Intelligent obstacle avoidance system for fruit case conveying based on omnidirectional mobile robot

Abstract

This invention relates to the field of intelligent conveying technology, specifically an intelligent fruit crate delivery and obstacle avoidance system based on an omnidirectional mobile robot. The system includes a spatiotemporal fusion mapping module, a rolling temporal path optimization module, a distributed cooperative path encoding module, a trajectory tracking and torque compensation control module, a visual servo fine-tuning placement module, and a task closed-loop self-checking log module. The system performs spatiotemporal fusion mapping on the operating environment information to obtain a dense temporal environment map; performs rolling temporal optimization on the motion state to obtain a continuous curvature smooth path; performs consistent encoding on the cooperative path information to obtain conflict-free path instructions; performs torque compensation adjustment on the intelligent fruit crate delivery path to obtain a stable crate delivery state; performs visual servo fine-tuning on the target robot to obtain a zero-impact crate placement result; and performs a state self-check on the zero-impact crate placement result to obtain a closed-loop task execution log. This invention can improve the efficiency of obstacle avoidance in intelligent fruit crate delivery.

Description

Technical Field

[0001] This invention relates to the field of intelligent conveying technology, and in particular to an intelligent conveying and obstacle avoidance system for fruit crates based on an omnidirectional mobile robot. Background Technology

[0002] In existing intelligent fruit distribution and conveying systems, the mapping of the working environment lacks a spatiotemporal joint processing mechanism. It relies solely on the fusion of single-frame perception data or static historical maps, which cannot effectively identify and eliminate non-persistent feature areas. This results in insufficient density and poor temporal continuity of the generated environmental map, providing limited accuracy for subsequent path planning. At the same time, traditional path planning does not incorporate the concept of rolling temporal optimization, and lacks forward-looking analysis of the robot's motion state. It is difficult to construct time-varying safety corridors that adapt to dynamic environments, and the curvature continuity of the path cannot be guaranteed.

[0003] In the collaborative delivery and execution of multi-robot systems, existing systems often employ centralized processing for conflict resolution. This lack of comprehensive integration and spatiotemporal gridding of collaborative path information leads to delayed conflict detection, unreasonable resource allocation, and a tendency to trigger path conflicts. Furthermore, insufficient decoupling and identification of geometric constraints and disturbance characteristics during trajectory tracking, coupled with a lack of feedforward adaptation and tuning for torque compensation, results in poor stability during box delivery and inadequate impact control during box placement, impacting delivery efficiency and fruit integrity. Therefore, improving the environmental mapping accuracy, path optimization smoothness, collaborative obstacle avoidance reliability, and delivery and placement stability of intelligent fruit box delivery has become an urgent problem to be solved. Summary of the Invention

[0004] To achieve the above objectives, the present invention provides an intelligent fruit crate delivery and obstacle avoidance system based on an omnidirectional mobile robot, characterized in that the system includes a spatiotemporal fusion mapping module, a rolling temporal path optimization module, a distributed cooperative path coding module, a trajectory tracking and torque compensation control module, a visual servo fine-tuning placement module, and a task closed-loop self-checking log module, wherein: The environment spatiotemporal fusion mapping module is used to perform spatiotemporal joint mapping of the target robot's operating environment information to obtain a dense temporal environment map of the target robot; The rolling temporal path optimization module is used to perform rolling temporal optimization on the motion state of the target robot based on a dense temporal environment map, so as to obtain a continuous curvature smooth path for the target robot. The distributed cooperative path encoding module is used to perform distributed conflict resolution on the cooperative path information of the target robot based on a continuous curvature smooth path, and to perform consistent encoding on the resolved information to obtain the conflict-free path instructions of the target robot. The trajectory tracking and torque compensation control module is used to perform trajectory tracking control on the intelligent fruit box conveying path of the target robot based on conflict-free path instructions, and to perform torque compensation adjustment on the intelligent fruit box conveying path to obtain a stable box conveying state of the target robot. The visual servo fine-tuning placement module is used to perform visual servo fine-tuning on the target robot based on the stable box delivery state, so as to obtain the zero-impact box placement result of the target robot. The task closed-loop self-check log module is used to perform a status self-check on the zero-impact box placement results in order to obtain the closed-loop task execution log of the target robot.

[0005] In a preferred embodiment, when the environmental spatiotemporal fusion mapping module performs spatiotemporal joint mapping of the target robot's operating environment information to obtain a dense temporal sequence environment map of the target robot, it is specifically used for: Acquire the target robot's operational environment perception data and historical dense environment map; Spatiotemporal encoding is performed on the operational environment perception data to obtain the current frame environmental features of the target robot; Map the current frame's environmental features to a historical dense environment map to identify the target robot's non-persistent feature regions; Based on non-persistent feature regions, invalid regions are dissolved in historical dense environment maps to obtain a corrected baseline map of the target robot. The modified baseline map is fused with the current frame's environmental features across frames to obtain the spatiotemporal fusion feature map of the target robot. Dense topology reconstruction is performed on the spatiotemporal fusion feature map to obtain a dense temporal environment map of the target robot.

[0006] In a preferred embodiment, when the rolling temporal path optimization module performs rolling temporal optimization on the motion state of the target robot based on a dense temporal environment map to obtain a continuous curvature smooth path for the target robot, it is specifically used for: A forward-looking analysis is performed on the current state of the target robot and the task objective to obtain the initial forward-looking path of the target; Based on the dense temporal environment map, the feasible region of the initial look-ahead path is extracted to obtain the time-varying safety corridor of the target robot. Based on the time-varying safety corridor, parametric interpolation fitting is performed on the initial look-ahead path of the target to obtain a smooth path description of the target robot. The curvature continuity of the smooth path description is enhanced to obtain a continuous curvature smooth path for the target robot.

[0007] In a preferred embodiment, when the rolling temporal path optimization module performs curvature continuity enhancement on the smooth path description to obtain a continuous curvature smooth path for the target robot, it is specifically used for: Mathematical parameter analysis is performed on the smooth path description to obtain the parameterized coordinate components of the smooth path description; Curvature derivation is performed on the parameterized coordinate components to obtain the curvature values ​​describing the smooth path. Based on curvature values, defects are identified in smooth path descriptions to obtain continuous defect information in the smooth path descriptions. Based on continuous defect information, the smooth path description is iteratively repaired to obtain the optimized path description of the target robot. The continuity of the optimized path description is verified to obtain a continuous curvature smooth path for the target robot.

[0008] In a preferred embodiment, when the distributed cooperative path encoding module performs distributed conflict resolution on the cooperative path information of the target robot based on a continuous curvature smooth path, and performs consistent encoding on the resolved information to obtain the conflict-free path instructions for the target robot, it is specifically used for: Real-time global fusion of continuous curvature smooth paths yields the cooperative path information of the target robot; Spatiotemporal rasterization mapping of the collaborative path information yields the spatiotemporal occupancy atlas of the target robot; Conflict detection is performed on the spatiotemporal occupancy atlas to obtain the spatiotemporal conflict region of the target robot; Consistent negotiation is conducted in the spatiotemporal conflict areas to obtain the spatiotemporal resource allocation agreement for the target robot; Based on the spatiotemporal resource allocation protocol, trajectory replanning is performed in spatiotemporal conflict areas to obtain a conflict-free coordinated path for the target robot. The conflict-free coordination path is encoded into an instruction sequence to obtain the conflict-free path instruction for the target robot.

[0009] In a preferred embodiment, when the trajectory tracking and torque compensation control module executes a conflict-free path instruction to perform trajectory tracking control on the intelligent fruit box conveying path of the target robot and to perform torque compensation adjustment on the intelligent fruit box conveying path to obtain a stable box conveying state of the target robot, it is specifically used for: By performing structured semantic parsing on conflict-free path instructions, the intelligent delivery path for fruit boxing of the target robot is obtained; The dual-flow decoupling identification of the intelligent conveying path for fruit boxes is performed to obtain the geometric constraint flow and disturbance characteristic flow of the intelligent conveying path for fruit boxes; High-fidelity trajectory reconstruction is performed on the geometrically constrained flow to obtain the anti-distortion reference trajectory flow of the target robot; Feedforward compensation is performed on the disturbance feature flow to obtain the torque compensation command flow of the target robot; By fusing the anti-distortion reference trajectory flow and the torque compensation command flow in the control domain, the stable box delivery state of the target robot is obtained.

[0010] In a preferred embodiment, when the trajectory tracking and torque compensation control module performs feedforward compensation on the disturbance feature flow to obtain the torque compensation command flow of the target robot, it is specifically used for: The perturbation feature flow is subjected to structured requirement deconstruction to obtain the compensation requirement template of the target robot; Based on the feedforward rule base of the target robot, the compensation requirement template is subjected to feedforward regularization and matching to obtain the original compensation action vector of the target robot. The original compensation motion vector is kinematically adapted and tuned to obtain the state adaptation compensation amount of the target robot. The state adaptation compensation amount and the anti-distortion reference trajectory flow are spatiotemporally coupled and calibrated to obtain the temporal alignment compensation sequence of the target robot; The timing alignment compensation sequence is mapped and encoded to obtain the torque compensation command stream of the target robot.

[0011] In a preferred embodiment, when the visual servo fine-tuning placement module performs visual servo fine-tuning on the target robot based on a stable box delivery state to obtain a zero-impact box placement result for the target robot, it is specifically used for: Visual pose deviation is extracted from the stable box conveying state to obtain the end-effector pose deviation description of the target robot; Based on the end-effector pose deviation description, deviation servo correction is performed on the target robot to obtain the initial fine-tuning motion of the target robot; Impact energy is estimated based on the preliminary fine-tuned motion, and the spatiotemporal distribution of impact energy of the target robot is obtained; Based on the spatiotemporal distribution of impact energy, the energy offset reconstruction is performed on the initially finely tuned motion to obtain the energy neutralization motion of the target robot. By integrating energy neutralization and motion quantities with contact transient compliant control, the zero-impact box placement result of the target robot is obtained.

[0012] In a preferred embodiment, when the visual servo fine-tuning placement module performs energy offset reconstruction on the initial fine-tuning motion quantity based on the spatiotemporal distribution of impact energy to obtain the energy-neutralized motion quantity, it is specifically used for: The spatiotemporal distribution of impact energy is decomposed into action modes to obtain the spatial impact mode set of the target robot; By performing a mirror inversion on the spatial impact mode set, the canceling energy mode set of the target robot is obtained; Based on the energy dissipation mode set, the preliminary fine-tuned motion quantities are kinematically injected and fused to obtain the composite motion quantities of the target robot; Energy convergence correction is applied to the composite motion to obtain the energy-neutralized motion of the target robot.

[0013] In a preferred embodiment, when the task closed-loop self-check log module performs a status self-check on the zero-impact bin placement results to obtain the closed-loop task execution log of the target robot, it is specifically used for: The zero-impact box placement results are measured in detail to obtain the quantitative state characteristics of the target robot; The quantified state characteristics are used to determine compliance, thus obtaining the state assessment result of the target robot. Root cause analysis is performed on non-compliant states in the status assessment results to obtain an anomaly diagnosis report for the target robot; By decoupling the deviation elements from the anomaly diagnosis report, the performance deviation positioning data of the target robot can be obtained. The status assessment results, anomaly diagnosis reports, and performance deviation location data are compiled into a traceable log to obtain the closed-loop task execution log of the target robot.

[0014] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention utilizes the spatiotemporal joint mapping technology of the environmental spatiotemporal fusion mapping module to generate a dense temporal sequence environmental map with high density and temporal continuity, providing accurate and comprehensive environmental data support for path planning. The rolling temporal domain path optimization module, through forward-looking analysis, time-varying safety corridor construction, and curvature continuity enhancement, outputs a continuous curvature smooth path, significantly improving the smoothness and stability of robot movement, reducing the probability of box swaying during transport, and providing a stable transport environment for fruits. The distributed collaborative path encoding module achieves conflict-free planning of collaborative paths through global fusion and distributed conflict resolution, improving path coordination efficiency during multi-robot operations and ensuring the orderly progress of the overall transport process.

[0015] 2. This invention employs a dual-flow decoupling identification and feedforward torque compensation mechanism in its trajectory tracking and torque compensation control module to achieve precise tracking and disturbance cancellation of the conveying path, significantly enhancing the stability of the box conveying and effectively preventing fruit displacement or damage caused by disturbances. The visual servo fine-tuning placement module achieves zero-impact box placement through pose deviation extraction, energy offset reconstruction, and contact transient compliance control, maximizing fruit integrity and improving the quality of box placement operations. The closed-loop task execution log generated by the task closed-loop self-check log module accurately captures task execution status and performance deviation data, providing reliable data for subsequent system maintenance and parameter optimization, enhancing the reliability and sustainability of system operation. Attached Figure Description

[0016] Figure 1 This is a system architecture diagram of an intelligent fruit box delivery and obstacle avoidance system based on an omnidirectional mobile robot, provided in an embodiment of the present invention.

[0017] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments belong to some, but not all, embodiments of the present invention. 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.

[0019] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “said” and “the” as used in the embodiments of this invention and the appended claims are also intended to include the plural forms, and “multiple” generally includes at least two unless the context clearly indicates otherwise.

[0020] Depending on the context, the word "if" or "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0021] Furthermore, the timing of the steps in the following method embodiments is merely an example and not a strict limitation.

[0022] In practice, the server-side equipment deployed in the intelligent fruit box delivery and obstacle avoidance system based on omnidirectional mobile robots may consist of one or more devices. This system can be implemented as a business instance, a virtual machine, or hardware devices. For example, it can be implemented as a business instance deployed on one or more devices in a cloud node. Simply put, it can be understood as software deployed on a cloud node to provide the intelligent fruit box delivery and obstacle avoidance system to various user terminals. Alternatively, it can be implemented as a virtual machine deployed on one or more devices in a cloud node, with application software installed to manage each user terminal. Or, it can also be implemented as a server-side system composed of numerous identical or different types of hardware devices, with one or more hardware devices configured to provide the intelligent fruit box delivery and obstacle avoidance system to various user terminals.

[0023] In terms of implementation, the intelligent fruit box delivery and obstacle avoidance system based on omnidirectional mobile robots and the user terminal are mutually compatible. That is, if the intelligent fruit box delivery and obstacle avoidance system based on omnidirectional mobile robots is implemented as an application installed on a cloud service platform, then the user terminal is a client that establishes a communication connection with the application; or if the intelligent fruit box delivery and obstacle avoidance system based on omnidirectional mobile robots is implemented as a website, then the user terminal is implemented as a webpage; or if the intelligent fruit box delivery and obstacle avoidance system based on omnidirectional mobile robots is implemented as a cloud service platform, then the user terminal is implemented as a mini-program in an instant messaging application.

[0024] like Figure 1 The diagram shown is a system architecture diagram of an intelligent fruit box delivery and obstacle avoidance system based on an omnidirectional mobile robot provided in an embodiment of the present invention.

[0025] The intelligent fruit box delivery and obstacle avoidance system 100 based on an omnidirectional mobile robot described in this invention can be installed on a cloud server. In terms of implementation, it can be used as one or more service devices, or as an application installed on the cloud (e.g., a mobile service operator's server, server cluster, etc.), or it can be developed into a website. Depending on the functions implemented, the intelligent fruit box delivery and obstacle avoidance system 100 based on an omnidirectional mobile robot may include an environmental spatiotemporal fusion mapping module 101, a rolling temporal path optimization module 102, a distributed collaborative path encoding module 103, a trajectory tracking and torque compensation control module 104, a visual servo fine-tuning placement module 105, and a task closed-loop self-checking log module 106. The module described in this invention can also be called a unit, which refers to a series of computer program segments that can be executed by an electronic device's processor and can perform a fixed function, stored in the electronic device's memory.

[0026] In this embodiment of the invention, in the intelligent fruit box conveying and obstacle avoidance system based on an omnidirectional mobile robot, each of the above modules can be implemented independently and can call other modules. Here, "calling" can be understood as one module connecting to multiple modules of another type and providing corresponding services to those connected modules. In the intelligent fruit box conveying and obstacle avoidance system based on an omnidirectional mobile robot provided by this embodiment of the invention, the applicable scope of the system architecture can be adjusted by adding modules and directly calling them without modifying the program code, achieving cluster-based horizontal expansion to quickly and flexibly expand the system. In practical applications, the above modules can be set in the same device or different devices, or they can be set in a virtual device, such as a service instance in a cloud server.

[0027] The following describes, with reference to specific embodiments, each component and its specific workflow of the intelligent fruit box conveying and obstacle avoidance system based on an omnidirectional mobile robot: The environment spatiotemporal fusion mapping module 101 is used to perform spatiotemporal joint mapping on the operating environment information of the target robot to obtain a dense temporal environment map of the target robot. In this embodiment of the invention, when the environmental spatiotemporal fusion mapping module performs spatiotemporal joint mapping of the target robot's operating environment information to obtain a dense temporal sequence environment map of the target robot, it is specifically used for: Acquire the target robot's operational environment perception data and historical dense environment map; Spatiotemporal encoding is performed on the operational environment perception data to obtain the current frame environmental features of the target robot; Map the current frame's environmental features to a historical dense environment map to identify the target robot's non-persistent feature regions; Based on non-persistent feature regions, invalid regions are dissolved in historical dense environment maps to obtain a corrected baseline map of the target robot. The modified baseline map is fused with the current frame's environmental features across frames to obtain the spatiotemporal fusion feature map of the target robot. Dense topology reconstruction is performed on the spatiotemporal fusion feature map to obtain a dense temporal environment map of the target robot.

[0028] The robot uses visual sensors, lidar, and other environmental perception devices to continuously collect real-time data on the outline shape, spatial position, distance dimensions, and obstacle distribution of objects within the work area. At the same time, it retrieves historical dense environment maps generated and saved during previous operations from the robot's storage medium, which contain detailed spatial structures and stability characteristics of the work environment. This ensures that both real-time work environment perception data and historical dense environment maps are obtained completely at the same time.

[0029] The time sequence of the collection of environmental perception data is clearly defined, and each set of collected environmental data is marked with a unique timestamp. Then, the environmental elements corresponding to these data are established with a precise correspondence with the spatial coordinate system of the work area. The scattered real-time environmental data is structured and integrated according to timestamps and spatial coordinates, and the originally discrete environmental information is transformed into a structured data set with clear time attributes and spatial location identifiers. Finally, the current frame environmental features that can accurately reflect the current state of the work environment are formed.

[0030] Based on the spatial coordinates of each environmental element in the current frame's environmental features, they are mapped one by one to the same spatial location in the historical dense environmental map. By comparing the differences between the environmental elements in the current frame's environmental features and the corresponding spatial locations in the historical dense environmental map, regions corresponding to environmental elements that exist in the historical dense environmental map but do not appear in the current frame's environmental features, as well as regions corresponding to environmental elements that appear in the current frame's environmental features but have no corresponding records in the historical dense environmental map and do not have long-term existence attributes, are selected. These selected regions are non-persistent feature regions.

[0031] In the historical dense environment map, the spatial range corresponding to all non-persistent feature areas is accurately located. By deleting all environmental feature information within this spatial range, invalid environmental data that no longer exists or temporarily appears in the historical map is completely removed. At the same time, long-term stable environmental features in the historical dense environment map, such as fixed shelves, passage boundaries, and permanent facilities, are completely preserved. After thoroughly cleaning up invalid information and retaining valid information, a corrected benchmark map that can accurately reflect the stable spatial structure of the working environment is formed.

[0032] Using the modified baseline map as the basic framework, all valid environmental elements in the current frame's environmental features, including newly emerging stable environmental features and unchanged original environmental features, are precisely superimposed onto the corresponding positions of the modified baseline map according to their corresponding timestamps and spatial coordinates. This achieves complementary integration of historical stable environmental information and current real-time environmental information, so that the integrated environmental data contains both stable structural information from historical maps and real-time environmental change information from the current frame, ultimately forming a spatiotemporal fusion feature map that simultaneously possesses temporal continuity and spatial integrity.

[0033] A comprehensive analysis of the spatial connections and adjacent location relationships among various environmental elements in the spatiotemporal fusion feature map is conducted. Based on the actual spatial layout of the work environment, scattered environmental feature points and feature regions are organically linked and integrated according to their true spatial location relationships. This process supplements the spatial relationship information between different environmental features, fills in any information gaps between features, and enables the map to fully present the spatial topology of the work environment. At the same time, it retains the timestamp information corresponding to each environmental feature, forming an environmental map that combines detailed spatial details with clear temporal dimension relationships, ultimately resulting in a dense temporal sequence environmental map.

[0034] The rolling temporal path optimization module 102 is used to perform rolling temporal optimization on the motion state of the target robot based on a dense temporal environment map, so as to obtain a continuous curvature smooth path for the target robot. In this embodiment of the invention, when the rolling temporal path optimization module performs rolling temporal optimization on the motion state of the target robot based on a dense temporal environment map to obtain a continuous curvature smooth path for the target robot, it is specifically used for: A forward-looking analysis is performed on the current state of the target robot and the task objective to obtain the initial forward-looking path of the target; Based on the dense temporal environment map, the feasible region of the initial look-ahead path is extracted to obtain the time-varying safety corridor of the target robot. Based on the time-varying safety corridor, parametric interpolation fitting is performed on the initial look-ahead path of the target to obtain a smooth path description of the target robot. The curvature continuity of the smooth path description is enhanced to obtain a continuous curvature smooth path for the target robot.

[0035] The rolling temporal path optimization module, when performing curvature continuity enhancement on the smooth path description to obtain a continuous curvature smooth path for the target robot, is specifically used for: Mathematical parameter analysis is performed on the smooth path description to obtain the parameterized coordinate components of the smooth path description; Curvature derivation is performed on the parameterized coordinate components to obtain the curvature values ​​describing the smooth path. Based on curvature values, defects are identified in smooth path descriptions to obtain continuous defect information in the smooth path descriptions. Based on continuous defect information, the smooth path description is iteratively repaired to obtain the optimized path description of the target robot. The continuity of the optimized path description is verified to obtain a continuous curvature smooth path for the target robot.

[0036] The robot acquires its real-time position, speed, and posture information through its onboard state detection equipment. At the same time, it clarifies the task objectives, such as the destination of the fruit box and the delivery time requirements. By correlating the current state information with the task objectives, it predicts the robot's future movement trend from its current position to the target position, plans a route that covers the key areas it needs to pass through and its general direction, and finally forms the initial forward-looking path to the target.

[0037] Based on the dense temporal environmental map generated by the environmental spatiotemporal fusion mapping module, environmental information such as fixed obstacles, dynamic interference factors and their real-time locations contained in the map is extracted. Taking the initial forward-looking path as the core, the passable space range around the path without collision risk is screened out. At the same time, the changing patterns of dynamic factors in the environment are fully considered, and a spatial area that can always meet the safe passage conditions over time is delineated. This area is the time-varying safety corridor of the target robot.

[0038] Using the boundary range and temporal variation law of the time-varying safety corridor as rigid constraints, several transition path points are added at uniform intervals between the key nodes of the initial forward path of the target. This makes the path connection between the key nodes and the transition path points, and between the transition path points themselves, natural and smooth, completely eliminating abrupt turns in the path, and forming a path outline that is completely within the time-varying safety corridor and has a coherent shape. This outline is the smooth path description of the target robot.

[0039] The overall path presented by the smooth path description is decomposed according to spatial dimensions. The specific spatial location of each path point on the path is sorted out one by one. The scattered path point location data is classified and organized according to spatial dimensions to form a data sequence in which the path point position changes sequentially with the travel order in each spatial dimension. These data sequences are the parameterized coordinate components of the smooth path description.

[0040] For the path point location data of each spatial dimension in the parametric coordinate components, analyze the spatial positional relationship between adjacent path points, determine the direction and degree of change of path direction, and convert the curvature state of each path segment into specific values ​​that can intuitively represent the degree of curvature. These values ​​are the curvature values ​​that describe the smooth path.

[0041] The formula for calculating the curvature value is as follows: ; in, Represents parameter points The curvature value at that point, The parameterized x-axis components represent the smooth path description. The parameterized ordinate components representing the smooth path description express The first derivative, express The first derivative, express The second derivative, express The second derivative; The parameterized x-axis component from the smooth path description is a data sequence of path point positions in the x-axis direction that change with the travel sequence after the smooth path description is decomposed according to spatial dimensions.

[0042] The parameterized ordinate component from the smooth path description is a data sequence of path point positions in the ordinate direction that change with the travel sequence after the smooth path description is decomposed according to spatial dimensions.

[0043] It is obtained by calculating the ratio of the change in the x-coordinate of adjacent path points in the parameterized x-coordinate component to the corresponding travel sequence interval, and then coherently organizing the calculation results of consecutive adjacent points. The first derivative.

[0044] It is obtained by calculating the ratio of the change in the ordinate of adjacent path points in the parameterized ordinate component to the corresponding travel sequence interval, and then coherently organizing the calculation results of consecutive adjacent points. The first derivative.

[0045] Through calculation The ratio of the change in adjacent derivative results to the corresponding traversal interval is obtained by coherently organizing consecutive adjacent results. The second derivative of .

[0046] Through calculation The ratio of the change in adjacent derivative results to the corresponding traversal interval is obtained by coherently organizing consecutive adjacent results. The second derivative of .

[0047] This formula is used to quantify the parameter points describing a smooth path. The curvature at a given point is determined by integrating the first and second derivatives of the parameterized abscissa and ordinate components. This results in a curvature value that accurately characterizes the curvature state of the path at that point, providing direct and accurate data support for subsequent identification of continuity defects in smooth paths based on curvature values.

[0048] As the absolute value of the cross product of the first and second derivatives of the parameterized abscissa and ordinate components increases, the curvature value also increases, indicating that the path becomes more curved at that parameter point.

[0049] As the cube root of the sum of the squares of the first derivatives of the parameterized abscissa and ordinate components increases, the curvature value decreases, indicating that the path is less curved at that parameter point.

[0050] The magnitude of curvature value directly reflects the difference in the path's curvature state. A larger curvature value indicates a more obvious path curvature and a greater likelihood of continuity defects. A more stable curvature value indicates a more consistent path curvature state, meeting the requirements of a continuous curvature smooth path.

[0051] The curvature values ​​of adjacent path segments in the smooth path description are compared one by one to observe the trend of curvature value changes. Path segments with sudden large increases or decreases in curvature value that do not meet the requirements of smooth transition are selected. The specific location, curvature change magnitude and other detailed information of these path segments are recorded and summarized to form the continuity defect information of the smooth path description.

[0052] For each curvature abrupt change path segment marked in the continuous defect information, the spatial coordinates of the path segment and the path points within a certain range around it are precisely adjusted so that the curvature value at the abrupt change position gradually transitions smoothly to the curvature value of the adjacent path segment. After the adjustment is completed, the curvature value of the area is recalculated and the defect is checked to see if it is eliminated. If curvature abrupt changes still exist, the adjustment process is repeated until all continuous defects are improved, and finally the optimized path description of the target robot is formed.

[0053] A comprehensive review of the curvature changes of all path segments in the optimized path description is conducted. Each path segment is checked to ensure a smooth transition in curvature, without any abrupt increases or decreases. This ensures that the overall curvature of the path is continuous and consistent, fully meeting the requirements for smooth robot movement. After verification and confirmation that the path meets the standards, a continuous curvature smooth path for the target robot is obtained.

[0054] The distributed cooperative path encoding module 103 is used to perform distributed conflict resolution on the cooperative path information of the target robot based on a continuous curvature smooth path, and to perform consistent encoding on the resolved information to obtain the conflict-free path instructions of the target robot. In this embodiment of the invention, when the distributed cooperative path encoding module performs distributed conflict resolution on the cooperative path information of the target robot based on a continuous curvature smooth path, and performs consistent encoding on the resolved information to obtain the conflict-free path instruction of the target robot, it is specifically used for: Real-time global fusion of continuous curvature smooth paths yields the cooperative path information of the target robot; Spatiotemporal rasterization mapping of the collaborative path information yields the spatiotemporal occupancy atlas of the target robot; Conflict detection is performed on the spatiotemporal occupancy atlas to obtain the spatiotemporal conflict region of the target robot; Consistent negotiation is conducted in the spatiotemporal conflict areas to obtain the spatiotemporal resource allocation agreement for the target robot; Based on the spatiotemporal resource allocation protocol, trajectory replanning is performed in spatiotemporal conflict areas to obtain a conflict-free coordinated path for the target robot. The conflict-free coordination path is encoded into an instruction sequence to obtain the conflict-free path instruction for the target robot.

[0055] The system collects continuous curvature smooth paths of all target robots participating in the intelligent delivery task of fruit boxes, including complete information such as the path direction, travel time nodes, and spatial locations traversed by each robot. At the same time, it summarizes the real-time operation status data of each robot. This scattered path information and status data are comprehensively integrated according to the time and spatial dimensions to eliminate information silos and form a holistic data set that can reflect the path correlation relationship when all robots work together, ultimately obtaining the collaborative path information of the target robots.

[0056] The work area is divided into several uniform spatial grids according to a fixed size. Each grid is assigned a unique spatial identifier. Then, combined with the travel time nodes of each robot in the collaborative path information, a corresponding time identifier is assigned to each time node. The spatial position of each robot in the collaborative path information at different time nodes is accurately mapped to the corresponding spatial grid. The corresponding robot identifier and time identifier are marked on each grid, forming an image set containing the grid occupancy of all robots in different time and space. This set is the spatiotemporal occupancy atlas of the target robot.

[0057] The occupancy records of each spatial grid in the spatiotemporal occupancy atlas are checked one by one at each time node. The occupancy of different robots in the same spatial grid at the same time node is compared. If the occupancy records of multiple robots overlap in the same spatiotemporal grid, it is determined that the area where the spatiotemporal grid is located is an area with path conflict. All spatiotemporal grids with such overlap are summarized, their spatial range and corresponding time interval are determined, and finally the spatiotemporal conflict area of ​​the target robot is obtained.

[0058] For each identified spatiotemporal conflict zone, all target robots involved in the conflict in that zone are organized to exchange information. Each robot synchronously reports its own task priority, path planning constraints, adjustable time windows, and other key information. Through information exchange and demand coordination among multiple robots, the usage rules for each spatiotemporal conflict zone are determined, and the passage order, time allocation, and spatial avoidance methods of different robots in the zone are clarified. Finally, a unified spatiotemporal resource allocation agreement for target robots is formed.

[0059] Based on the spatiotemporal resource allocation protocol, for each spatiotemporal conflict area, the original paths of the robots involved in the conflict are adjusted. Without violating the passage rules stipulated in the protocol, the robot's travel route in and around the conflict area is replanned, and the time nodes and spatial orientation of the path are optimized to ensure that the robot can strictly follow the spatiotemporal resources allocated by the protocol when passing through the area, without causing new conflicts with other robots, and finally obtains the conflict-free coordinated path of the target robot.

[0060] The conflict-free coordination path is decomposed in a structured manner, and key information such as the key spatial position, direction of travel, speed adjustment nodes, and time nodes in the path is extracted. This information is transformed into standardized instructions that the robot control system can recognize and execute. These instructions are arranged in an orderly manner according to the robot's movement sequence to form a coherent instruction sequence, and finally the conflict-free path instructions for the target robot are obtained.

[0061] The trajectory tracking and torque compensation control module 104 is used to perform trajectory tracking control on the intelligent fruit box conveying path of the target robot based on conflict-free path instructions, and to perform torque compensation adjustment on the intelligent fruit box conveying path to obtain a stable box conveying state of the target robot. In this embodiment of the invention, when the trajectory tracking and torque compensation control module executes a conflict-free path instruction to perform trajectory tracking control on the intelligent fruit box conveying path of the target robot and to perform torque compensation adjustment on the intelligent fruit box conveying path to obtain a stable box conveying state of the target robot, it is specifically used for: By performing structured semantic parsing on conflict-free path instructions, the intelligent delivery path for fruit boxing of the target robot is obtained; The dual-flow decoupling identification of the intelligent conveying path for fruit boxes is performed to obtain the geometric constraint flow and disturbance characteristic flow of the intelligent conveying path for fruit boxes; High-fidelity trajectory reconstruction is performed on the geometrically constrained flow to obtain the anti-distortion reference trajectory flow of the target robot; Feedforward compensation is performed on the disturbance feature flow to obtain the torque compensation command flow of the target robot; By fusing the anti-distortion reference trajectory flow and the torque compensation command flow in the control domain, the stable box delivery state of the target robot is obtained.

[0062] When the trajectory tracking and torque compensation control module performs feedforward compensation on the disturbance feature flow to obtain the torque compensation command flow for the target robot, it is specifically used for: The perturbation feature flow is subjected to structured requirement deconstruction to obtain the compensation requirement template of the target robot; Based on the feedforward rule base of the target robot, the compensation requirement template is subjected to feedforward regularization and matching to obtain the original compensation action vector of the target robot. The original compensation motion vector is kinematically adapted and tuned to obtain the state adaptation compensation amount of the target robot. The state adaptation compensation amount and the anti-distortion reference trajectory flow are spatiotemporally coupled and calibrated to obtain the temporal alignment compensation sequence of the target robot; The timing alignment compensation sequence is mapped and encoded to obtain the torque compensation command stream of the target robot.

[0063] The instruction sequence in the conflict-free path instruction is semantically split, and key information such as spatial location coordinates, direction of travel, speed requirements, and stopping points corresponding to each instruction is extracted. Then, these scattered key information are integrated according to time sequence and spatial logic to form a complete, continuous and clear intelligent delivery path for fruit boxes.

[0064] The intelligent conveying path for fruit boxes is feature-decomposed, extracting the inherent spatial structural constraints of the path itself, such as the curvature angle, passage width limit, and necessary node position constraints, to form a geometric constraint flow. At the same time, the feature information corresponding to various interference factors that may be encountered during the path execution process, such as changes in ground friction, uneven weight distribution of boxes, and minor external impacts, is separated to form a disturbance feature flow, thus achieving a thorough separation and identification of two different attribute features.

[0065] Based on geometrically constrained flow, the key spatial constraint information is fully preserved. Potential trajectory distortion risks, such as path node offset and unnatural bending transition, are corrected. By supplementing key transition points and adjusting the connection method of path segments, a trajectory that is highly consistent with the original path and without distortion is constructed, thus obtaining an anti-distortion reference trajectory flow.

[0066] The various disturbance information in the disturbance feature stream is decomposed according to dimensions such as type, scope of influence, intensity of action, and duration. The specific impact of each disturbance on the robot's delivery state and the problems that need to be solved through torque compensation are clarified, forming a compensation requirement template that includes information such as compensation target, compensation range, and compensation priority.

[0067] The feedforward rule base is a set of rules pre-built and stored based on a large amount of actual operation data and disturbance handling experience. It includes association rules between various disturbance characteristics and corresponding compensation actions, compensation intensity standards, action execution timing specifications, etc. Each rule in the rule base clarifies the type, execution magnitude, and time node of the compensation action to be taken under a specific disturbance scenario. It can directly provide a matching basis for the compensation requirement template. For each compensation requirement in the compensation requirement template, the rule base is used to find the completely corresponding disturbance handling rule, and the specific compensation action content is determined according to the rule to form the original compensation action vector.

[0068] The robot's current motion state, such as real-time speed, motion posture, and current position, is obtained. Based on this real-time state information, the motion amplitude, execution speed, and force application timing of the original compensation motion vector are adjusted to ensure that the compensation motion is fully adapted to the robot's current motion state. This avoids conflicts between the compensation motion and the robot's own motion, ensuring the effectiveness and safety of the compensation motion, and yielding the state adaptation compensation amount.

[0069] Based on the time axis and spatial coordinates of the anti-distortion reference trajectory flow, the execution time of the state adaptation compensation is adjusted so that the execution time of the compensation action is completely synchronized with the execution time of the corresponding position in the trajectory flow. At the same time, the spatial range of the compensation action is calibrated to ensure that the compensation action can be accurately applied to the position in the trajectory flow that needs compensation, thus forming a time-aligned compensation sequence.

[0070] Each compensation action in the timing alignment compensation sequence is converted into an instruction format that the robot control system can recognize. The amplitude, direction, execution time and other information of the compensation action are converted into standardized instruction codes and arranged in chronological order to form a continuous instruction stream, thus obtaining the torque compensation instruction stream.

[0071] By integrating the trajectory control signal corresponding to the anti-distortion reference trajectory flow and the torque control signal in the torque compensation command flow within the same control domain, the trajectory tracking control and torque compensation adjustment work together in a coordinated manner. The trajectory control signal ensures that the robot strictly follows the anti-distortion reference trajectory, while the torque compensation command flow cancels out the effects of various disturbances in real time. The combined effect of the two ensures that the robot remains stable during the transport process, with no shaking or deviation of the box, resulting in a stable box transport state.

[0072] The visual servo fine-tuning placement module 105 is used to perform visual servo fine-tuning on the target robot based on the stable box delivery state, so as to obtain the zero-impact box placement result of the target robot. In this embodiment of the invention, when the visual servo fine-tuning placement module performs visual servo fine-tuning on the target robot based on a stable box delivery state to obtain a zero-impact box placement result for the target robot, it is specifically used for: Visual pose deviation is extracted from the stable box conveying state to obtain the end-effector pose deviation description of the target robot; Based on the end-effector pose deviation description, deviation servo correction is performed on the target robot to obtain the initial fine-tuning motion of the target robot; Impact energy is estimated based on the preliminary fine-tuned motion, and the spatiotemporal distribution of impact energy of the target robot is obtained; Based on the spatiotemporal distribution of impact energy, the energy offset reconstruction is performed on the initially finely tuned motion to obtain the energy neutralization motion of the target robot. By integrating energy neutralization and motion quantities with contact transient compliant control, the zero-impact box placement result of the target robot is obtained.

[0073] When the visual servo fine-tuning placement module performs energy offset reconstruction on the initial fine-tuned motion quantity based on the spatiotemporal distribution of impact energy to obtain the energy-neutralized motion quantity, it is specifically used for: The spatiotemporal distribution of impact energy is decomposed into action modes to obtain the spatial impact mode set of the target robot; By performing a mirror inversion on the spatial impact mode set, the canceling energy mode set of the target robot is obtained; Based on the energy dissipation mode set, the preliminary fine-tuned motion quantities are kinematically injected and fused to obtain the composite motion quantities of the target robot; Energy convergence correction is applied to the composite motion to obtain the energy-neutralized motion of the target robot.

[0074] The robot uses a vision acquisition device to capture the actual spatial position and posture angle of the container during stable container transport. This acquired visual information is compared point by point with the preset standard placement posture. The offset in spatial coordinates and the difference in posture angle are accurately calculated. These deviation data are recorded in detail to form an end-effector posture deviation description that can comprehensively reflect the difference between the actual posture of the container and the standard posture.

[0075] Based on the spatial position offset and attitude angle difference specified in the end-effector pose deviation description, the specific direction and corresponding adjustment range that the robot needs to adjust are determined. By controlling the robot's actuator to make targeted motion adjustments in the determined direction and range, the pose deviation of the box is gradually offset, forming a preliminary fine-tuning motion that enables the box to gradually approach the standard placement pose.

[0076] By combining the range of motion, execution speed, and expected contact between the box and the placement surface during the initial fine-tuning of the exercise intensity, and taking into account the weight and material properties of the box, as well as the material and hardness of the placement surface, we analyze the magnitude of the impact energy that may be generated when the box is placed during the initial fine-tuning of the exercise intensity. We identify the release point of the impact energy in the time dimension and the area of ​​action in the spatial dimension, and obtain the spatiotemporal distribution of the impact energy.

[0077] The impact energy at different time points and in different spatial regions in the spatiotemporal distribution of impact energy is classified and decomposed according to its mode of action and intensity of influence. The direction of action, coverage and duration of each part of the impact energy are clarified. These decomposed impact energy characteristics are then classified and integrated to form a spatial impact mode set.

[0078] For each impact mode in the spatial impact mode set, based on the magnitude, direction, and action characteristics of its impact energy, a canceling energy mode with the same energy value but opposite action direction is constructed. All constructed canceling energy modes are systematically integrated to form a canceling energy mode set.

[0079] The motion instructions corresponding to each energy cancellation mode in the energy cancellation mode set are incorporated into the preliminary fine-tuning motion quantity according to kinematic principles. This allows the preliminary fine-tuning motion quantity to not only correct the box's posture deviation during execution, but also to offset the impact energy that may be generated during placement through the reverse energy generated by the energy cancellation mode, thus forming a composite motion quantity.

[0080] The energy balance of the composite exercise is checked to detect the effect of the cancellation of impact energy and neutralizing energy during the execution of the composite exercise. Based on the check results, the execution rhythm and force of the composite exercise are adjusted to ensure that the impact energy generated during the execution of the composite exercise is completely canceled, and finally the energy-neutralized exercise is obtained.

[0081] During the execution of energy neutralization motion, when the instant the box is about to contact the placement surface is detected, the magnitude of the force and the speed of the robot actuator are dynamically adjusted to make the contact process between the box and the placement surface smooth and avoid impact caused by excessive force at the moment of contact. By integrating energy neutralization motion with this contact transient smooth control method, a zero-impact box placement result is obtained.

[0082] The task closed-loop self-check log module 106 is used to perform a status self-check on the zero-impact box placement result in order to obtain the closed-loop task execution log of the target robot.

[0083] In this embodiment of the invention, when the task closed-loop self-check log module performs a status self-check on the zero-impact bin placement result to obtain the closed-loop task execution log of the target robot, it is specifically used for: The zero-impact box placement results are measured in detail to obtain the quantitative state characteristics of the target robot; The quantified state characteristics are used to determine compliance, thus obtaining the state assessment result of the target robot. Root cause analysis is performed on non-compliant states in the status assessment results to obtain an anomaly diagnosis report for the target robot; By decoupling the deviation elements from the anomaly diagnosis report, the performance deviation positioning data of the target robot can be obtained. The status assessment results, anomaly diagnosis reports, and performance deviation location data are compiled into a traceable log to obtain the closed-loop task execution log of the target robot.

[0084] By using high-precision position detection equipment, visual imaging equipment, and fruit condition detection equipment, the results of zero-impact box placement are comprehensively inspected. The system accurately measures various indicators such as the offset distance between the final placement position of the box and the standard position, the posture angle of the box after placement, whether there is damage on the surface of the box, and the integrity and arrangement of the fruit inside the box. These detected physical state information are converted into specific numerical data to form quantitative state characteristics that can objectively and comprehensively characterize the quality of box placement.

[0085] Pre-set compliance standards for box placement, including the allowable offset range of box placement position, the qualified range of posture angle, and the integrity standards of the box and fruit. Compare each numerical data in the quantitative status characteristics with the corresponding compliance standards one by one to determine whether each indicator meets the preset requirements. Clearly mark the qualified and unqualified indicator items, and summarize all comparison results to form the status assessment result of the target robot.

[0086] For each indicator that fails to meet the status assessment results, the entire process of box delivery corresponding to that indicator is traced back, from environmental mapping, path planning, collaborative obstacle avoidance, trajectory tracking, torque compensation to visual fine-tuning and placement, and each step is investigated to identify the factors that may cause the indicator to fail to meet the standard. The operation of each step is analyzed to see if it conforms to the preset process and whether there are any deviations in the output of related modules. The root cause of the indicator failure is determined, and the root cause, the impact process, and the specific manifestations of each non-compliant indicator are compiled and summarized to obtain the abnormality diagnosis report of the target robot.

[0087] The root cause of each abnormality recorded in the abnormality diagnosis report is broken down, and the individual elements that cause the performance deviation are separated. The nature, conditions of occurrence, degree of influence on the non-compliant indicators and the path of action of each element are clarified. The specific operation link, module function or operation step is accurately located. The information of these decomposed elements is classified and organized to form the performance deviation positioning data of the target robot.

[0088] Based on the execution time sequence and logical association of the packing task, the status assessment results, anomaly diagnosis reports, and performance deviation location data are systematically integrated. This includes supplementing information such as key time nodes during task execution, operating status records of each module, and changes in equipment parameters. This ensures that the integrated log can completely reconstruct the entire task execution process. At the same time, each data item is clearly marked with a traceability identifier to facilitate subsequent querying and verification, ultimately forming a closed-loop task execution log for the target robot.

[0089] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0090] This application embodiment can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0091] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A smart fruit crate delivery and obstacle avoidance system based on an omnidirectional mobile robot, characterized in that, The system includes an environmental spatiotemporal fusion mapping module, a rolling temporal path optimization module, a distributed collaborative path encoding module, a trajectory tracking and torque compensation control module, a visual servo fine-tuning placement module, and a task closed-loop self-checking log module, wherein: The environment spatiotemporal fusion mapping module is used to perform spatiotemporal joint mapping of the target robot's operating environment information to obtain a dense temporal environment map of the target robot; The rolling temporal path optimization module is used to perform rolling temporal optimization on the motion state of the target robot based on a dense temporal environment map, so as to obtain a continuous curvature smooth path for the target robot. The distributed cooperative path encoding module is used to perform distributed conflict resolution on the cooperative path information of the target robot based on a continuous curvature smooth path, and to perform consistent encoding on the resolved information to obtain the conflict-free path instructions of the target robot. The trajectory tracking and torque compensation control module is used to perform trajectory tracking control on the intelligent fruit box conveying path of the target robot based on conflict-free path instructions, and to perform torque compensation adjustment on the intelligent fruit box conveying path to obtain a stable box conveying state of the target robot. The visual servo fine-tuning placement module is used to perform visual servo fine-tuning on the target robot based on the stable box delivery state, so as to obtain the zero-impact box placement result of the target robot. The task closed-loop self-check log module is used to perform a status self-check on the zero-impact box placement results in order to obtain the closed-loop task execution log of the target robot.

2. The intelligent fruit box conveying and obstacle avoidance system based on an omnidirectional mobile robot as described in claim 1, characterized in that, The spatiotemporal fusion mapping module, when performing spatiotemporal joint mapping of the target robot's operating environment information to obtain a dense temporal sequence environment map of the target robot, is specifically used for: Acquire the target robot's operational environment perception data and historical dense environment map; Spatiotemporal encoding is performed on the operational environment perception data to obtain the current frame environmental features of the target robot; Map the current frame's environmental features to a historical dense environment map to identify the target robot's non-persistent feature regions; Based on non-persistent feature regions, invalid regions are dissolved in historical dense environment maps to obtain a corrected baseline map of the target robot. The modified baseline map is fused with the current frame's environmental features across frames to obtain the spatiotemporal fusion feature map of the target robot. Dense topology reconstruction is performed on the spatiotemporal fusion feature map to obtain a dense temporal environment map of the target robot.

3. The intelligent fruit box conveying and obstacle avoidance system based on an omnidirectional mobile robot as described in claim 1, characterized in that, The rolling temporal path optimization module, when performing rolling temporal optimization on the motion state of the target robot based on a dense temporal environment map to obtain a continuous curvature smooth path for the target robot, is specifically used for: A forward-looking analysis is performed on the current state of the target robot and the task objective to obtain the initial forward-looking path of the target; Based on the dense temporal environment map, the feasible region of the initial look-ahead path is extracted to obtain the time-varying safety corridor of the target robot. Based on the time-varying safety corridor, parametric interpolation fitting is performed on the initial look-ahead path of the target to obtain a smooth path description of the target robot. The curvature continuity of the smooth path description is enhanced to obtain a continuous curvature smooth path for the target robot.

4. The intelligent fruit box conveying and obstacle avoidance system based on an omnidirectional mobile robot as described in claim 3, characterized in that, The rolling temporal path optimization module, when performing curvature continuity enhancement on the smooth path description to obtain a continuous curvature smooth path for the target robot, is specifically used for: Mathematical parameter analysis is performed on the smooth path description to obtain the parameterized coordinate components of the smooth path description; Curvature derivation is performed on the parameterized coordinate components to obtain the curvature values ​​describing the smooth path. Based on curvature values, defects are identified in smooth path descriptions to obtain continuous defect information in the smooth path descriptions. Based on continuous defect information, the smooth path description is iteratively repaired to obtain the optimized path description of the target robot. The continuity of the optimized path description is verified to obtain the continuous curvature smooth path of the target robot.

5. The intelligent fruit box conveying and obstacle avoidance system based on an omnidirectional mobile robot as described in claim 1, characterized in that, The distributed cooperative path encoding module, when executing a continuous curvature smooth path to perform distributed conflict resolution on the cooperative path information of the target robot and then performing consistent encoding on the resolved information to obtain the conflict-free path instructions for the target robot, is specifically used for: Real-time global fusion of continuous curvature smooth paths yields the cooperative path information of the target robot; Spatiotemporal rasterization mapping of the collaborative path information yields the spatiotemporal occupancy atlas of the target robot; Conflict detection is performed on the spatiotemporal occupancy atlas to obtain the spatiotemporal conflict region of the target robot; Consistent negotiation is conducted in the spatiotemporal conflict areas to obtain the spatiotemporal resource allocation agreement for the target robot; Based on the spatiotemporal resource allocation protocol, trajectory replanning is performed in spatiotemporal conflict areas to obtain a conflict-free coordinated path for the target robot. The conflict-free coordination path is encoded into an instruction sequence to obtain the conflict-free path instruction for the target robot.

6. The intelligent fruit box conveying and obstacle avoidance system based on an omnidirectional mobile robot as described in claim 1, characterized in that, The trajectory tracking and torque compensation control module, when executing conflict-free path instructions to perform trajectory tracking control on the intelligent fruit box conveying path of the target robot and to adjust the torque compensation of the intelligent fruit box conveying path to obtain a stable box conveying state for the target robot, is specifically used for: By performing structured semantic parsing on conflict-free path instructions, the intelligent delivery path for fruit boxing of the target robot is obtained; The dual-flow decoupling identification of the intelligent conveying path for fruit boxes is performed to obtain the geometric constraint flow and disturbance characteristic flow of the intelligent conveying path for fruit boxes; High-fidelity trajectory reconstruction is performed on the geometrically constrained flow to obtain the anti-distortion reference trajectory flow of the target robot; Feedforward compensation is performed on the disturbance feature flow to obtain the torque compensation command flow of the target robot; By fusing the anti-distortion reference trajectory flow and the torque compensation command flow in the control domain, the stable box delivery state of the target robot is obtained.

7. The intelligent fruit box conveying and obstacle avoidance system based on an omnidirectional mobile robot as described in claim 6, characterized in that, When the trajectory tracking and torque compensation control module performs feedforward compensation on the disturbance feature flow to obtain the torque compensation command flow for the target robot, it is specifically used for: The perturbation feature flow is subjected to structured requirement deconstruction to obtain the compensation requirement template of the target robot; Based on the feedforward rule base of the target robot, the compensation requirement template is subjected to feedforward regularization and matching to obtain the original compensation action vector of the target robot. The original compensation motion vector is kinematically adapted and tuned to obtain the state adaptation compensation amount of the target robot. The state adaptation compensation amount and the anti-distortion reference trajectory flow are spatiotemporally coupled and calibrated to obtain the temporal alignment compensation sequence of the target robot; The timing alignment compensation sequence is mapped and encoded to obtain the torque compensation command stream of the target robot.

8. The intelligent fruit crate delivery and obstacle avoidance system based on an omnidirectional mobile robot as described in claim 1, characterized in that, When the visual servo fine-tuning placement module performs visual servo fine-tuning on the target robot based on a stable box delivery state to obtain a zero-impact box placement result for the target robot, it is specifically used for: Visual pose deviation is extracted from the stable box conveying state to obtain the end-effector pose deviation description of the target robot; Based on the end-effector pose deviation description, deviation servo correction is performed on the target robot to obtain the initial fine-tuning motion of the target robot; Impact energy is estimated based on the preliminary fine-tuned motion, and the spatiotemporal distribution of impact energy of the target robot is obtained; Based on the spatiotemporal distribution of impact energy, the energy offset reconstruction is performed on the initially finely tuned motion to obtain the energy neutralization motion of the target robot. By integrating energy neutralization and motion quantities with contact transient compliant control, the zero-impact box placement result of the target robot is obtained.

9. The intelligent fruit box conveying and obstacle avoidance system based on an omnidirectional mobile robot as described in claim 8, characterized in that, When the visual servo fine-tuning placement module performs energy offset reconstruction on the initial fine-tuned motion quantity based on the spatiotemporal distribution of impact energy to obtain the energy-neutralized motion quantity, it is specifically used for: The spatiotemporal distribution of impact energy is decomposed into action modes to obtain the spatial impact mode set of the target robot; By performing a mirror inversion on the spatial impact mode set, the canceling energy mode set of the target robot is obtained; Based on the energy dissipation mode set, the preliminary fine-tuned motion quantities are kinematically injected and fused to obtain the composite motion quantities of the target robot; Energy convergence correction is applied to the composite motion to obtain the energy-neutralized motion of the target robot.

10. The intelligent fruit box conveying and obstacle avoidance system based on an omnidirectional mobile robot as described in claim 1, characterized in that, The task closed-loop self-check log module, when performing a status self-check on the zero-impact bin placement results to obtain the closed-loop task execution log of the target robot, is specifically used for: The zero-impact box placement results are measured in detail to obtain the quantitative state characteristics of the target robot; The quantified state characteristics are used to determine compliance, thus obtaining the state assessment result of the target robot. Root cause analysis is performed on non-compliant states in the status assessment results to obtain an anomaly diagnosis report for the target robot; By decoupling the deviation elements from the anomaly diagnosis report, the performance deviation positioning data of the target robot can be obtained. The status assessment results, anomaly diagnosis reports, and performance deviation location data are compiled into a traceable log to obtain the closed-loop task execution log of the target robot.

Images