Robot environment perception data processing method and device, readable storage medium and intelligent cleaning equipment

Abstract

The application relates to an environment perception data processing method and device of a robot, and relates to the technical field of robots, and can improve the adaptive operation efficiency of the robot in a complex environment. The method comprises the following steps: acquiring environment perception data collected by the robot when performing a historical task, and a perception type corresponding to the environment perception data; encapsulating the environment perception data into a structured data object of a preset data structure, and acquiring a historical environment perception file according to the structured data object; when a new task is performed, a target historical environment perception file corresponding to a perception type of the new task is acquired, so that the robot can be controlled to process the new task according to the target historical environment perception file.

Description

Technical Field

[0001] This application relates to the field of robotics technology, and in particular to a method, apparatus, readable storage medium, and intelligent cleaning equipment for processing environmental perception data of a robot. Background Technology

[0002] With the development of robotics technology, cleaning robots are widely used in various complex home and commercial environments. In actual operation, in order to adapt to changing spatial terrain and achieve ideal cleaning results, cleaning robots usually need to perform precise path navigation and dynamically formulate detailed cleaning strategies.

[0003] However, in traditional control and processing methods, due to the complexity of cleaning scenarios, robots need to continuously utilize underlying multimodal sensors for real-time environmental detection and topology calculation, which significantly consumes the system's perception capabilities and core computing resources. With persistently high computing loads, when cleaning robots encounter unexpected situations (such as sudden encounters with unknown obstacles or edge steps), the system often needs to start from scratch, blindly exploring and re-mapping the site, resulting in severe delays in data processing and decision-making. This memoryless processing paradigm, which relies entirely on real-time computing power, makes it impossible for cleaning robots to respond promptly to unexpected situations, easily leading to serious safety accidents such as equipment collisions, getting stuck, or falls. Summary of the Invention

[0004] Therefore, it is necessary to provide a method, apparatus, computer equipment, computer-readable storage medium, and computer program product for processing environmental perception data of a robot to address the above-mentioned technical problems.

[0005] In a first aspect, this application provides a method for processing environmental perception data of a robot, including:

[0006] Acquire environmental perception data collected by the robot while performing historical tasks, as well as the perception type corresponding to the environmental perception data;

[0007] The environmental perception data is encapsulated into a structured data object with a preset data structure, and historical environmental perception files are obtained based on the structured data object;

[0008] When executing a new task, new map information for the new task and target historical environment perception file of the perception type corresponding to the new task are obtained, and integrated environment information corresponding to the new task is determined based on the new map information and the target historical environment perception file.

[0009] In one embodiment, encapsulating the environmental perception data into a structured data object with a preset data structure includes:

[0010] Obtain field configuration information for characterizing the environmental perception data features of the perception type;

[0011] If, based on the current data fields and current data types contained in the environmental perception data, it is determined that the environmental perception data matches the field configuration information, then the environmental perception data is encapsulated into a structured data object with a preset data structure.

[0012] If, based on the current data fields and current data types contained in the environmental perception data, it is determined that the environmental perception data does not match the field configuration information, then the encapsulation operation of the environmental perception data is terminated.

[0013] In one embodiment, after obtaining the field configuration information for characterizing the environmental perception data features of the perception type, the method further includes:

[0014] Obtain the specified field and expected data type from the field configuration information, and determine the current data field and current data type in the environment-aware data;

[0015] If any of the specified fields is missing from the current data field, or if any of the current data types does not match the expected data type, then it is determined that the environment-aware data does not match the field configuration information.

[0016] In one embodiment, encapsulating the environmental perception data into a structured data object with a preset data structure includes:

[0017] Obtain the global identifier corresponding to the environmental perception data, and obtain the current field value corresponding to each current data field in the environmental perception data;

[0018] Based on the perception type, the global identifier, and each of the current data fields and their current field values, a structured data object with the preset data structure is encapsulated.

[0019] In one embodiment, the process of encapsulating the structured data object based on the perception type, the global identifier, and each of the current data fields and their current field values ​​to obtain the preset data structure includes:

[0020] Get the data type of the current field value;

[0021] If the data type of the current field value belongs to a preset data type set, then the unified data storage structure is encapsulated based on the perception type, the global identifier, and each current data field and its current field value.

[0022] In one embodiment, obtaining the target historical environment perception file corresponding to the perception type of the new task includes:

[0023] Based on the new map information and the perception type of the new task, a search is performed in multiple saved historical environment perception files to obtain a candidate historical file set.

[0024] The target historical environment awareness file is determined based on the modification time of each candidate historical file in the candidate historical file set.

[0025] In one embodiment, the new task includes a cleaning task. After determining the integrated environmental information corresponding to the new task based on the new map information and the target historical environmental awareness file, the method further includes:

[0026] When performing a cleaning task, the target historical environment perception file is deserialized to obtain a structured data object of the perception type related to the cleaning task.

[0027] Based on the environmental perception data in the structured data object and the new map information, the robot is controlled to perform the cleaning task.

[0028] In one embodiment, obtaining the historical environment-aware file based on the structured data object includes:

[0029] By calling the unified data management interface shared by different perception types, the structured data object is serialized to obtain a structured text stream;

[0030] Obtain the map information and generation time of the historical task, and generate a filename based on the splicing result of the perception type, the map information, and the generation time;

[0031] The structured text stream is written into an entity file named with the filename to obtain the historical environment-aware file.

[0032] In one embodiment, the method further includes:

[0033] Obtain the number of existing files in the storage directory corresponding to the entity file, which corresponds to the perception type;

[0034] If the number of existing files is greater than or equal to a preset threshold, then the generation time of each existing file is obtained;

[0035] Based on the order of their generation times, determine the target cleanup file that is sorted first;

[0036] Delete the target cleanup file from the storage directory and save the entity file in the storage directory.

[0037] Secondly, this application also provides a robot environmental perception data processing device, comprising:

[0038] The perception data acquisition module is used to acquire environmental perception data collected by the robot when performing historical tasks, as well as the perception type corresponding to the environmental perception data.

[0039] The data structure encapsulation module is used to encapsulate the environmental perception data into a structured data object with a preset data structure, and to obtain historical environmental perception files based on the structured data object.

[0040] The perception memory reuse module is used to acquire new map information of the new task and target historical environment perception file of the perception type corresponding to the new task when executing a new task, and determine the integrated environment information corresponding to the new task based on the new map information and the target historical environment perception file.

[0041] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0042] Acquire environmental perception data collected by the robot while performing historical tasks, as well as the perception type corresponding to the environmental perception data;

[0043] The environmental perception data is encapsulated into a structured data object with a preset data structure, and historical environmental perception files are obtained based on the structured data object;

[0044] When executing a new task, new map information for the new task and target historical environment perception file of the perception type corresponding to the new task are obtained, and integrated environment information corresponding to the new task is determined based on the new map information and the target historical environment perception file.

[0045] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0046] Acquire environmental perception data collected by the robot while performing historical tasks, as well as the perception type corresponding to the environmental perception data;

[0047] The environmental perception data is encapsulated into a structured data object with a preset data structure, and historical environmental perception files are obtained based on the structured data object;

[0048] When executing a new task, new map information for the new task and target historical environment perception file of the perception type corresponding to the new task are obtained, and integrated environment information corresponding to the new task is determined based on the new map information and the target historical environment perception file.

[0049] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:

[0050] Acquire environmental perception data collected by the robot while performing historical tasks, as well as the perception type corresponding to the environmental perception data;

[0051] The environmental perception data is encapsulated into a structured data object with a preset data structure, and historical environmental perception files are obtained based on the structured data object;

[0052] When executing a new task, new map information for the new task and target historical environment perception file of the perception type corresponding to the new task are obtained, and integrated environment information corresponding to the new task is determined based on the new map information and the target historical environment perception file.

[0053] Sixthly, this application also provides an intelligent cleaning device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the method as described in any of the above embodiments.

[0054] In one embodiment, when the processor executes the computer program, it further implements:

[0055] Based on the integrated environmental information, the cleaning object and the corresponding task processing information are determined; wherein, the cleaning object includes at least one of steps, suspended areas, carpets, or obstacles;

[0056] Based on the task processing information, obtain the navigation strategy and cleaning strategy for the cleaning object;

[0057] The intelligent cleaning device is controlled to perform cleaning tasks according to the navigation strategy and the cleaning strategy.

[0058] The aforementioned robot environmental perception data processing method, apparatus, computer equipment, computer-readable storage medium, computer program product, and intelligent cleaning equipment acquire environmental perception data collected by the robot during historical task execution, as well as the perception type corresponding to the environmental perception data; encapsulate the environmental perception data into structured data objects with a preset data structure; obtain historical environmental perception files based on the structured data objects; and when executing a new task, acquire new map information for the new task, as well as a target historical environmental perception file corresponding to the perception type of the new task, and determine the integrated environmental information corresponding to the new task based on the new map information and the target historical environmental perception file. In this application, by acquiring environmental perception data and its corresponding perception types collected by the robot when performing historical tasks, and uniformly encapsulating them into structured data objects with a preset data structure and archiving them as historical environmental perception files, the standardization and integration of multimodal heterogeneous perception elements at the underlying memory and physical storage level are realized. Then, when performing new tasks, the current new map information is fused with the target historical environmental perception file to construct multi-dimensional integrated environmental information, realizing the efficient transfer of spatial environment prior knowledge in cross-task scenarios, effectively reducing the robot's repeated trial actions in known areas, reducing the algorithm computation load and power consumption caused by invalid detection, and thus improving the efficiency of adaptive operation in complex environments. Attached Figure Description

[0059] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0060] Figure 1 This is a flowchart illustrating a robot's environmental perception data processing method in one embodiment;

[0061] Figure 2 This is a schematic diagram of the memory reuse architecture of a robot in one embodiment;

[0062] Figure 3 This is a structural block diagram of the robot's environmental perception data processing device in one embodiment;

[0063] Figure 4 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0064] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0065] This application provides a method for processing environmental perception data of a robot, which can be applied to intelligent robot systems with environmental perception and task execution capabilities. To cover the broadest possible scope, the term "robot" can refer to any intelligent device with autonomous movement and environmental detection requirements, such as floor-mopping robots, cleaning robots, warehousing and logistics robots, or security patrol robots. This application does not strictly limit the specific product form of the robot.

[0066] In this application environment, the robot's internal hardware architecture typically includes a processor, memory, environmental perception components, and task execution components. The processor, as the robot's core computing and control unit, establishes communication connections with the memory, environmental perception components, and task execution components via an internal system bus. Environmental perception components can include combinations of one or more sensors such as LiDAR, depth cameras, ultrasonic sensors, fall sensors, infrared sensors, or collision sensors. These sensors are used to collect various types of environmental perception data in real time during robot task execution, such as data about steps, suspended areas, carpets, or obstacles. The memory includes non-volatile storage media and internal memory. It not only provides the memory environment for the robot's operating system and underlying programs but also allocates dedicated storage areas or log archive directories for persistently storing structured data objects encapsulated by the processor and generated historical environmental perception files.

[0067] In some application scenarios, when the robot performs historical tasks (such as initial mapping or cleaning), the environmental perception component transmits the collected multimodal perception data to the processor in real time. The processor uses a pre-defined unified multimodal perception data management framework to encapsulate this heterogeneous data into structured data objects with a unified structure, serializes them, and saves them to memory, forming a historical environmental perception file with map markers and timestamps. When the robot triggers and executes a new task in the same area, the processor first retrieves and extracts the corresponding target historical environmental perception file from memory, loads it into memory, and performs deserialization to obtain historical prior information. Then, based on this prior information, it schedules the task execution component and controls the robot to adaptively complete the new task. The aforementioned hardware architecture and its internal data flow process provide the underlying physical support and operating environment for the perception data memory reuse method proposed in this application.

[0068] In one exemplary embodiment, such as Figure 1As shown, a method for processing environmental perception data of a robot is provided, including the following steps S101 to S103. Wherein:

[0069] Step S101: Obtain the environmental perception data collected by the robot when performing historical tasks, as well as the perception type corresponding to the environmental perception data.

[0070] Historical tasks can refer to work processes that the robot has completed or started before the current moment, such as a complete cleaning, inspection, or mapping process. They can be generated based on preset scheduling instructions or triggering conditions and are mainly used to provide a historical state benchmark for subsequent scene operations.

[0071] Environmental perception data can be multimodal data that reflects the state of the surrounding physical space, captured by robots through various onboard sensing devices during movement or operation. It can be used to characterize static terrain or dynamic entities in space.

[0072] The perception type can be an enumerated information or label data used to distinguish and identify the category to which the above environmental perception data belongs. It can be obtained based on pre-set classification rules, including but not limited to various classifications such as step type, suspended type, carpet type, or obstacle type.

[0073] Specifically, when the environmental data processing flow is started, the robot continuously receives raw detection signals uploaded in real time or offline by the sensing components during historical operations, and performs feature analysis and normalization on these raw detection signals to extract the required environmental perception data. At the same time, in conjunction with a pre-built configuration table, the robot identifies and attaches the perception type that matches the environmental perception data, thereby initially sorting and consolidating the originally scattered and disordered perception results in the runtime memory.

[0074] Optionally, in addition to collecting the above data directly in real physical cleaning tasks, it is also possible to obtain prior environmental perception data and its perception type based on background simulation and deduction when the device is in a state of mapping exploration, hibernation synchronization, etc., so as to provide more comprehensive basic input support for the memory reuse framework of configuration drive in the overall solution without being limited to a single task form.

[0075] Step S102: Encapsulate the environmental perception data into a structured data object with a preset data structure, and obtain historical environmental perception files based on the structured data object.

[0076] Among them, the preset data structure usually refers to a predefined data layout model used to standardize multimodal spatial information. It can be constructed based on a combination of constraints of specific dimensions (such as a multi-dimensional storage framework of perception type, global identification information and variable-length field mapping table). It is mainly used to eliminate the storage and I / O differences between different data types and provide a consistent data underlying layer for cross-module interaction of heterogeneous perception information.

[0077] Structured data objects can be entity data containers built in runtime memory according to a preset data structure. They can be used to hold specific field values ​​with strict type constraints, or they can be instantiated after the collected environmental perception data is standardized, encapsulated, and validated for integrity.

[0078] Historical environment-aware files can be file entities that are persistently stored in local non-volatile media. They are mainly used to retain prior knowledge of the runtime scenario after the task ends or the system restarts. They can be generated by performing text stream serialization transformation on structured data objects in memory, and persistently writing them to disk in combination with specific naming rules (such as splicing region identifiers and generation timestamps) and file rolling replacement strategies.

[0079] Specifically, after extracting environmental perception data, the data is format-aligned according to the data specifications defined in the preset data structure, and then injected into a unified storage container according to rules, thereby encapsulating it into a structured data object with the preset data structure. Format alignment can involve matching field descriptions of the acquired environmental perception data, or configuring requirements for the perception type corresponding to the currently acquired perception data. Subsequently, historical environmental perception files are obtained based on the structured data object. During this process, the structured data object instantiated in memory is encoded or format-converted and persistently dumped to the corresponding entity file, thus obtaining and retaining historical environmental perception files reflecting the environmental characteristics of the area.

[0080] Step S103: When executing a new task, obtain the new map information of the new task and the target historical environment perception file of the perception type corresponding to the new task, and determine the integrated environment information corresponding to the new task based on the new map information and the target historical environment perception file.

[0081] A new task can refer to a work process that the robot triggers and prepares for or is executing at the current time point. It can be generated based on a single control command issued in real time, a preset periodic scheduling rule, or a specific event triggering condition.

[0082] New map information can be the spatial topology, grid data, or real-time constructed physical boundary features that the robot relies on in the currently triggered work process. It can be obtained based on real-time localization and mapping algorithms or pre-loaded static base maps.

[0083] The target historical environment perception file can be a specific file entity selected from multiple saved historical environment perception files and associated with the perception type of the current new task. It contains the prior spatial environment characteristics of a specific operating area within a historical time period, providing accurate environmental reference for the execution of the current task.

[0084] Integrating environmental information can be a global cognitive view formed by deeply fusing basic spatial physical contours with multi-dimensional semantic features. In some embodiments, the integrated environmental information can be based on a two-dimensional grid or three-dimensional point cloud as the basic spatial architecture. The map matrix generated by accurately registering spatial environment prior features with precise location coordinates (such as step anti-fall boundaries with height thresholds, carpet polygonal outlines marked with specific material resistance, or static obstacle bounding boxes recording specific length and width dimensions) can be generated based on map information obtained by real-time scanning and target historical environmental perception files extracted by deserialization through matrix superposition or point mapping fusion calculation in a global coordinate system.

[0085] Specifically, in the process of executing a new task, the first step is to acquire new map information for the new task and clarify the spatial structure of the current operation. Simultaneously, the historical environmental perception file corresponding to the new task's perception type is acquired. After extracting the aforementioned spatial basic information and the historical prior memory represented by the target historical environmental perception file, cross-comparison and mapping are performed based on the new map information and the target historical environmental perception file under a unified global coordinate system or local reference. This determines the integrated environmental information corresponding to the new task, enabling the direct reading of the specific environmental attributes mapped to the current coordinate position when generating cleaning trajectories or dynamically adjusting hardware drive parameters. This allows for the early and precise triggering of corresponding obstacle crossing, avoidance, or cleaning mode switching actions.

[0086] For example, in addition to a single retrieval and loading at the start of the task, the target historical environment perception file corresponding to the new task can be retrieved and acquired dynamically in stages during the execution of the new task, as the work space is dynamically switched or the sensor detects a specific boundary in real time, thereby realizing the streaming dynamic loading of prior information.

[0087] In this embodiment, by acquiring environmental perception data and its corresponding perception types collected by the robot when performing historical tasks, and uniformly encapsulating them into structured data objects with a preset data structure and archiving them as historical environmental perception files, the standardization and integration of multimodal heterogeneous perception elements at the underlying memory and physical storage level are realized. Then, when performing new tasks, the current new map information is fused with the target historical environmental perception file to construct multi-dimensional integrated environmental information, realizing the efficient transfer of spatial environment prior knowledge in cross-task scenarios, effectively reducing the robot's repeated trial actions in known areas, reducing the algorithm computation load and power consumption caused by invalid detection, and thus improving the efficiency of adaptive operation in complex environments.

[0088] In one embodiment, environmental perception data is encapsulated into a structured data object with a preset data structure, including:

[0089] Obtain field configuration information for environmental sensing data features used to characterize the sensing type; if the environmental sensing data matches the field configuration information based on the current data fields and current data types contained in the environmental sensing data, then encapsulate the environmental sensing data into a structured data object with a preset data structure; if the environmental sensing data does not match the field configuration information based on the current data fields and current data types contained in the environmental sensing data, then stop the encapsulation operation of the environmental sensing data.

[0090] The field configuration information can be a pre-built set of metadata used to describe and standardize the internal structure of specific categories of data, which can be generated based on the summary of attribute definitions for different environmental elements.

[0091] The current data field can be a specific attribute identifier contained in the actual collected environmental perception data, which is used to characterize the logical features of the input data under a specific measurement dimension.

[0092] The current data type can be a classification label that reflects the format or numerical representation of the data field at the underlying storage level, which can be obtained based on the identification of variable attributes in memory.

[0093] Specifically, during the process of encapsulating environmental perception data into structured data objects with a preset data structure, field configuration information for characterizing the perception data features of the perception type is obtained. After obtaining this configuration baseline, a targeted specification verification process is triggered. If, based on the current data fields and data type contained in the environmental perception data, it is determined that the environmental perception data matches the field configuration information, indicating that the currently input data conforms to the preset structural specifications, the data encapsulation process proceeds normally, encapsulating the environmental perception data into a structured data object with the preset data structure. Conversely, if, based on the current data fields and data type contained in the environmental perception data, it is determined that the environmental perception data does not match the field configuration information, the encapsulation operation of the environmental perception data is aborted, and non-compliant data is directly intercepted at runtime.

[0094] For example, in acquiring field configuration information for environmental sensing data features characterizing sensing types, the field configuration information centrally defines the data field features relied upon by different sensing categories, forming a dynamic configuration mapping table. When business requirements necessitate the integration of new sensing devices or the expansion of entirely new sensing types, only the corresponding field configuration record for the new sensing type needs to be added to the configuration-driven layer. The underlying data management framework will automatically read this newly added field configuration information during the data flow process and perform matching and verification on the input data accordingly, thereby automatically triggering subsequent structured data object encapsulation or interception and abort operations. This configuration table-driven verification logic replaces the previous development approach of independently writing underlying input and output control rules for each specific sensing module, enabling the framework to seamlessly adapt to new data structures and greatly reducing the cost and redundancy of underlying modifications when facing the expansion of sensing elements.

[0095] In this embodiment, by obtaining field configuration information and flexibly deciding whether to perform data encapsulation or abort operations based on the matching of the current data field and the current data type with the field configuration information, configuration-driven data validity verification and automated anomaly interception are realized. This avoids malformed or missing dirty data from polluting the memory storage structure and ensures data processing consistency from the input source.

[0096] In one embodiment, after obtaining field configuration information for environmental perception data features used to characterize the perception type, the method further includes:

[0097] The system retrieves the specified fields and expected data types from the field configuration information, and determines the current data fields and current data types in the environmental awareness data. If any specified field is missing from the current data fields, or if any current data type does not match the expected data type, it determines that the environmental awareness data does not match the field configuration information.

[0098] Among them, the specified field can be a core attribute identifier that is predefined in the configuration system and is indispensable for a certain type of environmental characteristics. It can be set based on the system's bottom line requirements for the integrity of the underlying data and is mainly used to constrain the key information dimensions that the input data needs to have.

[0099] The expected data type can be a valid storage format or structure specification bound to a specified field. It represents the expected numerical form of data in a specific dimension for the preset data storage architecture and is used to determine the serialization / deserialization behavior for perceived data.

[0100] Specifically, after obtaining the field configuration information for environmental sensing data features used to characterize the sensing type, the specified fields and expected data types are retrieved from the field configuration information, and the current data fields and current data types in the environmental sensing data are determined. Then, in the fine-grained comparison logic, if any specified field is missing from the current data field, or if any current data type does not match the expected data type, it is determined that the environmental sensing data does not match the field configuration information, thus providing a precise basis for subsequently suspending the writing of illegal data.

[0101] For example, the field configuration information not only defines each data dimension, but also assigns an attribute identifier (i.e., a constraint field) to each dimension to indicate whether it is a core required field, and a format constraint (i.e., a constraint expected data type) to regulate its numerical representation. During the specific matching verification, the system iterates through and parses the environment-aware data, checking whether its current data fields completely cover all the specified fields marked as required in the configuration system. Simultaneously, it extracts the current data type corresponding to each current data field and performs strict format alignment with the expected data type bound in the configuration system. If, during the verification process, it is found that any core specified field marked as required is missing from the actual input data stream, or that the current data type carrying the actual numerical value conflicts with the expected format specification, the system will determine that the environment-aware data does not match the field configuration information.

[0102] In this embodiment, by extracting the specified fields and expected data types from the field configuration information and comparing and verifying them with the current data fields and current data types, strict screening of the perceived data is achieved from the dual dimensions of field existence and type legality. This effectively intercepts malformed and missing data, further improves the security of the underlying data structure and the robustness of the framework operation.

[0103] In an exemplary embodiment, during the process of encapsulating environmentally perceived data into structured data objects with a preset data structure, the system obtains field configuration information that characterizes the environmentally perceived data features representing the perception type. This process can be implemented based on an underlying configuration-driven layer (e.g., ReuseDataConfigTable), replacing the traditional hard-coding method with a dynamic configuration table. In the field configuration information, the system centrally describes the metadata of each field through field descriptors (e.g., FieldDescriptor structures). This metadata may specifically include the field name corresponding to the JSON key name, the expected data type that determines subsequent serialization and deserialization behavior (e.g., type in the form of a FieldType enumeration), the specified field identifier used to drive the verification logic (i.e., the required attribute that indicates whether the field is a required field), the default value represented by a string (default_value), and the field description (description). To ensure compatibility with multimodal sensing elements, the configuration information for each sensing type is typically designed using a three-layer field structure: The first layer consists of common required fields shared by all types, such as the global identifier (id, expected data type UINT64) and edge shape (edge_shape, expected data type ARRAY_VECTOR2F); the second layer consists of general statistical fields, such as the number of collisions (bump_count or collision_count, expected data type INT), which are usually optional and unspecified fields; the third layer consists of type-specific fields, such as the height (height, expected data type FLOAT) and normal vector (step_normal, expected data type VECTOR2F) set for the step sensing type, or the user-defined attributes (user_set, expected data type BOOL) set for the carpet sensing type.

[0104] After obtaining the above configuration baseline, the system extracts the specified fields (i.e., fields with the required attribute set to true) and their corresponding expected data types from the field configuration information. Simultaneously, it parses and determines the current data fields and current data types in the environment-aware data. Subsequently, if any specified field is missing from the current data field (e.g., the required id field is omitted from the input data), or if any current data type does not match the expected data type (e.g., non-FLOAT data is actually passed to the height field), it is determined that the environment-aware data does not match the field configuration information, thus halting the encapsulation operation of the environment-aware data. Conversely, if the current data fields and current data types are determined to be a perfect match, the environment-aware data is successfully encapsulated into a structured data object with a preset data structure. With this configuration-driven verification architecture, when business scenarios need to expand to new perception types, developers only need to add a corresponding data type configuration record (DataTypeConfig) in the configuration table retrieval interface (e.g., ReuseDataConfigTable::GetConfigTable()). The underlying framework's serialization, deserialization, and verification logic can automatically adapt without modifying other underlying code. This achieves complete decoupling of verification rules from business code, resulting in near-zero expansion costs and the ability to effectively intercept illegal fields in the early stages of operation to prevent system crashes.

[0105] In one embodiment, environmental perception data is encapsulated into a structured data object with a preset data structure, including:

[0106] Obtain the global identifier corresponding to the environmental perception data, and obtain the current field value corresponding to each current data field in the environmental perception data; based on the perception type, global identifier, and each current data field and its current field value, encapsulate to obtain a structured data object with a preset data structure.

[0107] Among them, the global identifier can be a feature code that uniquely identifies a specific sensing event or environmental entity throughout the entire system operation cycle. It can be generated based on timestamp sequences, spatial coordinate system transformations, or specific sequence allocation rules, and is used to accurately locate and distinguish sensing records generated at different times or in different spatial locations in multimodal data streams.

[0108] The current field value can be substantial quantitative data bound to a specific environmental data field, which is used to truly carry the physical dimension information captured by the underlying sensing and detection devices.

[0109] Specifically, in the step of encapsulating environmental perception data into structured data objects with a preset data structure, the global identifier corresponding to the environmental perception data is obtained, and the current field value corresponding to each current data field in the environmental perception data is obtained. Subsequently, the internal data integration and assembly process is initiated. Based on the perception type, the global identifier, and each current data field and its current field value, the data is recombined in memory to encapsulate and obtain a structured data object with a preset data structure.

[0110] For example, in the specific unfolding process of a structured data object based on the perception type, global identifier, and each current data field and its current field value, the system constructs a unified triplet storage container. This storage container uses the perception type as a macro-level classification index, the global identifier as a unique primary key to distinguish individual data entities, and aggregates each current data field and its corresponding current field value through a dynamic mapping table. This structured reorganization method breaks down the barriers that previously required separate planning of underlying logic for different perception modules such as steps, carpets, or obstacles. By uniformly loading the scattered heterogeneous fields into this triplet logical entity, all types of perception information can be circulated, verified, and persisted following the same data foundation.

[0111] In this embodiment, by obtaining the global identifier and the values ​​of each current field, and encapsulating them into a structured data object with a unified structure based on the perception type, the global identifier, and the values ​​of each field, the complex heterogeneous perception information is seamlessly gathered into a standardized three-element storage architecture. This achieves the elimination of system redundancy caused by different perception features independently maintaining the underlying storage logic, and greatly reduces the coupling of the underlying architecture and the subsequent expansion costs.

[0112] In one embodiment, the preset data structure is constructed based on a preset set of data types and is associated with data writing interfaces with different generic parameters.

[0113] Based on the perception type, global identifier, and each current data field and its value, a structured data object with a preset data structure is encapsulated, including:

[0114] Based on the target generic parameters determined by the expected data type, the data writing interface is called.

[0115] Based on the data write interface, the perceptual type, global identifier, and each current data field and its current value are encapsulated into a structured data object. The data write interface is configured to match and validate the generic parameters against a preset data type set during the compilation phase, so as to trigger a compilation block when the generic parameters do not belong to the preset data type set.

[0116] Among them, the preset data type set usually refers to a set of legal structure formats pre-defined in the underlying architecture. It can be set based on the compatibility requirements of multimodal environment information and is mainly used to provide strict whitelist constraints for the types of memory containers.

[0117] Generic parameters are abstract type placeholders used in non-runtime phases such as interface declarations or class template definitions. They are used at the code design level to give the underlying interface the ability to handle various heterogeneous data types.

[0118] The target generic parameter can be an entity type identifier determined at the time of execution based on the actual expected data type, which reflects the real physical structure of the data that needs to be processed.

[0119] The data write interface refers to the access channel used to load multimodal data into a memory container. It can be built based on the aforementioned generalized architecture and aims to provide a unified data integration entry point. At the underlying level, this data write interface is configured to automatically match and validate declared generic parameters against a preset set of data types during the program's compilation phase. If a generic parameter is found to be outside the preset data type set, compilation is proactively halted.

[0120] Specifically, the preset data structure is built based on a preset set of data types and is associated with data write interfaces with different generic parameters. In the process of encapsulating data, the target generic parameter is first determined based on the expected data type, and the data write interface is called using the target generic parameter. Subsequently, based on the execution logic of the data write interface, each current data field and its value are written to the underlying hash mapping table, and encapsulated into a structured data object by combining the type awareness and the global identifier.

[0121] For example, in the design and execution of the underlying system architecture, the solution logically and strictly distinguishes between the structural constraints of the non-runtime phase and the data flow of the runtime phase. In the non-runtime phase (i.e., the program's framework construction and compilation period), the underlying preset data structure employs a secure union container, whose type boundaries are strictly limited to a preset data type set. The corresponding data writing interface introduces abstract generic parameters in its initial declaration and incorporates static type checking rules. When upper-layer business code attempts to instantiate and compile this interface using a specific type, the compiler compares the generic parameters derived during instantiation with the preset data type set; if the derived generic parameters are not within the valid set, a syntax error is thrown directly, triggering a compilation block, preventing the generation of programs with illegal write vulnerabilities.

[0122] During the actual operation of the system, when a real sensor data stream is received, the framework will determine a substantial target generic parameter (e.g., specifying it as a floating-point type or a specific vector matrix type) based on the expected data type defined for the data in the configuration system. Subsequently, the system uses this specific target generic parameter to dynamically call the data writing interface that has successfully passed the non-runtime verification, and stably encapsulates the perception type, global identifier, and specific field values ​​into a structured data object in an absolutely safe memory environment.

[0123] In this embodiment, by distinguishing between the static declaration of generic parameters and the actual call of the target generic parameters, and introducing a matching and verification mechanism for generic parameters and preset data type sets and a compilation blocking mechanism in the non-runtime compilation stage, the technical effect of significantly advancing memory type security review to before system operation is achieved. This eliminates the risk of runtime memory crash caused by illegal data type injection from the source of code construction, and greatly enhances the accuracy of the underlying unified data framework in dealing with multimodal heterogeneous perception data.

[0124] In one exemplary embodiment, the underlying system uses a unified data storage structure called GenericReuseDataItem to instantiate a preset data structure, serving as a unified storage container for all perception types. Internally, this container constructs a triplet physical architecture consisting of (data_type, id, fields): where data_type is an enumeration value used to identify the perception type to which the data belongs (e.g., steps, suspension, carpet, or obstacle); id is specifically declared as a globally unique identifier of type uint64_t, used to carry the global identifier; in some embodiments, a proto message can be defined for each perception type, achieving unified serialization through protobuf reflection; and fields are constructed as a hash mapping table with field names as keys and FieldValues ​​as values, used to dynamically aggregate each current data field and its current field value.

[0125] Furthermore, to ensure the secure loading of multimodal heterogeneous data, the preset data structure is built upon a preset set of data types and is associated with data write interfaces with different generic parameters. In the aforementioned underlying implementation, the FieldValue of the hash table mapping uses the std::variant union type from the C++ standard library for security definition. Its defined legal support range (i.e., the preset data type set) includes: primitive types such as int, uint64_t, float, double, and bool; spatial vector types such as Eigen::Vector2f, Eigen::Vector2i, and Eigen::Vector3f; and std::vector. <int>、std::vector<uint64_t>、std::vector <float>、std::vector <eigen::vector2f>、std::vector <eigen::vector2i>Variable-length array type. Correspondingly, this container provides external interfaces such as SetField. <t>(name,value)、GetField <t>Operation interfaces with template generic parameters T, such as `(name, default)` and `HasField(name)`, serve as data writing and access interfaces. When performing encapsulation, the system determines the corresponding target generic parameter based on the expected actual data type (e.g., when loading a two-dimensional edge shape, the target generic parameter is determined to be `std::vector`). <eigen::vector2f>), and call SetField <t>The (name, value) data writing interface securely encapsulates the type awareness, global identifier, and each current data field and its current field value into a structured data object. During this process, SetField... <t>The templated access interface is strictly configured to automatically perform static matching and validation between the generic parameter T declared in the code and the set of preset data types declared inside std::variant during the compilation phase of the underlying code. If the developer passes an illegal generic parameter that is not in the above-mentioned supported list when calling the interface, the compiler will not be able to deduce the corresponding specialized type, thus directly triggering a compilation block and throwing a syntax error. This architectural design allows the framework to capture and intercept all field type errors during the early compilation phase, completely eliminating the possibility of non-preset data types entering the underlying container.

[0126] In an optional embodiment, before obtaining the global identifier corresponding to the environmental perception data, the method further includes: if the environmental perception data is missing a non-specified field, obtaining the preset value of the corresponding non-specified field in the field configuration information; filling the missing fields in the environmental perception data according to the preset values ​​to obtain the field-filled environmental perception data; and encapsulating a structured data object with a preset data structure based on the perception type, the global identifier, and each current data field and its current field value, including: encapsulating the field-filled environmental perception data to obtain a structured data object with a corresponding preset data structure.

[0127] For example, when parsing the real-time incoming perception data stream, the system iterates through the FieldDescriptor structure records defined in the configuration driver layer. When the system finds that the current incoming data stream does not contain data dimensions whose required attribute in the FieldDescriptor structure is configured as non-mandatory (i.e., non-mandatory, non-specified fields, such as optional fields like bump_count or collision_count under general statistical dimensions), the system will not trigger interception or error logic. Instead, it will directly extract the default_value (e.g., the default initial value represented as a JSON string) declared in the FieldDescriptor structure as the preset value. Subsequently, the system uses the extracted default_value to automatically fill in the missing non-specified fields such as bump_count or collision_count into the current memory data stream, thereby completing the missing field filling of the environmental perception data. After completing the above state completion, the system then performs a subsequent encapsulation step based on the environmental perception data after field filling. This involves safely loading the complete data set containing the above preset values, along with the perception type and global identifier, into the underlying unified storage container (such as GenericReuseDataItem), and then encapsulating it into a structured data object with the corresponding preset data structure.

[0128] In one embodiment, obtaining the target historical environment perception file corresponding to the perception type of the new task includes:

[0129] Obtain the perception type of the new task and the new map information; based on the new map information and the perception type of the new task, search through multiple saved historical environment perception files to obtain a set of candidate historical files; based on the modification time of each candidate historical file in the candidate historical file set, determine the target historical environment perception file.

[0130] The perception type of the new task can be a classification of specific environmental elements in the currently triggered work process, which can be obtained based on the work requirements or task scheduling parameters of the new task.

[0131] New map information is used to characterize the infrastructure or topological features of the physical space where the current operation is located. It may contain specific attributes or identifiers to distinguish different spatial areas, such as a new task map identifier, a unique code to characterize the physical space where the current operation is located, which can be generated based on spatial mapping results or location matching algorithms.

[0132] The candidate historical file set refers to a set of file entities initially selected in the persistent storage directory. It can be formed by batch matching of archived files based on specific search conditions, and is used to narrow down the selection range of the final target files.

[0133] Modification time is a timestamp attribute that records the most recent content change or save of a file, and it can reflect the age of the data entity.

[0134] Specifically, when entering a new workflow, the new task awareness type and new task map identifier reflecting the characteristics of the task are extracted and used as the retrieval key to enter the non-volatile storage layer for targeted querying. In the saved file library, by strictly comparing the above key, files that meet the conditions are aggregated into a candidate historical file set; then, the file system attributes of each file in this set are read, and by sorting and comparing them in the time dimension, the file that best meets the timeliness requirements is selected from the candidate historical file set as the final target historical environment awareness file.

[0135] For example, during the actual execution of acquiring the target historical environment awareness file corresponding to the new task, the new task awareness type and new task map information are extracted, and these two are structurally concatenated according to a preset file naming convention to form a specific string matching pattern (e.g., sequentially combining the awareness type name, separator characters, and map identifiers). This specific matching pattern is used to scan the storage directory of locally saved records, performing pattern matching searches among multiple saved historical environment awareness files, extracting all files whose filenames match the concatenation rule, thus forming a candidate historical file set. After obtaining this set, each candidate historical file is traversed and read, extracting its corresponding modification time, and arranged and compared in chronological order from closest to furthest from the current time. Finally, the candidate historical file with the modification time closest to the current time is identified as the target historical environment awareness file and returned for loading.

[0136] In this embodiment, a candidate historical file set is obtained by initially searching for the new task perception type and new map information. Then, the target historical environment perception file is determined based on the modification time of the candidate historical files. This realizes the automatic isolation of historical perception data in multi-map or multi-floor scenarios and the accurate tracking of the latest version data. It effectively avoids cross-map data pollution between different physical areas and ensures that the most accurate and up-to-date historical environment prior knowledge can be retrieved and utilized in new tasks.

[0137] In one embodiment, the new task includes a cleaning task. After determining the integrated environmental information corresponding to the new task based on new map information and target historical environmental awareness files, the task further includes:

[0138] When performing a cleaning task, the target's historical environmental perception file is deserialized to obtain a structured data object of the perception type related to the cleaning task; based on the environmental perception data and new map information in the structured data object, the robot is controlled to perform the cleaning task.

[0139] The cleaning task can be the action process of the robot to perform ground cleaning operations in the target physical area. It can be generated based on the start command issued by the outside or the timed scheduling rules, and is mainly used to drive the equipment to complete the coverage and cleaning of a specific space.

[0140] Deserialization can be a process of re-parseing and restoring a specific text stream or data stream that is persistently stored on disk into an entity object in memory.

[0141] The target perception type refers to the classification of specific environmental elements that are invoked in the current work process. It can be determined based on the specific needs of the current task and includes, but is not limited to, various perception elements such as steps, suspended areas, carpets, and obstacles.

[0142] Specifically, upon receiving the instruction to initiate ground cleaning operations, the system reads the target historical environmental perception file acquired in the preceding steps and deserializes it according to a pre-agreed data parsing protocol. This process reverse-engineers the static file storage format, re-instantiating a structured data object corresponding to the target perception type in memory. Subsequently, a global spatial cognitive view, constructed by fusing the environmental perception data carried within this structured data object with new map information, is injected into the operation control flow as prior knowledge data of the current scene. Based on the spatial state characteristics in this prior knowledge data, the system guides and controls the robot to safely and adaptively execute the current cleaning task.

[0143] For example, during the cleaning task, when deserializing the target historical environment perception file, the data nodes in the file stream are read item by item according to the preset field metadata in the configuration driver layer (field configuration information), and a structured data object containing identifiers, perception categories, and specific feature values ​​is reassembled in a unified three-element storage container. After obtaining these structured data objects carrying historical prior environmental features, the underlying perception and decision-making no longer need to blindly explore from scratch as in the initial mapping, but directly use the environmental perception data as a cognitive benchmark. For example, when the loaded historical environmental perception data clearly records the existence of steps or fixed obstacle entities in a certain coordinate area, it can be superimposed on the two-dimensional or three-dimensional spatial grid of the current area. Based on this prior feature that integrates historical experience and current terrain, when controlling the robot to perform the cleaning task, path planning instructions for deceleration, avoidance, or edge cleaning can be issued in advance based on this prior feature, so that the robot only needs to perform incremental environmental state updates and cleaning operation scheduling based on historical prior knowledge.

[0144] In this embodiment, by deserializing the target historical environment perception file during the cleaning task and controlling the robot to perform the cleaning task based on the reconstructed structured data object, the efficient and lossless restoration of historical environment perception features to the memory runtime state and deep reuse across tasks are achieved. This reduces the blindness of the device in environmental detection at the beginning of a new task and reduces repeated attempts to explore the same stationary obstacle or special terrain, thereby comprehensively improving the robot's environmental adaptability and cleaning efficiency in complex scenarios.

[0145] In one embodiment, obtaining historical environment-aware files based on structured data objects includes:

[0146] By calling a unified data management interface shared by different perception types, structured data objects are serialized to obtain a structured text stream; map information and generation time of historical tasks are obtained, and a filename is generated based on the splicing result of perception type, map information and generation time; the structured text stream is written into an entity file named with the filename to obtain a historical environment perception file.

[0147] The unified data management interface can be a standardized access channel that is pre-built and shared by various sensing modules. Its main purpose is to avoid the differences in file read and write rules that are maintained independently by various business modules (sensing types).

[0148] Serialization is a process that converts a running entity object in memory into a linear data stream that can be continuously stored or transmitted.

[0149] Structured text streams can be continuous sequences of characters with specific typographic formats produced by the transformation process, which can preserve the inherent hierarchy and mapping relationships of the original data.

[0150] Generation time is a timestamp attribute that records the centralized processing of perceived data, and it can be used to reflect the timeliness of data entity files.

[0151] A filename is an external retrieval identifier assigned to a persistent file entity. It can be generated by concatenating strings based on specific dimensions and is used to accurately locate data within the file system. The entity file, on the other hand, is the data carrier that perceives the data as being stored on physical storage media.

[0152] Specifically, in the process of obtaining historical environmental perception files based on structured data objects, the structured data objects loaded in memory are serialized by calling a unified data management interface shared by different perception types to obtain a structured text stream. At the same time, the map information and generation time of the historical task are obtained, and a filename is generated based on the splicing result of the perception type, map information and generation time. Subsequently, the structured text stream is written into an entity file named with the filename to complete the physical retention of the data, thereby obtaining the historical environmental perception file.

[0153] For example, in the data persistence stage of the underlying system, the previous architecture, where each of the step perception module, carpet perception module, and obstacle perception module maintained its own dedicated data storage rules, has been replaced by a unified data management interface centrally managed at the underlying level. When the environmental data disk persistence process is triggered, the structured data objects generated by each perception module are converted into structured text streams with specific key-value mapping formats through this unified data management interface. To manage environmental prior data in a refined manner in multi-floor or multi-map scenarios, the system extracts map information reflecting the current spatial location and the generation time accurate to the millisecond level. Following a pre-defined string concatenation template, the current perception type, map information, and the generation time containing year, month, day, hour, minute, second, and milliseconds are structurally concatenated according to a specific connector order, and the resulting string sequence serves as the final filename. Subsequently, the structured text stream is written to an entity file named according to this format. This standardized serialization and naming mechanism allows perception data of different modalities to follow the same processing pipeline for physical archiving, forming a distribution system at the physical level that is isolated by map and ordered by time.

[0154] In this embodiment, the structured data object is serialized by calling the unified data management interface, and the file name is generated by combining the perception type, execution map information and generation time to save the entity file. This realizes the standardization and normalization of the multimodal perception data persistence link and the multi-dimensional fine isolation of historical files at the physical level.

[0155] In one embodiment, the method further includes:

[0156] Obtain the number of existing files corresponding to the perception type in the storage directory corresponding to the entity file; if the number of existing files is greater than or equal to a preset threshold, obtain the generation time of each existing file; determine the target cleanup file that is sorted first according to the order of their generation times; delete the target cleanup file from the storage directory and save the entity file in the storage directory.

[0157] The storage directory can be a file path allocated by the system, used for centralized archiving and retention of various historically generated environment-aware entity files.

[0158] The number of existing files can be the total number of valid files belonging to a specific perception type within the current storage directory.

[0159] The preset quantity threshold is a pre-defined upper limit constraint on the number of files to be retained.

[0160] The target cleanup file is the old file entity that is selected and prepared for erasure, which can be determined based on a sorting comparison along the time dimension.

[0161] Specifically, in the process of storing entity files on disk, the number of existing files corresponding to the perception type in the storage directory corresponding to the entity file is obtained; if it is determined that the number of existing files is greater than or equal to a preset threshold, the generation time of each existing file is further obtained; then, the files are sorted and compared in chronological order of their generation times to determine the target cleanup file that is sorted first; finally, the target cleanup file is deleted from the storage directory, and the newly generated entity file is saved in the storage directory.

[0162] For example, whenever a newly serialized entity file is about to be written to the local disk, the system pre-scans the file path responsible for storing record data and accurately counts the number of historical files belonging to the same perception type (e.g., all steps or all obstacles) under that path. If the total number of existing files is found to have reached or exceeded the pre-defined quota limit, the system automatically extracts the timestamp tag of each existing file in that storage directory. After reverse calculation and comparison of the time series, the file furthest from the current time point, i.e., the one with the longest generation time, is marked as the target cleanup file, and the underlying erase command is invoked to remove it from the physical sector. After completing this space release operation, the newly generated entity file is then safely written to that storage directory.

[0163] In this embodiment, by obtaining the number of existing files and combining it with a preset number threshold to perform rolling deletion and new file saving operations based on generation time, automated quota control and lifecycle closed-loop management of persistent storage of perception data are realized. This achieves the goal of effectively preventing the exhaustion of the limited storage resources of embedded devices by massive historical files and ensuring the long-term stability of the underlying operating environment, while ensuring that the device can continuously retrieve the latest scenario prior knowledge.

[0164] In an exemplary embodiment, during the process of retrieving historical environment awareness files based on structured data objects, the structured data objects are serialized to obtain a structured text stream by calling a unified data management interface shared by different awareness types. For example, the persistence operation can be performed using the unified CRUD interface provided by the GenericReuseDataManager. Specifically, its SaveToJson(path) method is called to serialize the data to a JSON entity file, or the SaveToJsonString() method is called to serialize the in-memory data into a structured text stream in JSON string format. Meanwhile, to prevent unexpected large object copying during runtime from causing memory errors, the underlying logic of the data manager is strictly configured to prohibit copy construction and assignment operations. When preparing to write to disk, the execution map identifier and generation time of the historical task are obtained, and a filename is generated based on the concatenation result of the awareness type, execution map identifier, and generation time. Following preset file naming rules, files are assembled using the format {data type name}_map{map ID}_{YYYYMMDD}_{HHMMSS}_{milliseconds}.json. For example, for step features, a filename like step_map123_20241220_143025_456.json is generated. Then, the number of existing files corresponding to the perception type in the storage directory (i.e., / data / log / record / directory) is obtained. If the number of existing files is greater than or equal to a preset threshold (i.e., a maximum of 9 files can be retained for a single perception type), the generation time of each existing file is obtained. Based on the order of their generation times, the oldest file is selected as the target cleanup file. A rolling replacement strategy is executed, automatically replacing and deleting the target cleanup file from the storage directory, and saving the entity file named after the latest generated filename in the storage directory. These are then packaged into the log archive to effectively control the storage usage limit in the embedded environment.

[0165] Furthermore, when executing a new task, the underlying file search strategy can be combined with the `FindLatestDataFile(type, map_id)` retrieval interface. Using `{type_name}_map{map_id}_*.json` as the file system's regular expression matching pattern, the corresponding candidate historical file set can be extracted. The most recently modified entity file is then selected as the target historical environment perception file, thus achieving automatic selection of historical versions isolated by map ID at the underlying level. After successfully locking the file, during the cleaning task, the `LoadFromJson(path)` method in the unified data management interface is called to directly deserialize the JSON file, or the `LoadFromJsonString(str)` method is called to deserialize the extracted JSON string stream. By deserializing the target historical environment perception file, a structured data object of the target perception type related to the cleaning task is reconstructed. Finally, based on the environmental perception data in the structured data object, the robot is controlled to execute the cleaning task with prior knowledge. In addition, throughout the entire data flow and memory reuse cycle, the underlying data manager also provides the AddGenericData(type, item) method for adding newly captured data items at runtime, the GetGenericData(type) method for retrieving all data of a specified type, the GetDataCount(type) method for checking the number of data items in memory, the Clear() or ClearType(type) method for clearing the memory cache as needed, and the ValidateAllData() method for performing full data validation on all elements flowing into the container, thereby building a highly autonomous and closed-loop multimodal data management chain.

[0166] To enable those skilled in the art to better understand the above steps, the following example illustrates the embodiments of this application, but it should be understood that the embodiments of this application are not limited thereto.

[0167] In one embodiment, the aforementioned robot system includes, for example, Figure 2 The robot perception data configuration-driven memory reuse framework shown includes a unified data storage structure, a configuration-driven layer, a data manager, and persistence and file management mechanisms.

[0168] In this embodiment, the system first constructs a unified data storage structure, GenericReuseDataItem, as component one. When processing multimodal sensing data, the system no longer declares separate structures for sensing types such as steps, suspension, carpets, or obstacles. Instead, it uniformly adopts a triple containing (data_type, id, fields) as the standard storage container for all sensing types. Here, data_type is configured as an enumeration value to accurately identify the sensing type to which the current data belongs; id is configured as a globally unique identifier of type uint64_t to distinguish different physical entities or sensing events; and fields is implemented as a hash mapping table with field names as keys and FieldValues ​​as values, enabling dynamic aggregation of heterogeneous features. To ensure memory safety when storing heterogeneous fields, FieldValue uses std::variant from the C++ standard library for unified definition. Its predefined set of supported data types includes: primitive numeric types such as int, uint64_t, float, double, and bool; vector types such as Eigen::Vector2f, Eigen::Vector2i, and Eigen::Vector3f; and std::vector. <int>、std::vector<uint64_t>、std::vector <float>、std::vector <eigen::vector2f>、std::vector <eigen::vector2i>The container is a variable-length array type. It also exposes interfaces such as GetField. <t>(name,default)、SetField <t>The templated access interfaces of (name, value) and HasField(name) enable the compiler to directly use the generic parameter T to perform strict static type checking against the above-mentioned preset data type set when writing and retrieving data at the underlying level. This allows the compiler to capture and block field type mismatch problems at compile time, completely avoiding memory overflow or crashes during robot operation.

[0169] When loading data into the unified container, the system introduces a configuration-driven layer, ReuseDataConfigTable, as component two, replacing the traditional underlying hard-coded logic with a dynamic configuration table. This configuration-driven layer centrally describes the metadata of each field by defining a FieldDescriptor structure. This structure specifically includes: the field name (name) corresponding to the serialized JSON key name, the enumeration type (type) that determines the subsequent serialization and deserialization parsing behavior, the boolean attribute (required) used to drive the underlying data validation logic to determine whether a field is required, the default completed value (default_value) represented in JSON string form, and the field description (description). The data assembly for each perception type strictly adheres to the three-layer field structure specification in the configuration table for verification: First, there is a layer of common mandatory fields shared by all perception types (e.g., requiring the inclusion of UINT64 type id and ARRAY_VECTOR2F type edge_shape); second, there is a layer of general statistical fields that can be included as needed (e.g., INT type bump_count or collision_count); and finally, there is a layer of type-specific fields set for specific perception tasks (e.g., for the step type, mandatory verification of FLOAT type height and VECTOR2F type step_normal, or for the carpet type, verification of BOOL type user_set). Based on this configuration-driven mechanism, when a business scenario requires the expansion of a completely new perception type, the development process only needs to add a corresponding DataTypeConfig record to the ReuseDataConfigTable::GetConfigTable() interface. The underlying serialization, deserialization, and integrity verification logic of the framework will automatically adapt to the new type, thus achieving perception type expansion with near-zero code modification cost.

[0170] Regarding runtime lifecycle management of the framework, this embodiment constructs a data manager, GenericReuseDataManager, as component three, to consolidate all scheduling operations for the perceived data. This manager provides a unified CRUD (Create, Read, Update, Delete) interface matrix to support seamless calls from upper-layer businesses: during the data collection and maintenance phase, newly generated structured data items are pushed into the memory pool by calling AddGenericData(type, item), the full set of environment memory data of a specified type is obtained by GetGenericData(type), the current cached data size is checked by GetDataCount(type), and a full or type-based memory clearing action is performed through the Clear() or ClearType(type) interface. During the data flow and persistence interaction phase, the manager provides the SaveToJson(path) interface to securely serialize in-memory objects to a local JSON file, the LoadFromJson(path) interface to deserialize from a JSON file to restore the in-memory state, and SaveToJsonString() and LoadFromJsonString(str) to meet string-level data stream conversion requirements. Furthermore, the ValidateAllData() interface can be used to trigger full validation of all data in the current memory pool. Specifically, to prevent accidental full copying of large multimodal data sets due to misoperation in complex computational scenarios, which could then degrade the system performance of embedded devices, the underlying code of this data manager is strictly declared to prohibit copy construction and assignment operations.

[0171] For cross-task memory reuse and storage quota management of the robot, the system relies on the persistence and file management mechanism of Component 4 to perform the underlying disk writing and reloading. When the robot receives the current cleaning end signal, the trigger timer immediately wakes up the unified data storage process. The data manager first extracts the current perception type name, the map ID where the robot is located, and the timestamp accurate to milliseconds, and strictly follows the file naming rules {data type name}_map{map ID}_{YYYYMMDD}_{HHMMSS}_{milliseconds}.json to concatenate strings (e.g., generate step_map123_20241220_143025_456.json). Then, it calls the persistence interface to serialize and write the corresponding type of data into the entity file, and packages it uniformly into the / data / log / record / log archive directory. During the writing process, the system synchronously executes a rolling replacement strategy: it detects the number of existing files of a specific perception type in the directory in real time. If the number exceeds the threshold of retaining a maximum of 9 files, it automatically deletes and replaces the oldest file based on file attributes, thus ensuring that storage usage always has a definite upper bound, perfectly adapting to embedded resource-constrained environments. When the robot subsequently starts a new cleaning task in this area, the system calls the file search strategy interface FindLatestDataFile(type, map_id) to search the directory using {type_name}_map{map_id}_*.json as the regular expression matching pattern. This mechanism can automatically filter out data from other map areas and select the file entity with the latest modification time for deserialization and loading, thereby achieving automatic selection of historical versions physically isolated by map ID. In this way, the robot can fully read historical environmental knowledge as prior memory at the beginning of a new task, greatly reducing repeated detection actions for known terrain and obstacles, and enhancing the overall system's environmental adaptability.

[0172] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0173] Based on the same inventive concept, this application also provides an environmental perception data processing device for implementing the above-described robot environmental perception data processing method. The solution provided by this device is similar to the implementation described in the above method. Therefore, the specific limitations in one or more robot environmental perception data processing device embodiments provided below can be found in the limitations of the robot environmental perception data processing method described above, and will not be repeated here.

[0174] In one exemplary embodiment, such as Figure 3 As shown, a robot environmental perception data processing device is provided, comprising: a perception data acquisition module 310, a data structure encapsulation module 320, and a perception memory reuse module 330, wherein:

[0175] The perception data acquisition module 310 is used to acquire environmental perception data collected by the robot when performing historical tasks, as well as the perception type corresponding to the environmental perception data.

[0176] The data structure encapsulation module 320 is used to encapsulate the environmental perception data into a structured data object with a preset data structure, and to obtain historical environmental perception files based on the structured data object.

[0177] The perception memory reuse module 330 is used to acquire new map information of the new task and target historical environment perception file of the perception type corresponding to the new task when executing a new task, and determine the integrated environment information corresponding to the new task based on the new map information and the target historical environment perception file.

[0178] In one embodiment, the data structure encapsulation module 320 is further configured to:

[0179] Obtain field configuration information for characterizing the environmental perception data features of the perception type;

[0180] If, based on the current data fields and current data types contained in the environmental perception data, it is determined that the environmental perception data matches the field configuration information, then the environmental perception data is encapsulated into a structured data object with a preset data structure.

[0181] If, based on the current data fields and current data types contained in the environmental perception data, it is determined that the environmental perception data does not match the field configuration information, then the encapsulation operation of the environmental perception data is terminated.

[0182] In one embodiment, the data structure encapsulation module 320 is further configured to:

[0183] Obtain the specified field and expected data type from the field configuration information, and determine the current data field and current data type in the environment-aware data;

[0184] If any of the specified fields is missing from the current data field, or if any of the current data types does not match the expected data type, then it is determined that the environment-aware data does not match the field configuration information.

[0185] In one embodiment, the data structure encapsulation module 320 is further configured to:

[0186] Obtain the global identifier corresponding to the environmental perception data, and obtain the current field value corresponding to each current data field in the environmental perception data;

[0187] Based on the perception type, the global identifier, and each of the current data fields and their current field values, a structured data object with the preset data structure is encapsulated.

[0188] In one embodiment, the data structure encapsulation module 320 is further configured to:

[0189] Get the data type of the current field value;

[0190] If the data type of the current field value belongs to a preset data type set, then the unified data storage structure is encapsulated based on the perception type, the global identifier, and each current data field and its current field value.

[0191] In one embodiment, the data structure encapsulation module 320 is further configured to:

[0192] By calling the unified data management interface shared by different perception types, the structured data object is serialized to obtain a structured text stream;

[0193] Obtain the map information and generation time of the historical task, and generate a filename based on the splicing result of the perception type, the map information, and the generation time;

[0194] The structured text stream is written into an entity file named with the filename to obtain the historical environment-aware file.

[0195] In one embodiment, the data structure encapsulation module 320 is further configured to:

[0196] Obtain the number of existing files in the storage directory corresponding to the entity file, which corresponds to the perception type;

[0197] If the number of existing files is greater than or equal to a preset threshold, then the generation time of each existing file is obtained;

[0198] Based on the order of their generation times, determine the target cleanup file that is sorted first;

[0199] Delete the target cleanup file from the storage directory and save the entity file in the storage directory.

[0200] In one embodiment, the perceptual memory multiplexing module 330 is further configured to:

[0201] Based on the new task map information and the perception type of the new task, a search is performed in multiple saved historical environment perception files to obtain a candidate historical file set.

[0202] The target historical environment awareness file is determined based on the modification time of each candidate historical file in the candidate historical file set.

[0203] In one embodiment, the new task includes a cleaning task, and the sensor memory reuse module 330 is further configured to:

[0204] When performing a cleaning task, the target historical environment perception file is deserialized to obtain a structured data object of the perception type related to the cleaning task.

[0205] Based on the environmental perception data and new map information in the structured data object, the robot is controlled to perform the cleaning task.

[0206] The various modules in the aforementioned robot's environmental perception data processing device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, so that the processor can call and execute the corresponding operations of each module.

[0207] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 4 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs in the non-volatile storage media to run. The database stores historical perception files. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network. When the computer program is executed by the processor, it implements a method for processing environmental perception data for a robot.

[0208] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0209] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0210] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0211] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0212] In one embodiment, a smart cleaning device is provided, the smart cleaning device including a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program implementing the steps in the above method embodiments when executed by the processor.

[0213] In one embodiment, when the processor executes a computer program, it also implements:

[0214] Based on integrated environmental information, the cleaning object and its corresponding task processing information are determined; the cleaning object includes at least one of steps, suspended areas, carpets, or obstacles; based on the task processing information, a navigation strategy and a cleaning strategy for the cleaning object are obtained; and the intelligent cleaning equipment is controlled to perform cleaning tasks according to the navigation strategy and the cleaning strategy.

[0215] The cleaning target can be a physical environmental entity extracted from the physical work space (such as steps, suspended areas, carpets, or various obstacles), which can be determined based on a previously generated multimodal spatial view. Task processing information can be a set of data representing the constraints or operational procedures required for a specific environmental entity, generated by combining the spatial semantics of the target entity with the current hardware performance status. Furthermore, navigation and cleaning strategies refer to the spatial path planning logic and physical cleaning parameter configuration relied upon when performing the cleaning task. These can be dynamically issued or adaptively scheduled based on the aforementioned task processing information to ensure safe and environmentally appropriate area coverage and cleaning.

[0216] Specifically, in the control process of the intelligent cleaning equipment performing relevant tasks, based on the integrated environmental information obtained from previous compilation, the cleaning object and the corresponding task processing information are determined in the current global or local coordinate system. The cleaning object includes at least one of steps, suspended areas, carpets, or obstacles. Subsequently, based on the parsed task processing information, a navigation strategy and a matching cleaning strategy for the specific cleaning object are dynamically acquired. Finally, the chassis drive module and cleaning head inside the intelligent cleaning equipment are controlled to strictly follow the navigation strategy and the cleaning strategy to collaboratively execute the specific cleaning task.

[0217] For example, in complex environments such as living rooms, after reading integrated environmental information containing rich semantics, the robot identifies a long-pile carpet area directly in front of it (i.e., the corresponding cleaning target). At this point, the robot retrieves the task processing information corresponding to the carpet, determining that the area requires deep vacuuming and that wet mopping should be avoided. Based on this processing rule, the robot quickly issues specific navigation strategies (e.g., planning a path that prioritizes covering the carpet edges and performs an internal zigzag traversal) and associated cleaning strategies (e.g., driving and raising the mop bracket and adjusting the main fan to maximum suction power). Upon receiving these integrated strategies, the underlying control center systematically schedules the wheel motors and vacuum components, precisely entering the area according to the set trajectory and hardware parameters for deep cleaning. Similarly, if the pre-identified cleaning target is a suspended step, the robot will, based on extracted anti-fall processing information, issue a navigation strategy of retreating and turning along the edge and a cleaning strategy of pausing the cleaning operation, thereby smoothly and safely guiding the equipment to avoid dangerous areas.

[0218] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0219] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0220] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0221] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0222] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.< / t> < / t> < / float> < / int> < / t> < / t> < / t> < / t> < / float> < / int>

Claims

1. A method for processing environmental perception data of a robot, characterized in that, The method includes: Acquire environmental perception data collected by the robot while performing historical tasks, as well as the perception type corresponding to the environmental perception data; The environmental perception data is encapsulated into a structured data object with a preset data structure, and historical environmental perception files are obtained based on the structured data object; When executing a new task, new map information for the new task and target historical environment perception file of the perception type corresponding to the new task are obtained, and integrated environment information corresponding to the new task is determined based on the new map information and the target historical environment perception file.

2. The method according to claim 1, characterized in that, The step of encapsulating the environmental perception data into a structured data object with a preset data structure includes: Obtain field configuration information for characterizing the environmental perception data features of the perception type; If, based on the current data fields and current data types contained in the environmental perception data, it is determined that the environmental perception data matches the field configuration information, then the environmental perception data is encapsulated into a structured data object with a preset data structure. If, based on the current data fields and current data types contained in the environmental perception data, it is determined that the environmental perception data does not match the field configuration information, then the encapsulation operation of the environmental perception data is terminated.

3. The method according to claim 2, characterized in that, After obtaining the field configuration information for characterizing the environmental perception data features of the perception type, the method further includes: Obtain the specified field and expected data type from the field configuration information, and determine the current data field and current data type in the environment-aware data; If any of the specified fields is missing from the current data field, or if any of the current data types does not match the expected data type, then it is determined that the environment-aware data does not match the field configuration information.

4. The method according to claim 3, characterized in that, The step of encapsulating the environmental perception data into a structured data object with a preset data structure includes: Obtain the global identifier corresponding to the environmental perception data, and obtain the current field value corresponding to each current data field in the environmental perception data; Based on the perception type, the global identifier, and each of the current data fields and their values, a structured data object with the preset data structure is encapsulated.

5. The method according to claim 4, characterized in that, The preset data structure is constructed based on a preset set of data types and is associated with data writing interfaces with different generic parameters. The structured data object, which encapsulates the preset data structure based on the perception type, the global identifier, and each of the current data fields and their values, includes: Based on the target generic parameters determined by the expected data type, the data writing interface is called; Based on the data writing interface, the perception type, the global identifier, and each of the current data fields and their values ​​are encapsulated into the structured data object; wherein, the data writing interface is configured to perform matching and verification between the generic parameter and the preset data type set during the compilation phase, so as to trigger compilation blocking when the generic parameter does not belong to the preset data type set.

6. The method according to claim 1, characterized in that, Obtaining the target historical environment perception file corresponding to the perception type of the new task includes: Based on the new map information and the perception type of the new task, a search is performed in multiple saved historical environment perception files to obtain a candidate historical file set. The target historical environment awareness file is determined based on the modification time of each candidate historical file in the candidate historical file set.

7. The method according to claim 1, characterized in that, The new task includes a cleaning task. After determining the integrated environmental information corresponding to the new task based on the new map information and the target historical environmental awareness file, the task further includes: When performing a cleaning task, the target historical environment perception file is deserialized to obtain a structured data object of the perception type related to the cleaning task. Based on the environmental perception data in the structured data object and the new map information, the robot is controlled to perform the cleaning task.

8. The method according to claim 7, characterized in that, The step of obtaining the historical environment-aware file based on the structured data object includes: By calling the unified data management interface shared by different perception types, the structured data object is serialized to obtain a structured text stream; Obtain the map information and generation time of the historical task, and generate a filename based on the splicing result of the perception type, the map information, and the generation time; The structured text stream is written into an entity file named with the filename to obtain the historical environment-aware file.

9. The method according to claim 8, characterized in that, The method further includes: Obtain the number of existing files in the storage directory corresponding to the entity file, which corresponds to the perception type; If the number of existing files is greater than or equal to a preset threshold, then the generation time of each existing file is obtained; Based on the order of their generation times, determine the target cleanup file that is sorted first; Delete the target cleanup file from the storage directory and save the entity file in the storage directory.

10. A robot environmental perception data processing device, characterized in that, The device includes: The perception data acquisition module is used to acquire environmental perception data collected by the robot when performing historical tasks, as well as the perception type corresponding to the environmental perception data. The data structure encapsulation module is used to encapsulate the environmental perception data into a structured data object with a preset data structure, and to obtain historical environmental perception files based on the structured data object. The perception memory reuse module is used to acquire new map information of the new task and target historical environment perception file of the perception type corresponding to the new task when executing a new task, and determine the integrated environment information corresponding to the new task based on the new map information and the target historical environment perception file.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the method as described in any one of claims 1-9.

12. An intelligent cleaning device, characterized in that, The intelligent cleaning device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1-9.

13. The intelligent cleaning device according to claim 12, characterized in that, When the processor executes the computer program, it is also configured to perform the following steps: Based on the integrated environmental information, the cleaning object and the corresponding task processing information are determined; wherein, the cleaning object includes at least one of steps, suspended areas, carpets, or obstacles; Based on the task processing information, obtain the navigation strategy and cleaning strategy for the cleaning object; The intelligent cleaning device is controlled to perform cleaning tasks according to the navigation strategy and the cleaning strategy.

Images