical measured data to construct a historical dataset containing multiple samples. This mechanism effectively avoids the engineering limitations of construction sites lacking the conditions for real-time online and non-destructive measurement of chemical viscosity. The system constructs a least-squares optimization objective function, expressed by the following formula: In this formula, Let represent the objective function of the sum of squared errors between the predicted viscosity and the measured viscosity. This represents the total number of samples in the historical dataset used in the calculation. It means greater than zero and less than or equal to zero. natural numbers, Indicates the first The historical measured viscosity corresponding to each sample Indicates the first The historical ambient temperature corresponding to each sample Indicates the first The historical relative humidity corresponding to each sample Indicates the first The historical initial viscosity ratio corresponding to each sample Indicates the first The system calculates the historical duration corresponding to each sample by solving this objective function. Find the minimum value to obtain the optimal model parameter coefficients. to And it is solidified into the physical state calculation model of auxiliary materials.

[0065] After obtaining the current dynamic viscosity and dynamic curing rate of the auxiliary material, the time margin for the current auxiliary material transportation task is calculated by combining the failure viscosity threshold. Under the physical constraints of the auxiliary material viscosity monotonically increasing with time and the dynamic curing rate being greater than zero, the basic prediction formula is used. A preliminary solution is performed. In this formula, Indicates the remaining effective time. This indicates the failure viscosity threshold defined by the construction process. This indicates the dynamic viscosity of the auxiliary material at the current moment. This represents the dynamic curing rate at the current moment. Due to the nonlinear characteristics of the actual curing process of the auxiliary materials, a forward iterative simulation is performed based on the remaining effective time of the base material, accumulating the viscosity nonlinear increment cycle by cycle. To ensure the precise executability of this forward iterative process in the underlying control algorithm, an explicit difference equation is used for recursive calculation. The specific discrete recursive formula is as follows:

[0066] ;

[0067] ;

[0068] In this formula, Indicates the loop step index of the forward iteration. and These represent the cumulative duration of the current iteration step and the next iteration step, respectively. This indicates the preset time step, which is preferably 60 seconds. and These represent the predicted viscosity of the current iteration step and the next iteration step, respectively. and These represent the current ambient temperature input value and the current relative humidity input value, respectively. to These are the coefficients of the acquired model parameters. They will be used during the initialization iteration. The value is assigned to the current duration obtained from the previous step. ,Will The value is assigned to the current dynamic viscosity of the auxiliary material. When the viscosity value predicted by the iterative simulation Reaching the failure viscosity threshold When the iteration stops, the total accumulated time is recorded. Duration since the current system time The difference is confirmed as the final output dynamic remaining valid time.

[0069] After the calculation is completed, the dynamic viscosity of the auxiliary material and the dynamic remaining effective time are pushed to the global data flow bus.

[0070] This step enables a precise mathematical expression of the internal physicochemical changes of building materials, successfully transforming the complex solidification state of materials into time constraints and physical penalty parameters that can be directly called upon by subsequent obstacle avoidance strategies and path planning modules.

[0071] S3. Adjust the surface roughness penalty weight according to the dynamic viscosity of the auxiliary material, adjust the obstacle expansion radius parameter according to the dynamic remaining effective time, and generate a spatial cost map based on the adjusted surface roughness penalty weight and obstacle expansion radius parameter.

[0072] Furthermore, in S3, the steps of adjusting the surface roughness penalty weight and adjusting the obstacle expansion radius parameter specifically include:

[0073] Obtain the basic penalty weight, calculate the weight gain based on the dynamic viscosity of the auxiliary material. The weight gain is negatively correlated with the value of the dynamic viscosity of the auxiliary material. The basic penalty weight and the weight gain are superimposed to calculate and output the surface roughness penalty weight.

[0074] Obtain the physical contour radius of the chassis, calculate the expansion gain based on the dynamic remaining effective time. The expansion gain is positively correlated with the value of the dynamic remaining effective time. The physical contour radius of the chassis and the expansion gain are superimposed to calculate the obstacle expansion radius parameter.

[0075] Furthermore, in S3, the steps for generating the spatial cost map specifically include:

[0076] Obtain the static obstacle map and surface roughness layer of the current work area;

[0077] Generate an obstacle expansion cost layer based on a static obstacle map and obstacle expansion radius parameters;

[0078] The surface roughness layer and the surface roughness penalty weight are weighted and calculated to generate the surface roughness penalty cost layer.

[0079] Perform a fusion calculation on the static obstacle map, the obstacle expansion cost layer, and the surface roughness penalty cost layer to output a spatial cost map.

[0080] Furthermore, the static obstacle map, obstacle expansion cost layer, and surface roughness penalty cost layer are fused and calculated to output a spatial cost map, specifically including:

[0081] Extract the coordinates of the restricted boundaries from the static obstacle map, and assign the raster values ​​of the corresponding restricted boundary coordinates as restricted constants;

[0082] For the coordinates of passable areas other than the coordinates of the restricted boundary, extract their inflation cost in the obstacle inflation cost layer and their roughness cost in the surface roughness penalty cost layer.

[0083] Calculate the roughness gradient value of the passable area coordinates in the preset spatial neighborhood, multiply the roughness gradient value with the expansion cost value to generate the coupling penalty cost value that characterizes the risk of slippage and deviation.

[0084] The expansion cost, roughness cost, and coupling penalty cost are input into the exponential fusion function for smoothing calculation to obtain the feasible fusion cost.

[0085] Based on the distribution sequence of the original spatial raster coordinates, the forbidden constant and the passable fusion cost are reconstructed by performing a matrix, and the spatial cost map is output.

[0086] Specifically, the dynamic viscosity and remaining effective time of the auxiliary material output from the previous steps are read from the global data flow bus. The system's preset basic penalty weight is obtained, and the weight gain is calculated based on the auxiliary material's dynamic viscosity. To prevent numerical divergence caused by the auxiliary material's dynamic viscosity in extremely low ranges, an inverse proportional function with a lower limit truncation is used to calculate the superimposed weight, expressed by the following formula: Simultaneously, the physical contour radius of the chassis is obtained, and the expansion gain is calculated based on the dynamic remaining effective time. To prevent excessively large dynamic remaining effective time from causing abnormal expansion of the expansion radius and covering the passable area, a linear scaling function with upper limit normalization is used to calculate the superposition radius, and its formula is expressed as follows: In the two formulas above, This represents the calculated surface roughness penalty weight. This represents the system's preset base penalty weight. This represents the system's preset penalty ratio constant. This indicates the dynamic viscosity of the auxiliary material input in the preceding step. This indicates the system's preset minimum viscosity protection threshold. This represents the calculated obstacle expansion radius parameter. This represents the obtained physical contour radius of the chassis. This represents the system's preset expansion ratio constant. This indicates the dynamic remaining valid time of the preceding input. This represents the maximum effective time reference window preset by the system. In this scheme, the basic penalty weight is preferably 1.5, the penalty ratio constant is preferably 10, the minimum viscosity protection threshold is preferably 10 Pascals per second, the chassis physical contour radius is preferably 0.35 meters, the expansion ratio constant is preferably 0.3 meters, and the maximum effective time reference window is preferably 3600 seconds.

[0087] After parameter calculation, a static obstacle map and a surface roughness layer of the current working area are acquired through the airborne environmental perception module. Both are stored in a two-dimensional raster format, with a coordinate system constructed using the physical edge of the map as the origin, and their raster physical resolution is strictly consistent with the output accuracy of the airborne environmental perception module. For the static obstacle map, an accurate Euclidean distance transformation algorithm is used to traverse the entire map raster, calculating the true distance field from each passable raster to the nearest static obstacle boundary raster. Based on this, the dilation cost is calculated for each passable raster, expressed by the formula: In this formula, This represents the inflation value of the currently passable grid. This represents the Euclidean distance from the boundary of the nearest static obstacle to the currently accessible grid cell, calculated using the precise Euclidean distance transformation algorithm. This represents the radius of expansion of the obstacle.

[0088] For the surface roughness layer, the original roughness values ​​of each grid cell are globally weighted and calculated with the surface roughness penalty weight. To prevent excessively large weights from causing values ​​to exceed limits, an upper limit truncation operation is performed to generate a surface roughness penalty cost layer, expressed by the formula: In this formula, This represents the roughness cost after the current grid is truncated. Indicates the surface roughness penalty weight. This represents the original roughness value of the current raster in the surface roughness layer. This represents the system's preset upper limit for local cost truncation. In this scheme, the upper limit for local cost truncation is preferably 1.

[0089] After generating the aforementioned independent feature layers, spatial feature fusion calculations are performed. For each grid cell in the passable region (excluding obstacles), its roughness gradient value in a 3x3 spatial neighborhood is calculated. To ensure global executability, a nearest-edge pixel copying strategy is used to fill the boundary values ​​for edge grid cells at map boundary locations. Subsequently, the roughness change rate is calculated using the center difference method, expressed by the formula: The roughness gradient value is directly multiplied by the dilation cost of the grid, and an upper limit truncation is performed simultaneously to generate a coupled penalty cost representing the risk of chassis near-obstacle slippage. The formula is expressed as follows: In the two formulas above, This represents the gradient value of the rate of change of roughness of the current raster within a unit space. This indicates the row index coordinate of the current raster in the two-dimensional map matrix. This indicates the column index coordinate of the current raster in the two-dimensional map matrix. and These represent the original roughness values ​​of the current raster's positive and negative adjacent raster cells in the row direction, respectively. and These represent the original roughness values ​​of the current raster's positive and negative adjacent raster values ​​in the column direction, respectively. Indicates the physical resolution size of the raster. This represents the cost of the coupling penalty after truncation. This represents the inflation cost of the current raster. This indicates the upper limit of local cost truncation. In this scheme, the preferred grid physical resolution size is 0.05 meters.

[0090] Subsequently, the truncated anisotropic cost values ​​are synchronously input into the exponential fusion function for smoothing calculation to obtain the feasible fusion cost value, expressed by the formula as follows: In this formula, This represents the feasible fusion cost obtained through smoothing calculations. This represents the system's preset global cost upper limit constant. Indicates the inflation weighting factor. Represents the roughness weighting factor. Let represent the coupling penalty weighting factor, where all three are non-negative real numbers and satisfy the strict normalization constraint that the sum of the three equals one. This indicates inflationary value. Represents the cost of roughness. This represents the coupling penalty cost. The coordinates of prohibited boundaries are extracted from the static obstacle map, and the corresponding raster values ​​are directly assigned to a system-preset prohibited constant. This prohibited constant is strictly greater than the global cost upper limit constant. In this scheme, the preferred global cost upper limit constant is 253, the preferred prohibited constant is 254, the preferred inflation weight factor is 0.4, the preferred roughness weight factor is 0.3, and the preferred coupling penalty weight factor is 0.3. Furthermore, the prohibited constant does not participate in the conventional addition of the total path cost in the subsequent path planning algorithm; instead, it exists as a Boolean absolute collision marker. During the path node expansion search, a collision rejection judgment is performed beforehand: if the cost reading value of the current candidate raster is equal to or greater than the prohibited constant, the algorithm will directly discard the node and refuse to add it to the open list.

[0091] Finally, based on the distribution sequence of the original spatial coordinates, the prohibition constant and the drivable fusion cost matrix are reconstructed, outputting a complete spatial cost map and pushing it to the global data flow bus. This step effectively transforms the rheological material state into spatial navigation constraints. Through numerical boundary protection and smooth fusion of multi-source costs, the absolute stability of the underlying algebraic operations is ensured, and adaptive scaling of the chassis obstacle avoidance buffer distance and path penalty cost is achieved, guaranteeing transportation safety under complex operating conditions.

[0092] S4. Generate a global avoidance path based on the spatial cost map and calculate the predicted total time along the global avoidance path by combining the chassis kinematic parameters;

[0093] Furthermore, in S4, the steps of generating a global avoidance path based on the spatial cost map using a path planning algorithm, and calculating the predicted total time along the global avoidance path in conjunction with chassis kinematic parameters, specifically include:

[0094] Obtain the current starting coordinates and target coordinates of the chassis, execute the path planning algorithm in the spatial cost map, and generate a global avoidance path containing discrete path nodes;

[0095] Analyze the chassis kinematic parameters and extract the rated linear velocity and rated angular acceleration;

[0096] The basic straight-line driving time is calculated by combining the spatial distance between discrete path nodes and the rated linear velocity. The corner nodes in the discrete path nodes are extracted, and the turning acceleration and deceleration time is calculated based on the angle difference of the corner nodes and the rated angular acceleration.

[0097] Add the time taken for the basic straight-line driving to the time taken for turning and acceleration / deceleration, and output the predicted total time.

[0098] Specifically, a spatial cost map is read from the data stream output from the preceding steps. The value of each grid cell in this spatial cost map has been established with a one-to-one correspondence between the actual road surface skidding risk and traffic constraints. To allow it to be directly coupled with physical spatial distance for subsequent integral calculations, data parsing and mapping normalization are first performed on the spatial cost map. Grids with values ​​equal to a preset prohibition constant are extracted and directly marked as inaccessible constraints. For the remaining passable grids, a mapping operation is performed, dividing the values ​​by a preset global cost upper limit constant, to linearly normalize them to a dimensionless range of zero to one, forming a normalized fusion cost value for the algorithm. Simultaneously, the current starting coordinates of the chassis output from the positioning module and the target coordinates issued by the task system are read. Both the starting coordinates and the target coordinates are derived from the spatial cost map. Figure 1 It uses a consistent raster coordinate system for expression, and can directly participate in path search calculations without additional coordinate transformations.

[0099] A cost-cumulative heuristic search algorithm is executed on the spatial cost map to construct a global avoidance path. Node expansion adopts an eight-neighbor rule, that is, expansion is performed on the current node along eight adjacent grids in four directions: horizontal, vertical, and diagonal. For adjacent nodes in the horizontal or vertical direction, the physical distance between nodes is taken as the physical resolution size of a single grid cell. For diagonally adjacent nodes, the physical distance between nodes is taken as... To ensure geometric consistency in path length calculation, the comprehensive cost function for any node is defined during the search process as follows: .in, Represents a node The overall evaluation cost Represents the distance from the starting coordinates to the current node. The cumulative cost of the path, Indicates the current node Heuristic cost estimation to target coordinates This represents the heuristic dynamic weighting coefficients. The calculation method is as follows .in, This represents the distance from the starting node to the current node. The total number of path segments traversed. It means greater than zero and less than or equal to zero. Natural number index, Indicates the first The value of normalized fusion of the endpoint raster of the path segment. Indicates the first The actual physical length of the segment path, when the first segment path... When the segment is horizontal or vertical Pick When the first When the segment is diagonal Pick , This represents the actual physical resolution size corresponding to a single grid cell, preferably 0.1 meters. This represents the preset dimensionless cost conversion coefficient, preferably 0.1. It should be noted that, through this improved formula, this scheme explicitly decouples the node cost from the purely physical movement distance. With additional risk and penalty costs This avoids the physical distance being counted twice, ensuring that the physical meaning of path cost is clear and fair for comparison. Euclidean distance is used for estimation, and its expression is: .in, and These represent the horizontal and vertical grid coordinates of the target node, respectively. and Representing the current node The horizontal and vertical grid coordinates. Weighting coefficients are introduced into the comprehensive cost function. The decoupling calculation of the basic movement distance ensures strict uniformity in magnitude and dimension between the historical cumulative cost and the future heuristic cost, guaranteeing the effective guidance of the heuristic function and avoiding the risk of the algorithm degenerating into breadth-first search due to the addition of mixed quantities. When the original map value of the target expansion node is detected to be equal to the preset prohibition constant, the node is skipped directly and does not participate in the expansion, triggering branch pruning. The search terminates when the target node is first removed from the open list with the minimum comprehensive cost, at which point the path cost from the starting node to the target node has reached the global optimum. After the search is completed, the output consists of a node sequence. The resulting global avoidance path consists of nodes that correspond to actual passable grid locations in the spatial cost map. This indicates the total number of path nodes.

[0100] A continuity analysis is performed on the global avoidance path, mapping the discrete node sequence to the actual physical path. For adjacent nodes... and Its horizontal grid coordinates are respectively and The vertical grid coordinates are respectively and Calculate the corresponding physical distance .in, It means greater than zero and less than Natural number index, Indicates the first The actual physical length of each path segment. The total path length is obtained by summing the lengths of all path segments. Extract the rated linear velocity from the chassis kinematics parameter configuration data package. This parameter corresponds to the stable cruising speed of the chassis on a road surface with a standard coefficient of adhesion, preferably 0.5 meters per second. The basic straight-line driving time is calculated as follows: .in, This indicates the time taken for the chassis to travel along the basic straight-line path of the global avoidance route.

[0101] To accurately reflect the impact of changes in direction on travel time, corner identification and time calculation are performed on path nodes. For three consecutive nodes in the path, they are as follows: , and The corresponding horizontal grid coordinates are denoted as follows: , and The corresponding vertical grid coordinates are denoted as follows: , and Calculate the path vectors of the first and second segments separately, and obtain the node values ​​through vector dot product. absolute value of the angle amplitude at the location .in, Represents a node The absolute value of the change in path direction ranges from zero to... .when Greater than the angle threshold At that time, the node It is determined to be a corner node, where The preferred value is 0.1 radians. Assume all corner nodes have a total... The first, in the set of corner nodes The absolute value of the angle difference between the three corner nodes is denoted as . ,in greater than zero and less than or equal to Natural number index.

[0102] Extract the rated angular acceleration of the chassis The preferred parameter is 1.0 radians per square second. Under the symmetrical dynamic constraints of uniform angular acceleration to the midpoint and then uniform angular deceleration to the completion of the turn during the chassis steering process, and assuming that the angular velocity at the beginning and end of each turn is zero, the steering time at a single turning node is... .in, Indicates the first The time taken for a single turn at each corner node. The total turning time is obtained by summing the times taken at all corner nodes. .in, This represents the cumulative time taken for the chassis to perform all steering maneuvers along the global avoidance path. The predicted total time is obtained by adding the base straight-line driving time to the total turning time. .in, This represents the predicted total time for the chassis to travel along the global avoidance path from the starting coordinates to the target coordinates. This predicted total time... A rigorous comparison is performed with the dynamic remaining effective time of auxiliary materials output in the previous steps: if If the remaining effective time exceeds the dynamic limit, it is determined that the path will cause the auxiliary materials to solidify and fail during transportation, thus abandoning the path and triggering replanning or system alarm; if the time constraint is met, the result is pushed to the global data flow bus as a core time indicator for later use.

[0103] By introducing spatial normalization mapping and node dynamics analysis, seamless coupling between the underlying environmental layer data and the chassis mechanical execution capability is achieved, enabling the generated obstacle avoidance path to simultaneously possess absolute obstacle avoidance safety, material chemical state effectiveness, and engineering time predictability.

[0104] S5. If the predicted total time is less than the dynamic remaining effective time, the global avoidance path is confirmed as an executable route; otherwise, a termination instruction is generated.

[0105] Furthermore, in S5, the steps for confirming a global evasion path as an executable route specifically include:

[0106] Extract the original coordinate sequence corresponding to the global avoidance path;

[0107] The curve fitting algorithm is invoked to smooth the original coordinate sequence, generating a smooth space curve.

[0108] Extract the discrete coordinate data of the smoothed space curve, encapsulate the discrete coordinate data into a control task data package containing discrete coordinate data, velocity constraint parameters and task identifier, push the control task data package into the control buffer queue, and mark the global avoidance path status as an executable route.

[0109] Specifically, the total prediction time is read from the data stream output from the preceding steps. And the dynamic remaining valid time issued in real time by the task scheduling module. Both are measured in seconds and operate on a unified time base. The execution time feasibility assessment calculation is performed when... Strictly less than When the current global avoidance path is determined to be an executable route that meets the task time limit constraints; when Greater than or equal to If the path is determined to be unfinishable under the current operating conditions, a termination command is generated and sent to the execution control module via the data bus. The termination command contains at least a task identifier and a termination status code, which is used to trigger the chassis to enter a safe stop or wait for replanning state.

[0110] After determining that the global avoidance path is an executable route, coordinate mapping is performed on the global avoidance path node sequence output by the previous steps. The horizontal and vertical grid coordinates corresponding to each node are extracted and combined with the grid physical resolution size defined in the previous steps. Converting this to actual physical space coordinates, the conversion formula is: the actual physical horizontal coordinate of a node equals its horizontal raster coordinate multiplied by... The actual physical vertical coordinate is equal to its vertical grid coordinate multiplied by 1. The coordinate unit is meters. The cumulative arc length parameter between adjacent nodes is calculated synchronously, and the cumulative arc length parameter corresponding to the starting node is defined. For the first greater than one There are nodes In this formula, This represents the first step calculated and output by the preceding steps. The actual physical length of the path segment. It means greater than zero and less than Natural number index, This indicates the accumulation along the path from the starting node to the nth node. The actual physical arc length of each node.

[0111] A cubic spline curve fitting algorithm is used to smooth the transformed physical space coordinate sequence to eliminate the sudden speed changes and mechanical shocks caused by the path polylines. Lateral parametric curves are then constructed. With longitudinal parameter curve In the The polynomial formula for the interval of arc length of segment nodes is expressed as:

[0112] ;

[0113] as well as ;

[0114] In the above formula, The cumulative arc length of the path is the independent variable. , , , They represent the first The fitting coefficients of the cubic, quadratic, linear, and constant terms of the transverse parametric curve segment. , , , They represent the first The fitting coefficients for the corresponding terms of the longitudinal parametric curves are given. All the fitting coefficients are obtained by solving the following set of boundary condition equations: At each node, the endpoints of the fitted curves coincide with the actual physical coordinates of the node, thus satisfying the positional continuity condition; at the junction of two adjacent curve segments, the first derivative of the end of the preceding curve segment is equal to the first derivative of the beginning of the following curve segment, thus satisfying the directional continuity condition; the second derivative of the end of the preceding curve segment is equal to the second derivative of the beginning of the following curve segment, thus satisfying the curvature continuity condition; at the start and end nodes, the second derivatives of both the transverse and longitudinal parametric curves are strictly zero, thus satisfying the natural spline boundary condition.

[0115] Discretization sampling is performed based on the constructed smooth space curve, according to a preset sampling interval. Uniformly extract the entire curve, the first... The cumulative arc length parameter corresponding to each sampling point is: In this formula, This represents a sampling index that starts from a gradually increasing natural number, with the sampling termination condition being... The sampling interval does not exceed the total path length obtained from the previous step. The preferred distance is 0.05 meters. [The following appears to be a separate, unrelated sentence:] Each sampling point... By substituting the numerical values ​​into the polynomials corresponding to the arc length interval, the actual lateral and longitudinal coordinates of each sampling point in physical space are directly calculated. For each sampling point, the tangential direction angle of the chassis motion at that point is calculated using the analytical value of the first derivative of the spline function, expressed by the formula: In this formula, Represents the arctangent function in the four quadrants. Indicates the first Tangential direction angle at each sampling point and These represent the horizontal and vertical parameter curves respectively, with respect to the arc length parameter. The value of the first derivative at that point.

[0116] Simultaneously calculate the absolute value of the curve curvature at each sampling point, expressed by the formula: In this formula, the double vertical lines enclosing the molecule represent the absolute value operator. The smooth space curve represents the first... The absolute value of curvature at each sampling point and These represent the horizontal and vertical parameter curves respectively, with respect to the arc length parameter. The second derivative value at that point. Based on this absolute value of curvature, the velocity constraint parameters at that sampling point are calculated. Less than the preset minimum curvature threshold At that time, the allowable linear velocity at that point is directly set to be equal to the rated linear velocity. ;when Greater than or equal to When linear velocity is constrained by both curvature and lateral acceleration, the corresponding calculation formula is: In this formula, This represents the function that performs the minimum value operation. Indicates the first The allowable linear velocity at each sampling point This indicates the system's preset maximum permissible lateral acceleration of the chassis. The preferred value is 0.01 per meter. The preferred value is 0.8 meters per square second.

[0117] The actual physical horizontal and vertical coordinate data of all sampling points, along with the corresponding tangential direction angles and permissible linear velocity parameters, are extracted in a structured manner and combined to generate a control task data package. This control task data package also includes a global task identifier and timestamp information to maintain timing traceability. The control task data package is pushed into the underlying control buffer queue, which is strictly read and executed point by point by the chassis drive module according to the first-in-first-out principle. Simultaneously, the status identifier of the global avoidance path is updated to an executable route and synchronously written to the global status management module. This step transforms discrete environmental nodes into continuous underlying execution trajectories through a rigorous mathematical model and generates point-by-point dynamic speed limits under anti-skid curvature limit constraints, ensuring smooth driving of the chassis under the dual constraints of time limit requirements and physical sideslip limits.

[0118] S6. Generate a local trajectory based on the executable route, and adjust the centripetal acceleration boundary and angular velocity boundary according to the dynamic viscosity of the auxiliary material to generate control instructions that match the local trajectory.

[0119] Furthermore, in S6, the steps for generating control instructions that match the local trajectory specifically include:

[0120] Extract the curvature of the look-ahead path within a preset look-ahead distance range in the local trajectory. Using the current linear velocity and current angular velocity of the chassis as a reference, and combining the chassis's acceleration and deceleration constraint parameters with the adjusted centripetal acceleration boundary and angular velocity boundary, construct a dynamic velocity sampling space with upper and lower limits for linear velocity and angular velocity.

[0121] Multiple sets of sampled velocity pairs are generated in the dynamic velocity sampling space. An evaluation function is constructed that includes trajectory matching cost, obstacle distance cost, and velocity smoothing cost. The evaluation function is used to score the multiple sets of sampled velocity pairs. The sampled velocity pairs whose evaluation values ​​meet the preset extreme value conditions are selected as the target linear velocity and target angular velocity for matching the local trajectory.

[0122] The target linear velocity and target angular velocity are input into the chassis inverse kinematics model to calculate the target speed sequence of the chassis drive motor and generate control commands.

[0123] Specifically, the control task data packets generated in the preceding steps are extracted from the control buffer queue of the chassis controller. The discrete coordinate sequence, orientation angle sequence, and velocity constraint sequence within the data packets are parsed and read, serving as a reference for the local trajectory. Simultaneously, the dynamic viscosity of the auxiliary material on the current working surface is read from the sensor data stream of the auxiliary material dispensing system. The dynamic viscosity of the auxiliary material on the road surface directly alters the slip characteristics between the tire and the ground. Based on this, a dynamic boundary adjustment model is constructed, and the adjusted centripetal acceleration and angular velocity boundaries are calculated. The corresponding calculation formulas are as follows:

[0124] ;

[0125] ;

[0126] In this formula, This indicates the adjusted centripetal acceleration boundary. This indicates the adjusted angular velocity boundary. This represents the real-time dynamic viscosity of the auxiliary material, measured in Pascals per second. This indicates the basic centripetal acceleration limit of the chassis on a standard dry road surface. Indicates the limit of the basic angular velocity. This represents the viscosity reduction factor due to centripetal acceleration. This represents the viscosity reduction factor based on angular velocity. In this embodiment, The preferred value is 0.8 meters per square second. The preferred value is 0.15. The preferred value is 1.2 radians per second. The preferred value is 0.2. The adjusted dynamic boundary parameters will be directly used for hard clipping of the subsequent dynamic velocity sampling space.

[0127] Based on the current physical coordinates of the chassis and the preceding discrete coordinate sequence, a trajectory segment within a preset look-ahead distance is extracted, preferably 1.5 meters. The preceding direction angle sequence of adjacent sampling points within this segment is then extracted, combined with the sampling interval output from the previous steps. Calculate the look-ahead path curvature within the trajectory segment using the following formula:

[0128] ;

[0129] In this formula, This represents the summation operator. This represents the curvature of the look-through path, in meters. This indicates the total number of physical interval segments within the look-ahead trajectory segment. The double vertical bar enclosing the numerator represents the absolute value operator. and These represent the preceding direction angles corresponding to the next sampling point and the previous sampling point within the segment, respectively. This is the uniform sampling interval generated in the preceding steps. Simultaneously, the velocity constraint values ​​corresponding to all discrete points within this look-ahead trajectory segment in the preceding data packet are extracted, and the minimum value among them is taken as the local velocity constraint limit for the current control cycle. The current linear velocity and angular velocity of the chassis sensors are read, and combined with the acceleration and deceleration constraint parameters set by the chassis at the factory, a dynamic velocity sampling space with upper and lower limits for linear velocity and angular velocity is constructed. The boundary extremum calculation formula is as follows:

[0130] ;

[0131] ;

[0132] ;

[0133] ;

[0134] In this formula, This represents the function that performs the minimum value operation. This represents the function for finding the maximum value, and the square root symbol represents the square root operator. and These represent the upper and lower limits of linear velocity in the dynamic velocity sampling space, respectively. Indicates the current linear velocity of the chassis. and These represent the maximum linear acceleration and maximum linear deceleration allowed by the chassis hardware, respectively. This represents the execution cycle of the local control algorithm, preferably 0.1 seconds. This represents the preset minimum curvature threshold, preferably 0.01 per meter, used to prevent numerical divergence caused by the denominator being zero when the path approaches a straight line. This represents the local velocity constraint limit obtained above. In the formula, the lower velocity limit and zero are taken as the maximum value to prevent the lower limit of linear velocity from becoming negative under strong deceleration conditions. and These represent the upper and lower limits of angular velocity in the dynamic velocity sampling space, respectively. This indicates the current angular velocity of the chassis. and These represent the maximum permissible angular acceleration and maximum angular deceleration of the chassis, respectively. This dynamic velocity sampling space requires that the linear velocity and angular velocity of any legally sampled velocity pair must strictly fall within the aforementioned limiting boundaries.

[0135] Within the aforementioned dynamic velocity sampling space, multiple sets of sampled velocity pairs are generated according to a preset velocity resolution. For each sampled velocity pair consisting of sampled linear velocity and sampled angular velocity, the future spatial trajectory of the chassis within the prediction time window is predicted based on the kinematic dead reckoning model. The trajectory discretization recursive formula is as follows:

[0136] ;

[0137] ;

[0138] ;

[0139] In this formula, and This represents the lateral and longitudinal coordinates of the chassis at the next recursive time step. and This represents the horizontal and vertical coordinates of the current recursive time step. and This indicates the chassis orientation angle between the next moment and the current moment. and These represent the sampling linear velocity and sampling angular velocity of the current group, respectively. and Let cosine and sine be trigonometric functions respectively. This represents the simulation recursion step size of the dead reckoning model. It synchronously reads the local cost map updated in real-time by the environmental perception module, calculates the actual spatial Euclidean distance from each discrete point on the predicted trajectory to the nearest obstacle, and extracts the minimum distance value as the global obstacle distance for that set of trajectories. To ensure safety, a collision rejection mechanism is implemented; if the global obstacle distance is... If the sampled velocity pair is smaller than the safe radius of the chassis's physical outline, then the entire sampled velocity pair is discarded. For the retained sampled velocity pairs, a score is calculated using an evaluation function, the formula of which is:

[0140] ;

[0141] In this formula, This represents the overall evaluation value of the current sampling rate pair. , , , representing the dimensionless weighting coefficients for trajectory matching cost, obstacle distance cost, and velocity smoothing cost, respectively, with preferred values ​​of 0.4, 0.4, and 0.2. The nearest Euclidean distance is used to find the match between the end of the predicted trajectory and the preceding discrete coordinate sequence, and the nearest preceding discrete point is identified as the target point. This represents the absolute value of the angular deviation between the direction angle at the end of the predicted trajectory and the corresponding direction angle at the target point, and is a constant equal to pi. Used to normalize the absolute value to the interval between zero and one; This indicates the system's preset safe cutoff distance, when Greater than In this case, the ratio is directly taken as one; This represents the compensation amount for the smallest positive real number whose denominator is zero. Achieve linear normalization of velocity cost. Iterate through all legal sampled velocity pairs, select the optimal sampled velocity pair with the largest comprehensive evaluation value, and confirm them as the target linear velocity and target angular velocity for matching the local trajectory.

[0142] The calculated target linear velocity and target angular velocity are input into the chassis inverse kinematics model, and the underlying drive mapping calculation is performed. Taking a two-wheel differential chassis as an example, the calculation formula is as follows:

[0143] ;

[0144] ;

[0145] In this formula, and These represent the calculated target speeds of the left and right drive motors, respectively, in revolutions per minute (rpm). This represents the target linear velocity selected in the previous step. This indicates the target angular velocity selected in the previous step. Indicates the calibrated physical radius of the chassis drive wheels. This represents the calibrated physical wheelbase between the centers of the left and right drive wheels of the chassis, with a constant coefficient of thirty divided by pi. This is used to convert the angular velocity of the drive wheels from radians per second to motor speeds in revolutions per minute. The target speed sequence is encapsulated as a control command containing electrical pulse width modulation signals and sent to the motor drivers on both sides via the fieldbus for execution.

[0146] By converting the rheological properties of auxiliary materials into a rigid cutting boundary in the underlying velocity space, and combining kinematic trajectory prediction with a multi-dimensional safety obstacle avoidance evaluation model, an anti-slip and safe motor speed control command is output.

[0147] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An intelligent obstacle avoidance control method for a tiling robot during construction, characterized in that, Includes the following steps: S1. Obtain the chassis kinematic parameters, including the rated linear velocity and rated angular acceleration, as well as the ambient temperature and relative humidity parameters; S2. Input the ambient temperature and relative humidity parameters into the auxiliary material physical state calculation model, and output the dynamic viscosity and dynamic remaining effective time of the auxiliary material. S3. Adjust the surface roughness penalty weight according to the dynamic viscosity of the auxiliary material, adjust the obstacle expansion radius parameter according to the dynamic remaining effective time, and generate a spatial cost map based on the adjusted surface roughness penalty weight and the obstacle expansion radius parameter. S4. Based on the spatial cost map, execute the path planning algorithm to generate a global avoidance path, and combine the chassis kinematic parameters to calculate the predicted total time along the global avoidance path; S5. If the predicted total time is less than the dynamic remaining effective time, the global avoidance path is confirmed as an executable route; otherwise, a termination instruction is generated. S6. Generate a local trajectory based on the executable route, and adjust the centripetal acceleration boundary and angular velocity boundary according to the dynamic viscosity of the auxiliary material to generate control commands that match the local trajectory.

2. The intelligent obstacle avoidance control method for a tiling robot during construction as described in claim 1, characterized in that, In S2, the steps of outputting the dynamic viscosity and dynamic remaining effective time of the auxiliary material specifically include: Input the ambient temperature and relative humidity parameters into the physical state calculation model of the auxiliary material, and read the initial ratio viscosity, preparation time and failure viscosity threshold of the auxiliary material. Based on the ambient temperature and relative humidity parameters, the initial viscosity ratio and the time difference between the current time and the preparation time, the dynamic viscosity and dynamic curing rate of the auxiliary material are calculated. Based on the dynamic viscosity of the auxiliary material, the dynamic curing rate, and the failure viscosity threshold, the remaining time required for the dynamic viscosity of the auxiliary material to reach the failure viscosity threshold is calculated, and the dynamic remaining effective time is output. The physical state calculation model for the auxiliary materials is a model trained based on historical ambient temperature, historical relative humidity, historical initial viscosity, duration of time, and historical measured viscosity.

3. The intelligent obstacle avoidance control method for a tiling robot during construction as described in claim 2, characterized in that, In S3, the steps of adjusting the surface roughness penalty weight and adjusting the obstacle expansion radius parameter specifically include: Obtain the basic penalty weight, calculate the weight gain based on the dynamic viscosity of the auxiliary material, the weight gain is negatively correlated with the value of the dynamic viscosity of the auxiliary material, and calculate the basic penalty weight and the weight gain by superimposing them to output the surface roughness penalty weight. Obtain the physical contour radius of the chassis, calculate the expansion gain based on the dynamic remaining effective time, the expansion gain is positively correlated with the value of the dynamic remaining effective time, and calculate the obstacle expansion radius parameter by superimposing the physical contour radius of the chassis and the expansion gain.

4. The intelligent obstacle avoidance control method for a tiling robot during construction as described in claim 1, characterized in that, In S3, the step of generating the spatial cost map specifically includes: Obtain the static obstacle map and surface roughness layer of the current work area; An obstacle expansion cost layer is generated based on the static obstacle map and the obstacle expansion radius parameter; The surface roughness layer and the surface roughness penalty weight are weighted and calculated to generate a surface roughness penalty cost layer. The static obstacle map, the obstacle expansion cost layer, and the working surface roughness penalty cost layer are fused together to output a spatial cost map.

5. The intelligent obstacle avoidance control method for a tiling robot in the construction process according to claim 4, characterized in that, The step of performing a fusion calculation on the static obstacle map, the obstacle expansion cost layer, and the working surface roughness penalty cost layer to output a spatial cost map specifically includes: Extract the coordinates of the restricted boundary from the static obstacle map, and assign the grid value corresponding to the restricted boundary coordinates as a restricted constant; For the coordinates of the passable area other than the restricted boundary coordinates, extract the expansion cost value in the obstacle expansion cost layer and the roughness cost value in the working surface roughness penalty cost layer. Calculate the roughness gradient value of the passable area coordinates in a preset spatial neighborhood, and multiply the roughness gradient value with the expansion cost value to generate a coupling penalty cost value that characterizes the risk of slippage. The expansion cost, the roughness cost, and the coupling penalty cost are input into the exponential fusion function for smoothing calculation to obtain the feasible fusion cost. Based on the distribution sequence of the original spatial grid coordinates, the prohibition constant and the passable fusion cost matrix are reconstructed to output a spatial cost map.

6. The intelligent obstacle avoidance control method for a tiling robot in the construction process according to claim 1, characterized in that, In S4, the step of generating a global avoidance path based on the spatial cost map using the path planning algorithm, and calculating the predicted total time along the global avoidance path in conjunction with the chassis kinematic parameters, specifically includes: Obtain the current starting coordinates and target coordinates of the chassis, execute the path planning algorithm in the spatial cost map, and generate a global avoidance path containing discrete path nodes; Analyze the chassis kinematic parameters and extract the rated linear velocity and rated angular acceleration; The basic straight-line driving time is calculated by combining the spatial distance between the discrete path nodes and the rated linear speed. The corner nodes in the discrete path nodes are extracted, and the turning acceleration and deceleration time is calculated based on the angle difference of the corner nodes and the rated angular acceleration. Add the basic straight-line driving time to the turning acceleration and deceleration time to output the predicted total time.

7. The intelligent obstacle avoidance control method for a tiling robot in the construction process according to claim 1, characterized in that, In S5, the step of confirming the global evasion path as an executable route specifically includes: Extract the original coordinate sequence corresponding to the global avoidance path; The original coordinate sequence is smoothed by calling a curve fitting algorithm to generate a smooth space curve. Extract the discrete coordinate data of the smoothed space curve, encapsulate the discrete coordinate data into a control task data packet containing the discrete coordinate data, velocity constraint parameters and task identifier, push the control task data packet into the control buffer queue, and mark the global avoidance path status as an executable route.

8. The intelligent obstacle avoidance control method for a tiling robot in the construction process according to claim 1, characterized in that, In S6, the step of generating control instructions that match the local trajectory specifically includes: Extract the look-ahead path curvature within a preset look-ahead distance range in the local trajectory, and construct a dynamic velocity sampling space with upper and lower limits of linear velocity and angular velocity by taking the current linear velocity and current angular velocity of the chassis as a reference, and combining the acceleration and deceleration constraint parameters of the chassis and the adjusted centripetal acceleration boundary and angular velocity boundary. Multiple sets of sampled velocity pairs are generated within the dynamic velocity sampling space. An evaluation function is constructed that includes trajectory matching cost, obstacle distance cost, and velocity smoothing cost. The evaluation function is used to score the multiple sets of sampled velocity pairs. The sampled velocity pairs whose evaluation values ​​satisfy the preset extreme value conditions are selected as the target linear velocity and target angular velocity for matching the local trajectory. The target linear velocity and the target angular velocity are input into the chassis inverse kinematics model to calculate the target speed sequence of the chassis drive motor and generate control commands.

Images