An agent-based low-speed vehicle navigation method

Abstract

The present application relates to the technical field of path planning, and particularly relates to a low-speed vehicle navigation method based on an intelligent agent, which comprises the following steps: acquiring environment and vehicle parameters in real time; screening a first navigation area according to slope; determining a second navigation area according to time sequence correlation; determining a temporary navigation area according to path fluency; screening a target navigation area according to a visual language model; adjusting parameters according to reflection and sudden stop; and outputting updated navigation information. The present application collects multi-dimensional parameters in real time, combines a visual language map model in the navigation intelligent agent, and performs multi-level determination and optimization on the vehicle driving path. Through dynamic analysis of the ground environment characteristics and the vehicle motion state, and weighted comprehensive evaluation of the navigation parameters, and through adjustment of the collection length and the slope threshold to optimize the navigation sensing range, the problem that the vehicle cannot realize efficient and safe autonomous navigation in a complex warehouse environment due to excessive dependence on a static global map is effectively solved.

Description

Technical Field

[0001] This invention relates to the field of path planning technology, and in particular to a method for navigating low-speed vehicles based on intelligent agents. Background Technology

[0002] With the rapid development of industrial parks and large-scale warehousing and logistics environments, low-speed vehicles need to complete autonomous navigation efficiently and safely in complex and ever-changing parking areas. At the same time, factors such as dense shelving, narrow aisles, and multi-vehicle collaborative operations make vehicle travel paths susceptible to interference and constraints, posing a huge challenge to achieving stable and reliable autonomous navigation.

[0003] Chinese Patent Application Publication No. CN113946152A discloses a global path planning method, system, and low-speed commercial unmanned vehicle. The method includes: obtaining the entire map using an inertial navigation device, outputting all coordinate position information of the driving trajectory, and determining the current position of the vehicle through a positioning module; using a hash lookup algorithm to adapt all coordinate data of the map into a hash table and configuring index information for the coordinate data; inputting the destination coordinates and using a hash lookup algorithm to obtain the destination position index information; retrieving the current position and destination position information of the vehicle, and combining them with the vehicle's kinematic parameters to generate the vehicle trajectory route.

[0004] Therefore, the global path planning method has the following problems: It relies on a global map and inertial navigation equipment, making it difficult to respond in real time to dynamically changing warehouse environments and sudden obstacles, resulting in insufficient flexibility; the hash lookup algorithm of this method involves a large amount of computation when indexing and retrieving large-scale coordinate data, which can easily lead to a decrease in path planning speed and affect the real-time navigation performance of the vehicle; this method generates trajectories based only on the endpoint coordinates and vehicle kinematic parameters, lacking comprehensive consideration of complex factors such as channel congestion, thus limiting the safety and smoothness of the navigation path. Summary of the Invention

[0005] To address this issue, the present invention provides a low-speed vehicle navigation method based on an intelligent agent, which overcomes the problem in existing technologies where vehicles cannot achieve efficient and safe autonomous navigation in complex warehouse environments due to over-reliance on static global maps by sensing environmental parameters in real time and dynamically adjusting the navigation path.

[0006] To achieve the above objectives, the present invention provides a low-speed vehicle navigation method based on an intelligent agent, comprising:

[0007] Real-time data acquisition is obtained of the ground reflectivity and relative slope within a radius of a moving vehicle and a preset acquisition length within the parking area of ​​the industrial park warehouse; the proportion of shelf obstructions and the width of the drivable passage in each undetermined area within the parking area; and the deceleration frequency and emergency stop frequency of each mobile charging robot during its movement.

[0008] Several first navigation areas are determined based on the relative slope, a preset relative slope threshold, and the undetermined area where the vehicle is located;

[0009] Based on the ground reflectivity ratio, the shelf obstruction ratio in each of the first navigation areas, and the width of the drivable passage within the next preset first determination time, a time-series correlation determination is performed to determine several second navigation areas.

[0010] Based on the deceleration frequency and emergency stop frequency of the mobile charging robot during its movement to each of the second navigation areas within the next preset second determination time, the path smoothness is determined to identify several temporary navigation areas.

[0011] The target navigation region is selected from all the temporary navigation regions based on the preset visual language map model set in the navigation intelligent body and the navigation parameters of each temporary navigation region acquired in real time.

[0012] The preset collection length or the preset relative slope threshold is adjusted based on the ground reflectivity ratio and the emergency stop frequency of the vehicle heading to the target navigation area within the next preset adjustment time.

[0013] The output provides navigation prompts for the target navigation area after adjusting the preset acquisition length or the preset relative slope threshold.

[0014] Furthermore, the process of determining several first navigation areas based on the relative slope, a preset relative slope threshold, and the undetermined area where the vehicle is located includes:

[0015] When the relative slope is less than the preset relative slope threshold, the undetermined area is determined to be the first candidate area;

[0016] When the relative slope of all undetermined regions within the adjacent preset screening range of each first candidate region is less than the preset relative slope threshold, the first candidate region is determined to be the second candidate region;

[0017] Several first navigation regions are determined based on the distribution characteristics of all second candidate regions.

[0018] Furthermore, the process of determining several first navigation regions based on the distribution characteristics of all second candidate regions includes:

[0019] The Euclidean distance between any two adjacent second candidate regions is calculated based on the spatial coordinate information of each second candidate region to obtain several candidate distances;

[0020] When the candidate distance is less than a preset aggregation distance threshold, the two corresponding second candidate regions are divided into the same aggregation unit to obtain several aggregation regions;

[0021] Several first navigation regions are determined based on the continuous distribution characteristics of the aggregation regions.

[0022] Furthermore, the process of determining several first navigation regions based on the continuous distribution characteristics of the aggregation regions includes:

[0023] Calculate the spatial connectivity determination value between each of the aggregated regions to obtain several degrees of continuity;

[0024] When the continuity is greater than a preset continuity threshold, the corresponding aggregation region is determined to be the first navigation region, thereby determining a plurality of first navigation regions.

[0025] Furthermore, the process of determining several second navigation areas by performing a time-series correlation determination based on the ground reflectivity ratio, the shelf obstruction ratio in each of the first navigation areas, and the width of the drivable passageway within the next preset first determination time period includes:

[0026] A correlation judgment index is determined based on the ground reflectivity ratio, the shelf obstruction ratio, and the width of the drivable passage within the preset first judgment time period.

[0027] When the correlation determination index is greater than a preset determination index threshold, the first navigation region is determined to be the second navigation region, thereby identifying several second navigation regions.

[0028] Further, the process of determining several temporary navigation areas by judging the path smoothness based on the deceleration frequency and the emergency stop frequency of the mobile charging robot during its movement to each of the second navigation areas within the next preset second determination time period includes:

[0029] Calculate path smoothness based on all the deceleration frequencies and all the emergency stop frequencies;

[0030] When the path smoothness is greater than a preset smoothness threshold, the second navigation area is determined to be the temporary navigation area.

[0031] Furthermore, the process of selecting the target navigation region from all the temporary navigation regions based on the preset visual language map model set in the navigation intelligent body and the navigation parameters of each temporary navigation region acquired in real time includes:

[0032] The vehicle's current speed, vehicle steering angle, and forward visibility distance from the navigation parameters are input into the preset visual language map model to obtain several first filtering areas;

[0033] The straight-line distance between the vehicle and the target parking space in the temporary navigation area and the vehicle heading deviation in the navigation parameters are input into the preset visual language map model to obtain several second filtering areas;

[0034] The intersection of the first and second filtering regions is determined to be the candidate region.

[0035] The target navigation region is selected from all the temporary navigation regions based on all the navigation parameters in all candidate regions.

[0036] Furthermore, the process of selecting the target navigation region from all the temporary navigation regions based on all the navigation parameters in all candidate regions includes:

[0037] Calculate the comprehensive characterization value of each candidate region based on all the navigation parameters in each candidate region;

[0038] The candidate region corresponding to the maximum value of all the comprehensive characterization values ​​is marked to filter out the target navigation region.

[0039] Furthermore, the process of adjusting the preset acquisition length or the preset relative slope threshold based on the ground reflectivity ratio and the emergency stop frequency of the vehicle traveling to the target navigation area within the next preset adjustment period includes:

[0040] Calculate the standard deviation of the total ground reflectivity from the initial time of the preset adjustment period to each time to obtain several reflectivity fluctuation values;

[0041] Calculate the standard deviation of all emergency stop frequencies from the initial time of the preset adjustment duration to each time to obtain several emergency stop frequency fluctuation values;

[0042] Calculate the Pearson correlation coefficients of all the aforementioned reflectivity fluctuation values ​​and all the aforementioned emergency stop frequency fluctuation values ​​to obtain the adjustment judgment value;

[0043] Adjust the preset acquisition length or the preset relative slope threshold according to the adjustment judgment value and the preset adjustment judgment range.

[0044] Furthermore, the process of adjusting the preset acquisition length or the preset relative slope threshold according to the adjustment judgment value and the preset adjustment judgment range includes:

[0045] When the adjustment judgment value is greater than the maximum value of the preset adjustment judgment range, the preset acquisition length is increased according to the relative deviation between the adjustment judgment value and the maximum value of the preset adjustment judgment range;

[0046] When the adjustment judgment value is less than the minimum value of the preset adjustment judgment range, the preset relative slope threshold is increased according to the relative deviation between the adjustment judgment value and the minimum value of the preset adjustment judgment range.

[0047] Compared with existing technologies, the beneficial effects of this invention are as follows: by collecting parameters such as the proportion of ground reflection around the vehicle, relative slope, proportion of shelf obstructions, width of passageways, and operating status of mobile charging robots in real time, and combining them with the visual language map model within the navigation intelligence body, the vehicle's driving path is determined and optimized at multiple levels. This achieves closed-loop control from initial navigation area selection, temporal correlation, path smoothness evaluation to final target navigation area determination. In this process, through dynamic analysis of ground environment features and vehicle motion status, as well as weighted comprehensive evaluation of navigation parameters, the vehicle can autonomously select the optimal driving path in complex warehouse environments. At the same time, by adjusting the collection length and slope threshold to optimize the navigation perception range, the frequency of sudden stops and decelerations is reduced, improving traffic efficiency and safety. This achieves intelligent, stable, and reliable improvement in low-speed vehicle navigation, effectively solving the problem that vehicles cannot achieve efficient and safe autonomous navigation in complex warehouse environments due to over-reliance on static global maps.

[0048] Furthermore, by determining the relative slope of each undetermined area within the parking area in real time, candidate areas with lower slopes and continuous distribution can be effectively screened out. This is further combined with the spatial relationship between adjacent areas to form a stable first navigation area, thereby ensuring that the center of gravity and power distribution of low-speed vehicles are balanced during driving, reducing the risk of vehicle slippage or unstable speed, while optimizing the continuity and safety of the driving path, improving navigation efficiency and the reliability of passage within the warehouse environment.

[0049] Furthermore, by calculating the Euclidean distance between the second candidate regions and aggregating spatially similar regions based on a preset aggregation distance threshold, a continuously distributed aggregation region is formed, thereby effectively identifying the first navigation region with a compact and coherent structure. This not only ensures the continuity and stability of path selection for low-speed vehicles during driving, but also improves the reliability of navigation decisions and driving efficiency. At the same time, it reduces frequent turning and deceleration caused by path discontinuity or local obstacles, enhancing the vehicle's adaptability in complex warehouse environments.

[0050] Furthermore, by calculating the spatial connectivity between each aggregation region, the continuity and accessibility of the aggregation region can be quantitatively evaluated, thereby effectively selecting regions with good continuous distribution as the first navigation region. In this embodiment, continuity is used to measure the connection between regions, so that the navigation path is spatially coherent and smooth, and can take into account both vehicle driving safety and traffic efficiency, ensuring that the selected navigation region is both tightly clustered and easy for vehicles to pass through smoothly.

[0051] Furthermore, by performing time-series correlation analysis on the ground reflectivity ratio, shelf obstruction ratio, and drivable aisle width within a preset first determination time period, the system can comprehensively evaluate the traffic conditions and safety of each first navigation area. When the correlation determination index exceeds a preset threshold, the system automatically selects the area with better conditions as the second navigation area, thereby achieving dynamic response to environmental changes, ensuring the feasibility of the navigation path and the stability of vehicle driving, and effectively utilizing the interrelationship between spatial information, optical reflection characteristics, and aisle structure features to optimize path selection.

[0052] Furthermore, by statistically analyzing the deceleration frequency and emergency stop frequency of the mobile charging robot during its movement towards the second navigation area, this embodiment can quantify the smoothness of each path, thereby scientifically evaluating the traffic efficiency of different areas. When the smoothness of the path exceeds a preset threshold, the corresponding area can be determined as a temporary navigation area, effectively ensuring that the vehicle avoids areas that may lead to frequent deceleration or emergency stops when selecting a path, while also ensuring the safety and efficiency of navigation, and realizing the dynamic matching of vehicle traffic behavior and environmental constraints.

[0053] Furthermore, by combining the vehicle's current speed, steering angle, and forward visibility with the straight-line distance and heading deviation from the target parking space, the system can effectively consider the coordination between the vehicle's own motion state and spatial position, achieving a multi-dimensional evaluation of the temporary navigation area. This allows for the selection of the optimal navigation area that is both conducive to smooth vehicle driving and close to the target parking space, improving the reliability of navigation decisions and the efficiency of path execution. At the same time, it takes into account the vehicle's dynamic response and the constraints of the surrounding environment, ensuring the continuity and safety of the navigation process.

[0054] Furthermore, by comprehensively evaluating various navigation parameters in the candidate areas, such as vehicle speed, steering angle, forward visibility distance, straight-line distance to the target parking space, and heading deviation, a comprehensive characteristic value for each candidate area is calculated. This quantifies the navigation quality of each area and achieves a unified measurement of multi-dimensional information. The area with the highest comprehensive characteristic value is designated as the target navigation area. This effectively balances vehicle dynamics and environmental constraints, ensuring safe and stable navigation while allowing for rapid approach to the target parking space, thus improving overall navigation efficiency and reliability.

[0055] Furthermore, by analyzing the changes in the proportion of ground reflection and the frequency of sudden stops during the vehicle's journey to the target navigation area, the fluctuation values ​​are calculated, and adjustment judgment values ​​are obtained through correlation evaluation. This allows for dynamic adjustment of the preset data acquisition length or the preset relative slope threshold. This method links the environmental complexity during vehicle movement with the vehicle's motion state, enabling adaptive optimization of navigation parameters. This allows the vehicle to maintain stable driving under different ground conditions and reduces sudden stops, improving navigation safety and path efficiency.

[0056] Furthermore, by comparing the adjustment judgment value with the upper and lower limits of the preset adjustment judgment range, the relative deviation is calculated, and the preset acquisition length or preset relative slope threshold is dynamically increased to achieve adaptive adjustment of navigation parameters. This method can match the actual performance of the vehicle under different ground reflection and driving conditions with the parameter settings of the navigation environment, enabling the vehicle to maintain stable operation in complex road conditions, optimize route selection, improve driving safety and navigation efficiency, and ensure that the navigation system can flexibly respond to environmental fluctuations. Attached Figure Description

[0057] Figure 1 This is a flowchart of the agent-based low-speed vehicle navigation method in this embodiment;

[0058] Figure 2 This is a logic diagram for determining the first candidate region in this embodiment;

[0059] Figure 3 This is a logic diagram for determining the aggregation region in this embodiment;

[0060] Figure 4 This is a logic diagram for determining the first navigation region in this embodiment. Detailed Implementation

[0061] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0062] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0063] Please see Figure 1The diagram shows a flowchart of a low-speed vehicle navigation method based on an intelligent agent in this embodiment. This embodiment provides a low-speed vehicle navigation method based on an intelligent agent, including: real-time acquisition of the ground reflectivity ratio and relative slope within a parking area of ​​an industrial park warehouse, centered on a moving vehicle and with a preset acquisition length as the radius; the proportion of shelf obstructions and the width of drivable passageways in each undetermined area within the parking area; and the deceleration frequency and emergency stop frequency of each mobile charging robot during its movement; determining several first navigation areas based on the relative slope, a preset relative slope threshold, and the undetermined area where the vehicle is located; and further determining the navigation method based on the ground reflectivity ratio, the proportion of shelf obstructions in each of the first navigation areas, and the width of the drivable passageways within the next preset first determination time. The system performs a time-series correlation determination to identify several second navigation areas; it then determines several temporary navigation areas based on the deceleration frequency and emergency stop frequency of the mobile charging robot during its movement to each of the second navigation areas within a next preset second determination time period; it further selects a target navigation area from all the temporary navigation areas based on a preset visual language map model set within the navigation intelligence and the navigation parameters of each temporary navigation area acquired in real time; it adjusts the preset acquisition length or the preset relative slope threshold based on the ground reflectivity ratio and emergency stop frequency of the vehicle heading to the target navigation area within a next preset adjustment time period; and finally outputs navigation prompts for the target navigation area re-determined after adjusting the preset acquisition length or the preset relative slope threshold.

[0064] In this embodiment, the site of a warehouse parking area in an industrial park has local ground reflections, obstructions from shelves and large equipment, and multiple branching passages with varying widths. The system is equipped with a vehicle-mounted color camera, an inertial measurement unit and a visual SLAM module, a depth point cloud sensor, and a motion control unit for a mobile charging robot with speed and steering angle feedback. The parameters collected in real time include the ground reflectivity and relative slope in the detection area around the vehicle, the proportion of shelf obstructions and the width of the drivable passage in each area to be determined, and the deceleration frequency and emergency stop frequency of the mobile charging robot during its movement. The ground reflectivity is obtained by extracting ground pixels through semantic segmentation after the environmental image is acquired by the onboard camera, and then calculating the proportion of bright pixels by performing grayscale or high dynamic range brightness statistics on the ground pixels and comparing them with the reflectivity threshold. The relative slope is obtained by fusing the local elevation information output by visual SLAM with the pitch angle of the inertial measurement unit, and is expressed as a local height difference or attitude angle. The width of the drivable passage is obtained by measuring the width of the continuous free grid band after the occupied grid or point cloud projection plane generated by SLAM. The proportion of shelf obstructions is obtained by ground segmentation and clustering identification of shelves and large equipment clusters by point cloud sensors, and then calculating the volume or occupancy ratio of the point cloud of such obstacles in the detection area. The deceleration frequency and emergency stop frequency of the mobile charging robot are obtained by the speed and acceleration log statistics of its motion control system. A speed decrease event per unit time is counted as a deceleration event, and a stop judgment triggered per unit time is counted as an emergency stop event. The system simultaneously maintains navigation parameters for each temporary navigation area in real time. These parameters include the vehicle's current speed, steering angle, forward visibility distance, straight-line distance between the vehicle and the target parking space in the temporary navigation area, and vehicle heading deviation. The vehicle's current speed is calculated by merging odometer readings from the wheel odometer or inertial measurement unit. The steering angle is obtained from feedback from the steering angle sensor or steering mechanism. The forward visibility distance is obtained by combining depth point cloud data or a forward-facing camera with depth estimation. The straight-line distance between the vehicle and the target parking space in the temporary navigation area is calculated using Euclidean straight-line distances calculated from the vehicle's pose and the target parking space's coordinates. The vehicle heading deviation is calculated as the difference between the current heading angle and the azimuth angle pointing towards the target parking space. The outputs from these sensors are input into the visual language map model in a time-series manner for multimodal fusion and evaluation. The fusion results are used for temporal correlation determination, path fluency determination, and comprehensive characterization value calculation, ensuring that the parameter acquisition methods, calculation steps, and data flow are fully disclosed and implementable in practice.

[0065] In this embodiment, the navigation prompts include the specific location coordinates, boundary range, entrance direction indication, and recommended entry order or route information of the target navigation area, which are used to guide the vehicle to accurately enter and park in the selected target navigation area.

[0066] The preset acquisition length is the detection distance used to determine the vehicle's surrounding sensing radius. It depends on the vehicle's speed, the sensor's maximum detection distance, and the complexity of the terrain, and is typically set between 2 and 6 meters. In this embodiment, it is set to 4 meters, which reduces redundant data and improves real-time path calculation efficiency while ensuring necessary sensing coverage. The preset relative slope threshold is a slope limit value used to determine road surface feasibility. It depends on the vehicle's chassis passability, tire friction coefficient, and load conditions, and is typically set between 2 and 8 degrees. In this embodiment, it is set to 5 degrees, which effectively avoids slippage or instability caused by steep slopes, improving navigation and parking safety. The preset [missing information - likely a number]... The first determination time is a time window used to calculate parameters such as the ground reflectivity ratio and shelf obstruction for time-series correlation determination. It depends on the vehicle speed and the rate of dynamic change of the environment, and is usually set between 2 and 10 seconds. In this embodiment, it is set to 5 seconds, which can balance determination accuracy and response speed and identify feasible navigation areas in a timely manner. The second determination time is a time window used to statistically analyze the deceleration frequency and emergency stop frequency of the mobile charging robot to determine the path smoothness. It depends on the vehicle motion characteristics and path complexity, and is usually set between 3 and 12 seconds. In this embodiment, it is set to 6 seconds, which can effectively capture short-term motion fluctuations and determine path smoothness.

[0067] The pre-set visual-language map model set within the navigation intelligent body is a multimodal perception model based on deep convolutional neural networks and visual-language joint encoding. The model includes a visual encoder, a language encoder, and a fusion module. The visual encoder uses ResNet50 to extract environmental image features, the language encoder uses a bidirectional Transformer to process navigation task instructions and target descriptions, and the fusion module maps visual and linguistic features to a unified embedding space through an attention mechanism. Model training uses a large-scale industrial park simulation dataset and a real-world dataset, including scenarios with different lighting, ground reflections, channel layouts, and vehicle motion states. During training, a combination of supervised learning and contrastive learning is used, with loss functions including cross-entropy loss and similarity maximization loss. After training, the model can generate a priority score for navigation areas based on the input environmental image and navigation parameters, namely vehicle speed, steering angle, forward visibility distance, straight-line distance between the vehicle and the target parking space, and heading deviation, thus enabling the selection of target navigation areas from temporary navigation areas.

[0068] By collecting parameters in real time, such as the proportion of ground reflection around the vehicle, relative slope, proportion of shelf obstructions, width of passageways, and operating status of mobile charging robots, and combining them with the visual language map model within the navigation intelligence, the vehicle's driving path is determined and optimized at multiple levels. This achieves closed-loop control from initial navigation area selection, temporal correlation, path smoothness evaluation to final target navigation area determination. During this process, through dynamic analysis of ground environment features and vehicle motion status, as well as weighted comprehensive evaluation of navigation parameters, the vehicle can autonomously select the optimal driving path in complex warehouse environments. At the same time, by adjusting the collection length and slope threshold to optimize the navigation perception range, the frequency of sudden stops and decelerations is reduced, improving traffic efficiency and safety. This achieves intelligent, stable, and reliable improvement in low-speed vehicle navigation, effectively solving the problem that vehicles cannot achieve efficient and safe autonomous navigation in complex warehouse environments due to over-reliance on static global maps.

[0069] Please see Figure 2 As shown, this is a logic diagram for determining the first candidate region in this embodiment. In this embodiment, the process of determining several first navigation regions based on the relative slope, a preset relative slope threshold, and the undetermined region where the vehicle is located includes: when the relative slope is less than the preset relative slope threshold, determining the undetermined region as a first candidate region; when the relative slope of all undetermined regions within the adjacent preset filtering range of each first candidate region is less than the preset relative slope threshold, determining the first candidate region as a second candidate region; and determining several first navigation regions based on the distribution characteristics of all second candidate regions.

[0070] By determining the relative slope of each undetermined area within the parking area in real time, candidate areas with lower slopes and continuous distribution can be effectively screened out. Furthermore, by combining the spatial relationship between adjacent areas, a stable first navigation area can be formed, thereby ensuring that the center of gravity and power distribution of low-speed vehicles are balanced during driving, reducing the risk of vehicle slippage or unstable speed, while optimizing the continuity and safety of the driving path, improving navigation efficiency and the reliability of passage within the warehouse environment.

[0071] Please see Figure 3 As shown, this is a logic diagram for determining the aggregation region in this embodiment. In this embodiment, the process of determining several first navigation regions based on the distribution characteristics of all second candidate regions includes: calculating the Euclidean distance between any two adjacent second candidate regions based on the spatial coordinate information of each second candidate region to obtain several candidate distances; when the candidate distance is less than a preset aggregation distance threshold, dividing the corresponding two second candidate regions into the same aggregation unit to obtain several aggregation regions; and determining several first navigation regions based on the continuous distribution characteristics of the aggregation regions.

[0072] The preset aggregation distance threshold is a distance parameter used to determine whether adjacent second candidate areas should be divided into the same aggregation unit. It depends on the spatial layout of the warehouse parking area and the vehicle driving accuracy. It is usually set between 20 cm and 100 cm. In this embodiment, it is set to 50 cm, which can ensure navigation continuity while avoiding excessive aggregation that leads to inaccurate path selection.

[0073] By calculating the Euclidean distance between the second candidate regions and aggregating spatially similar regions based on a preset aggregation distance threshold, a continuously distributed aggregation region is formed, thereby effectively identifying a compact and coherent first navigation region. This not only ensures the continuity and stability of path selection for low-speed vehicles during driving, but also improves the reliability of navigation decisions and driving efficiency. At the same time, it reduces frequent turning and deceleration caused by path discontinuity or local obstacles, enhancing the vehicle's adaptability in complex warehouse environments.

[0074] Please see Figure 4 As shown, this is the determination logic diagram for determining the first navigation region in this embodiment. In this embodiment, the process of determining several first navigation regions based on the continuous distribution characteristics of the aggregated regions includes: calculating the spatial connectivity determination value between each aggregated region to obtain several continuity values, where C=2×E / [N×(N-1)], C is the continuity value, E is the number of edges with spatial connection relationships between aggregated regions, and N is the number of aggregated regions; when the continuity value is greater than a preset continuity threshold, the corresponding aggregated region is determined to be the first navigation region, thereby determining several first navigation regions.

[0075] The preset continuity threshold is a criterion for judging the spatial connectivity of aggregated areas. It depends on the number and layout density of aggregated areas and is usually set between 0 and 1. In this embodiment, it is set to 0.6, which can ensure that only aggregated areas with relatively tight spatial connections and convenient for continuous vehicle travel are selected as the first navigation area, thereby improving the continuity of the navigation path and the safety of passage.

[0076] By calculating the spatial connectivity between each aggregation region, the continuity and accessibility of the aggregation region can be quantitatively evaluated, thereby effectively selecting regions with good continuous distribution as the first navigation region. In this embodiment, continuity is used to measure the connection between regions, so that the navigation path is spatially coherent and smooth, and can take into account both vehicle driving safety and traffic efficiency, ensuring that the selected navigation region is both tightly clustered and easy for vehicles to pass through smoothly.

[0077] Specifically, the process of determining several second navigation areas by performing a time-series correlation determination based on the ground reflectivity ratio, the proportion of shelf obstructions in each of the first navigation areas, and the width of the drivable passageway within the next preset first determination time period includes: determining a correlation determination index based on the ground reflectivity ratio, the proportion of shelf obstructions, and the width of the drivable passageway within the preset first determination time period, wherein... Q is the correlation determination index, T1 is the preset first determination time, a is the preset proportion weight, R(t) is the ground reflection proportion at time t within the preset first determination time, R' is the preset reflection proportion threshold, b is the preset occlusion weight, Y(t) is the proportion of shelf obstructions at time t within the preset first determination time, Y' is the preset obstruction proportion threshold, c is the preset width weight, I(t) is the width of the drivable passage at time t within the preset first determination time, and I' is the preset width threshold; when the correlation determination index is greater than the preset determination index threshold, the first navigation area is determined to be the second navigation area, thereby determining several second navigation areas.

[0078] The preset judgment index threshold is a value determined based on historical navigation data and actual driving safety requirements. It depends on the variation range of parameters such as ground reflection, the proportion of shelf obstructions, and the width of the passage. It is usually set between 0.5 and 0.9. In this embodiment, it is set to 0.7, which can effectively distinguish between areas with good traffic conditions and restricted areas, thereby ensuring the safe and reliable selection of the second navigation area.

[0079] By performing time-series correlation analysis on the ground reflectivity ratio, shelf obstruction ratio, and drivable aisle width within a preset first judgment period, the system can comprehensively evaluate the traffic conditions and safety of each first navigation area. When the correlation judgment index exceeds a preset threshold, the system automatically selects the area with better conditions as the second navigation area, thereby achieving dynamic response to environmental changes, ensuring the feasibility of the navigation path and the stability of vehicle driving, and effectively utilizing the interrelationship between spatial information, optical reflection characteristics, and aisle structure features to optimize path selection.

[0080] Specifically, the process of determining several temporary navigation areas by judging the path smoothness based on the deceleration frequency and the emergency stop frequency of the mobile charging robot during its movement to each of the second navigation areas within the next preset second determination time period includes: calculating the path smoothness based on all the deceleration frequencies and all the emergency stop frequencies, wherein... F represents path smoothness, T2 represents the preset second determination time, fs(t) represents the deceleration frequency at time t within the preset second determination time, and fh(t) represents the emergency stop frequency at time t within the preset second determination time; when the path smoothness is greater than the preset smoothness threshold, the second navigation area is determined to be the temporary navigation area.

[0081] The preset smoothness threshold is a reference value used to determine whether the path is smooth. It depends on the statistical distribution of the deceleration frequency and emergency stop frequency of the mobile charging robot in different traffic environments. It is usually set between 0.5 and 0.9. In this embodiment, it is set to 0.75, which can effectively filter out temporary navigation areas with smooth traffic and less interference, and improve the safety and efficiency of vehicle navigation.

[0082] By statistically analyzing the deceleration frequency and emergency stop frequency of the mobile charging robot during its movement to the second navigation area, this embodiment can quantify the smoothness of each path, thereby scientifically evaluating the traffic efficiency of different areas. When the smoothness of the path exceeds a preset threshold, the corresponding area can be determined as a temporary navigation area, effectively ensuring that the vehicle avoids areas that may cause frequent deceleration or emergency stops when selecting a path, while also ensuring the safety and efficiency of navigation, and realizing the dynamic matching of vehicle traffic behavior and environmental constraints.

[0083] Specifically, the process of selecting a target navigation region from all the temporary navigation regions based on the preset visual language map model set in the navigation intelligent body and the navigation parameters of each temporary navigation region acquired in real time includes: inputting the vehicle's current speed, vehicle steering angle, and forward visibility distance from the navigation parameters into the preset visual language map model to obtain several first filtering regions; inputting the straight-line distance between the vehicle and the target parking space in the temporary navigation region and the vehicle's heading deviation from the navigation parameters into the preset visual language map model to obtain several second filtering regions; determining the intersection of the first filtering regions and the second filtering regions as the candidate region; and selecting the target navigation region from all the temporary navigation regions based on all the navigation parameters in all the candidate regions.

[0084] By combining the vehicle's current speed, steering angle, and forward visibility with the straight-line distance and heading deviation from the target parking space, the system effectively considers the coordination between the vehicle's own motion state and spatial position, enabling a multi-dimensional evaluation of the temporary navigation area. This allows for the selection of the optimal navigation area that is both conducive to smooth vehicle movement and close to the target parking space, improving the reliability of navigation decisions and the efficiency of path execution. At the same time, it takes into account the vehicle's dynamic response and the constraints of the surrounding environment, ensuring the continuity and safety of the navigation process.

[0085] Specifically, the process of selecting a target navigation region from all temporary navigation regions based on all navigation parameters in all candidate regions includes: calculating a comprehensive characterization value for each candidate region based on all navigation parameters in each candidate region, wherein, S is the comprehensive characterization value, wi is the weight corresponding to the i-th navigation parameter, Pi is the i-th navigation parameter, Pimax is the maximum value of the i-th navigation parameter in all candidate regions, and Pimin is the minimum value of the i-th navigation parameter in all candidate regions; the candidate regions corresponding to the maximum values ​​of all the comprehensive characterization values ​​are marked to filter out the target navigation region.

[0086] In this embodiment, the weight corresponding to the vehicle's current speed depends on the degree of influence of speed on path selection during navigation, and is usually set between 0 and 1. In this embodiment, it is set to 0.25, which can reflect the importance of vehicle driving stability in the comprehensive performance value; the weight corresponding to the vehicle's steering angle depends on the influence of steering changes on path safety and feasibility, and is usually set between 0 and 1. In this embodiment, it is set to 0.20, which can reflect the influence of steering difficulty in the comprehensive performance value; the weight corresponding to the forward visibility distance depends on the influence of the visibility range on obstacle avoidance ability, and is usually set between 0 and 1. In this embodiment, it is set to 0.20, which can reflect the safety of the forward environment in the comprehensive performance value; the weight corresponding to the straight-line distance between the vehicle and the target parking space in the temporary navigation area depends on the influence of distance on navigation priority, and is usually set between 0 and 1. In this embodiment, it is set to 0.25, which can reflect the path proximity in the comprehensive performance value; the weight corresponding to the vehicle's heading deviation depends on the influence of heading on driving efficiency and path accuracy, and is usually set between 0 and 1. In this embodiment, it is set to 0.10, which can reflect the heading adjustment requirements in the comprehensive performance value.

[0087] By comprehensively evaluating various navigation parameters in the candidate areas, such as vehicle speed, steering angle, forward visibility distance, straight-line distance to the target parking space, and heading deviation, a comprehensive characteristic value for each candidate area is calculated. This quantifies the navigation quality of each area and achieves a unified measurement of multi-dimensional information. The area with the highest comprehensive characteristic value is designated as the target navigation area. This effectively balances vehicle dynamics and environmental constraints, ensuring safe and stable navigation while quickly approaching the target parking space, thus improving overall navigation efficiency and reliability.

[0088] Specifically, the process of adjusting the preset acquisition length or the preset relative slope threshold based on the ground reflectivity ratio and emergency stop frequency of the vehicle traveling to the target navigation area within the next preset adjustment period includes: calculating the standard deviation of all ground reflectivity ratios from the initial time of the preset adjustment period to each time to obtain several reflectivity ratio fluctuation values; calculating the standard deviation of all emergency stop frequencies from the initial time of the preset adjustment period to each time to obtain several emergency stop frequency fluctuation values; calculating the Pearson correlation coefficient of all reflectivity ratio fluctuation values ​​and all emergency stop frequency fluctuation values ​​to obtain an adjustment determination value; and adjusting the preset acquisition length or the preset relative slope threshold based on the adjustment determination value and the preset adjustment determination range.

[0089] By analyzing the changes in the proportion of ground reflection and the frequency of sudden stops during a vehicle's journey to the target navigation area, the fluctuation values ​​are calculated, and adjustment judgment values ​​are obtained through correlation evaluation. This allows for dynamic adjustment of the preset data acquisition length or preset relative slope threshold. This method links the environmental complexity during vehicle movement with the vehicle's motion state, enabling adaptive optimization of navigation parameters. This allows the vehicle to maintain stable driving under different ground conditions and reduces sudden stops, improving navigation safety and path efficiency.

[0090] Specifically, the process of adjusting the preset acquisition length or the preset relative slope threshold according to the adjustment judgment value and the preset adjustment judgment range includes: when the adjustment judgment value is greater than the maximum value of the preset adjustment judgment range, increasing the preset acquisition length according to the relative deviation between the adjustment judgment value and the maximum value of the preset adjustment judgment range, wherein, L' is the increased preset acquisition length, L is the original preset acquisition length, k1 is the preset length adjustment coefficient, U is the adjustment judgment value, and Umax is the maximum value of the preset adjustment judgment range. When the adjustment judgment value is less than the minimum value of the preset adjustment judgment range, the preset relative slope threshold is increased according to the relative deviation between the adjustment judgment value and the minimum value of the preset adjustment judgment range. R' is the increased preset relative slope threshold, R is the original preset relative slope threshold, k2 is the preset threshold adjustment coefficient, and Umin is the minimum value of the preset adjustment judgment range.

[0091] The preset length adjustment coefficient is used to adjust the increase of the preset acquisition length. It depends on the vehicle's sensitivity to changes in ground reflection and is usually set between 0.01 and 0.1. In this embodiment, it is set to 0.05 to achieve smooth and effective acquisition length adjustment. The preset threshold adjustment coefficient is used to adjust the increase of the preset relative slope threshold. It depends on the vehicle's driving stability under different slopes and is usually set between 0.01 and 0.1. In this embodiment, it is set to 0.05 to ensure that the slope threshold adjustment is both sensitive and safe. The maximum value of the preset adjustment judgment range is used to determine whether to increase the preset acquisition length. It depends on the complexity of the target navigation area and is usually set between 0.8 and 1.0. In this embodiment, it is set to 0.9 to reasonably trigger acquisition length adjustment in high-fluctuation environments. The minimum value of the preset adjustment judgment range is used to determine whether to increase the lower limit of the preset relative slope threshold. It depends on the range of ground slope changes and is usually set between 0.2 and 0.5. In this embodiment, it is set to 0.3 to trigger appropriate slope threshold adjustment in low-fluctuation environments.

[0092] By comparing the adjustment judgment value with the upper and lower limits of the preset adjustment judgment range, the relative deviation is calculated, and the preset acquisition length or preset relative slope threshold is dynamically increased to achieve adaptive adjustment of navigation parameters. This method can match the actual performance of the vehicle under different ground reflection and driving conditions with the parameter settings of the navigation environment, enabling the vehicle to maintain stable operation in complex road conditions, optimize route selection, improve driving safety and navigation efficiency, and ensure that the navigation system can flexibly respond to environmental fluctuations.

[0093] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for navigating low-speed vehicles based on intelligent agents, characterized in that, include: Real-time data acquisition is obtained of the ground reflectivity and relative slope within a radius of a moving vehicle and a preset acquisition length within the parking area of ​​the industrial park warehouse; the proportion of shelf obstructions and the width of the drivable passage in each undetermined area within the parking area; and the deceleration frequency and emergency stop frequency of each mobile charging robot during its movement. Several first navigation areas are determined based on the relative slope, a preset relative slope threshold, and the undetermined area where the vehicle is located; Based on the ground reflectivity ratio, the shelf obstruction ratio in each of the first navigation areas, and the width of the drivable passage within the next preset first determination time, a time-series correlation determination is performed to determine several second navigation areas. Based on the deceleration frequency and emergency stop frequency of the mobile charging robot during its movement to each of the second navigation areas within the next preset second determination time, the path smoothness is determined to identify several temporary navigation areas. The target navigation region is selected from all the temporary navigation regions based on the preset visual language map model set in the navigation intelligent body and the navigation parameters of each temporary navigation region acquired in real time. The preset collection length or the preset relative slope threshold is adjusted based on the ground reflectivity ratio and the emergency stop frequency of the vehicle heading to the target navigation area within the next preset adjustment time. The output provides navigation prompts for the target navigation area after adjusting the preset acquisition length or the preset relative slope threshold.

2. The low-speed vehicle navigation method based on intelligent agents according to claim 1, characterized in that, The process of determining several first navigation areas based on the relative slope, a preset relative slope threshold, and the undetermined area where the vehicle is located includes: When the relative slope is less than the preset relative slope threshold, the undetermined area is determined to be the first candidate area; When the relative slope of all undetermined regions within the adjacent preset screening range of each first candidate region is less than the preset relative slope threshold, the first candidate region is determined to be the second candidate region. Several first navigation regions are determined based on the distribution characteristics of all second candidate regions.

3. The low-speed vehicle navigation method based on intelligent agents according to claim 2, characterized in that, The process of determining several first navigation regions based on the distribution characteristics of all second candidate regions includes: The Euclidean distance between any two adjacent second candidate regions is calculated based on the spatial coordinate information of each second candidate region to obtain several candidate distances; When the candidate distance is less than a preset aggregation distance threshold, the two corresponding second candidate regions are divided into the same aggregation unit to obtain several aggregation regions; Several first navigation regions are determined based on the continuous distribution characteristics of the aggregation regions.

4. The low-speed vehicle navigation method based on intelligent agents according to claim 3, characterized in that, The process of determining several first navigation regions based on the continuous distribution characteristics of the aggregation regions includes: Calculate the spatial connectivity determination value between each of the aggregated regions to obtain several degrees of continuity; When the continuity is greater than a preset continuity threshold, the corresponding aggregation region is determined to be the first navigation region, thereby determining a plurality of first navigation regions.

5. The low-speed vehicle navigation method based on intelligent agents according to claim 4, characterized in that, The process of determining several second navigation areas by performing a time-series correlation determination based on the ground reflectivity ratio, the shelf obstruction ratio in each of the first navigation areas, and the width of the drivable passageway within the next preset first determination time period includes: A correlation judgment index is determined based on the ground reflectivity ratio, the shelf obstruction ratio, and the width of the drivable passage within the preset first judgment time period. When the correlation determination index is greater than a preset determination index threshold, the first navigation region is determined to be the second navigation region, thereby identifying several second navigation regions.

6. The low-speed vehicle navigation method based on intelligent agents according to claim 5, characterized in that, The process of determining several temporary navigation areas by judging the path smoothness based on the deceleration frequency and the emergency stop frequency of the mobile charging robot during its movement to each of the second navigation areas within the next preset second determination time period includes: Calculate path smoothness based on all the deceleration frequencies and all the emergency stop frequencies; When the path smoothness is greater than a preset smoothness threshold, the second navigation area is determined to be the temporary navigation area.

7. The low-speed vehicle navigation method based on intelligent agents according to claim 6, characterized in that, The process of selecting the target navigation region from all the temporary navigation regions based on the preset visual language map model set in the navigation intelligence and the navigation parameters of each temporary navigation region acquired in real time includes: The vehicle's current speed, vehicle steering angle, and forward visibility distance from the navigation parameters are input into the preset visual language map model to obtain several first filtering areas; The straight-line distance between the vehicle and the target parking space in the temporary navigation area and the vehicle heading deviation in the navigation parameters are input into the preset visual language map model to obtain several second filtering areas; The intersection of the first and second filtering regions is determined to be the candidate region. The target navigation region is selected from all the temporary navigation regions based on all the navigation parameters in all candidate regions.

8. The low-speed vehicle navigation method based on intelligent agents according to claim 7, characterized in that, The process of selecting a target navigation region from all the temporary navigation regions based on all the navigation parameters from all the candidate regions includes: Calculate the comprehensive characterization value of each candidate region based on all the navigation parameters in each candidate region; The candidate region corresponding to the maximum value of all the comprehensive characterization values ​​is marked to filter out the target navigation region.

9. The low-speed vehicle navigation method based on intelligent agents according to claim 8, characterized in that, The process of adjusting the preset acquisition length or the preset relative slope threshold based on the ground reflectivity ratio and the emergency stop frequency of the vehicle traveling to the target navigation area within the next preset adjustment time includes: Calculate the standard deviation of the total ground reflectivity from the initial time of the preset adjustment period to each time to obtain several reflectivity fluctuation values; Calculate the standard deviation of all emergency stop frequencies from the initial time of the preset adjustment duration to each time to obtain several emergency stop frequency fluctuation values; Calculate the Pearson correlation coefficients of all the aforementioned reflectivity fluctuation values ​​and all the aforementioned emergency stop frequency fluctuation values ​​to obtain the adjustment judgment value; Adjust the preset acquisition length or the preset relative slope threshold according to the adjustment judgment value and the preset adjustment judgment range.

10. The low-speed vehicle navigation method based on intelligent agents according to claim 9, characterized in that, The process of adjusting the preset acquisition length or the preset relative slope threshold according to the adjustment judgment value and the preset adjustment judgment range includes: When the adjustment judgment value is greater than the maximum value of the preset adjustment judgment range, the preset acquisition length is increased according to the relative deviation between the adjustment judgment value and the maximum value of the preset adjustment judgment range; When the adjustment judgment value is less than the minimum value of the preset adjustment judgment range, the preset relative slope threshold is increased according to the relative deviation between the adjustment judgment value and the minimum value of the preset adjustment judgment range.

Images