CN-121832551-B - Space intelligent guide instruction generation and execution verification method based on environment geometric constraint and electronic equipment

CN121832551BCN 121832551 BCN121832551 BCN 121832551BCN-121832551-B

Abstract

The application provides a space intelligent guide instruction generation and execution verification method based on environment geometric constraint and electronic equipment, and belongs to the technical field of space intelligent and man-machine collaborative navigation. The method comprises the steps of obtaining multi-mode environment data, executing three-dimensional computer graphic modeling processing to construct a three-dimensional geometric model representing environment characteristics, analyzing spatial semantic components in natural language task instructions, mapping characteristic information analyzed from the spatial semantic components to corresponding three-dimensional modeling objects in the three-dimensional geometric model to generate a geometric constraint characteristic set representing reasonable constraint of a guiding space, executing geometric consistency analysis and space path solving based on the geometric constraint characteristic set in the three-dimensional geometric model to generate a spatial intelligent guiding instruction with collision avoidance attributes, collecting running track data in the process of executing the instruction by an agent in real time, executing space geometric matching calculation on the running track data by utilizing the three-dimensional geometric model, and generating a track verification result representing the execution accuracy of the instruction.

Inventors

ZHANG YU
HE WENWU
Song Zhouying
ZHU XUPING
CHENG YUHANG
ZHANG XIAOBO
CHEN WENBO

Assignees

北京飞渡科技股份有限公司

Dates

Publication Date: 20260508
Application Date: 20260311

Claims (9)

1. The method for generating and executing the verification of the intelligent space guiding instruction based on the geometric constraint of the environment is characterized by comprising the following steps: Step 1, acquiring multi-modal environment data and executing three-dimensional computer graphic modeling processing to construct a three-dimensional geometric model representing environment characteristics, wherein the three-dimensional geometric model comprises a space topological structure and geometric attribute description of barrier boundary entities, and the geometric attribute description of the barrier boundary entities is expressed based on geometric primitives; Analyzing space semantic components in a natural language task instruction, and mapping feature information analyzed from the space semantic components to corresponding three-dimensional modeling objects in the three-dimensional geometric model to generate a geometric constraint feature set representing guide space rationality constraint, wherein the geometric constraint feature set comprises a distance and an angle threshold representing a space obstacle avoidance safety boundary; step 3, executing geometric consistency analysis and space path solving based on the geometric constraint feature set in the three-dimensional geometric model to generate a space intelligent guide instruction with collision avoidance attribute; Step 4, acquiring running track data of the intelligent agent in real time in the process of executing the space intelligent guiding instruction, and executing space geometric matching calculation on the running track data by utilizing the three-dimensional geometric model to generate a track verification result representing the execution accuracy of the instruction; In step 4, the process of executing the space geometric matching calculation includes converting the moving track data into a pose sequence under a coordinate system where the three-dimensional geometric model is located, calculating a Euclidean distance between the pose sequence and a constraint boundary defined by the geometric constraint feature set, and generating the track verification result according to whether the Euclidean distance exceeds a preset deviation threshold value.
2. The method for generating and executing the verification of the space intelligent guidance command based on the environment geometric constraint of claim 1, wherein in step 1, the process of constructing the three-dimensional geometric model includes the steps of acquiring depth images in the multi-mode environment data by using a depth camera, generating three-dimensional point cloud data through a point meshing process, executing semantic segmentation processing on the three-dimensional point cloud data to extract the obstacle boundary entity, executing topology analysis on the three-dimensional point cloud data, and identifying the adjacency relation between a passable area and a non-passable area to generate the space topology structure in the three-dimensional geometric model.
3. The method for generating and executing the verification command based on the space intelligent guidance command of the environment geometric constraint according to claim 1, wherein in the step 2, the parsed feature information comprises task verbs and space nouns extracted from the space semantic components, and the process for generating the geometric constraint feature set comprises the steps of extracting the task verbs and the space nouns by utilizing a semantic parsing module, establishing a mapping relation between the task verbs and the space pose based on the three-dimensional geometric model, and establishing topological association between the space nouns and the three-dimensional modeling object, so that the geometric constraint feature set is calculated.
4. The method for generating and executing the verification of the space intelligent guidance command based on the environment geometric constraint according to claim 3, wherein in step 3, the process of executing the geometric consistency analysis comprises the steps of extracting a distance and angle threshold value representing a space obstacle avoidance safety boundary in the geometric constraint feature set, searching relative pose of the intelligent object and the obstacle boundary entity in the three-dimensional geometric model in real time, and executing logic comparison of the relative pose and the distance and angle threshold value to determine whether the current path search space meets the geometric consistency analysis.
5. The method for generating and executing the verification of the space intelligent guide command based on the environment geometric constraint according to claim 1, wherein the method further comprises the steps of establishing a closed loop learning unit, receiving the track verification result by utilizing the closed loop learning unit, and executing model parameter dynamic update processing according to a deviation source represented by the track verification result so as to optimize the analysis natural language task command and execute the geometric consistency analysis.
6. The method for generating and executing the verification of the spatial intelligent guidance command based on the geometric constraint of the environment according to claim 5, wherein in the step 5, executing the dynamic update processing of the model parameters includes enhancing the weight for collision risk assessment in the geometric consistency analysis to optimize the distance and the angle threshold for characterizing the spatial obstacle avoidance safety boundary if the geometric error of the trace verification result characterizes the geometric error threshold of the design, and updating the word vector weight for intention understanding reasoning in the semantic parsing module if the trace verification result characterizes the semantic matching error.
7. The method for generating and executing the verification of the space intelligent guide command based on the environment geometric constraint according to claim 1, wherein the method further comprises the steps of establishing a scene mode recognition and self-adaptive learning mechanism to count scene deviation distribution under different environment characteristics to generate a scene deviation distribution map and configure an optimization strategy set for the scene deviation distribution map, and in a follow-up task, if the real-time environment characteristics are detected to be matched with the history record, automatically calling the optimization strategy set corresponding to the scene deviation distribution map to optimize the space intelligent guide command.
8. The method for generating and executing the verification of the space intelligent guide command based on the environment geometric constraint according to claim 7, wherein the method further comprises the steps of establishing a task review mechanism, executing playback processing on the space intelligent guide command and the running track data in the process of executing the historical task, and optimizing the step of generating the space intelligent guide command with collision avoidance attribute based on a self-supervision learning algorithm so as to adjust the adaptation of the three-dimensional geometric model to a dynamic scene.
9. An electronic device comprising a memory and a processor, the memory having stored thereon a computer program executable on the processor, the processor executing the method steps of any of claims 1 to 8 when the computer program is executed.

Description

Space intelligent guide instruction generation and execution verification method based on environment geometric constraint and electronic equipment Technical Field The application relates to the technical field of space intelligent and man-machine collaborative navigation, in particular to a space intelligent guide instruction generation and execution verification method based on environmental geometric constraint and electronic equipment. Background Along with the wide application of the intelligent robot, the augmented reality equipment and the autonomous inspection in the complex indoor environment, how to realize the deep fusion of the natural language instruction and the physical space environment, guide the intelligent body to safely and accurately execute the navigation task becomes a research hot spot in the field of space intelligence. The efficient space guidance not only needs to accurately understand semantic intentions of users, but also needs to convert the intentions into executable instructions conforming to physical geometric laws, and performs real-time deviation verification and closed-loop optimization on the execution process. In the existing space navigation guidance scheme, a path planning method based on a semantic map is generally adopted. The scheme comprises the steps of firstly pre-constructing an environment map containing semantic tags, then obtaining a navigation instruction of a user through voice recognition or text analysis, positioning a target coordinate point according to the semantic tags, finally calculating a path from a current point to the target point by using a global path planning algorithm, and issuing the path to a bottom control module of the robot for execution. However, this existing navigation guidance scheme has obvious technical drawbacks. Because its semantic understanding is in a relatively isolated state from the physical space geometric modeling, the generated instructions often lack real-time constraints on environmental geometric details (e.g., narrow channel width, obstacle edge morphology), resulting in the generated paths being semantically correct, but there may be problems of collision risk or insufficient space rationality at the physical geometry level. Meanwhile, the existing scheme lacks real-time geometric alignment verification on the instruction execution process, and cannot identify fine deviation between the running track and the original geometric constraint, so that the method lacks dynamic deviation correction feedback and self-evolution learning capability, and is difficult to adapt to a high-dynamic and high-complexity space interaction environment. Disclosure of Invention In order to solve the technical problems, the application provides a space intelligent guide instruction generation and execution verification method based on environmental geometric constraint and electronic equipment so as to at least alleviate the technical problems. A space intelligent guide instruction generation and execution verification method based on environment geometric constraint comprises the following steps: Step 1, acquiring multi-mode environment data and executing three-dimensional computer graph modeling processing to construct a three-dimensional geometric model representing environment characteristics, wherein the three-dimensional geometric model comprises a space topological structure and geometric attribute description of obstacle boundary entities; Analyzing space semantic components in a natural language task instruction, and mapping feature information analyzed from the space semantic components to corresponding three-dimensional modeling objects in the three-dimensional geometric model to generate a geometric constraint feature set for representing a guide space rationality constraint; step 3, executing geometric consistency analysis and space path solving based on the geometric constraint feature set in the three-dimensional geometric model to generate a space intelligent guide instruction with collision avoidance attribute; And 4, acquiring running track data of the intelligent agent in real time in the process of executing the space intelligent guide instruction, executing space geometric matching calculation on the running track data by utilizing the three-dimensional geometric model, and generating a track verification result representing the execution accuracy of the instruction. Optionally, in step 1, the process of constructing the three-dimensional geometric model includes acquiring a depth image in the multi-modal environment data by using a depth camera and generating three-dimensional point cloud data through a point meshing process, performing semantic segmentation process on the three-dimensional point cloud data to extract the obstacle boundary entity and express a geometric attribute description of the obstacle boundary entity based on geometric primitives, performing topology analysis on the three-dimensional point cl