102 results about "Active perception" patented technology

The invention discloses a visual language navigation method and system based on a monocular camera and a language instruction, and the method comprises the steps: S1, feature field construction: employing a 3DGS to construct a 3DGS feature field by employing visual information obtained by a monocular RGB-D camera; s2, implicit partial complementation: inferring the representation of a missing region through a context feature relationship to generate a complete feature map; s3, waypoint prediction: generating an aerial view feature map based on the 3DGS feature field, and predicting nearby navigable waypoints through a waypoint predictor; s4, based on active perception of uncertainty, when the robot is uncertain in navigation decision, more visual information is acquired and missing visual information is supplemented by rotating a camera; and S5, constructing a topological map based on the constructed panoramic feature map and the predicted path points, and carrying out navigation decision making by using a panoramic visual language navigation planning model in combination with a language instruction. The problem that monocular vision information is incomplete is effectively solved, and the success rate and efficiency of navigation are remarkably improved.

The invention relates to the technical field of information monitoring, in particular to a remote monitoring method and system for intelligently sensing situation, vibration and orientation. Identifying and acquiring a dynamic construction behavior mode, constructing a space-time dependent topological graph of the dynamic construction behavior mode by monitoring a traceability pattern and an orientation tracking community, and inferring state potential energy of the space-time dependent topological graph according to evolution of the vibration sensing signal and the horizontal sensing data so as to determine a global construction sensing situation of the target monitoring area; and performing phase space evolution trajectory evaluation on the global construction perception situation through an evaluation phase space of a spatial evolution evaluation criterion to obtain a stable node pattern, and introducing a Lyapunov function to analyze whether the divergence degree of the stable node pattern interferes with the buried facility or not so as to control remote monitoring equipment to send out a monitoring early warning signal. According to the invention, all-around real-time monitoring can be carried out on construction behaviors around the buried facility, early warning of potential risks is realized through active perception, and the safety and operation efficiency of the buried facility are improved.

The invention relates to an intelligent robot active cooperation assembling method based on multi-mode large model cooperation driving. The method is used for improving the perceptual understanding, autonomous reasoning and action execution capabilities of the robot in a dynamic complex industrial environment, so that a more efficient, safer and more flexible man-machine autonomous cooperation assembly mode is realized. According to the method, based on brain-like structure design, a multi-center system including a sensing center, a reasoning center and a motor center is constructed, multi-modal large model sensing understanding and semantic reasoning capabilities are fused, and multi-modal large model sensing understanding and semantic reasoning capabilities are obtained through information collaboration and autonomous decision-making among the centers. The robot is driven to realize active perception, semantic understanding, task reasoning and autonomous action execution in an assembly scene, and the adaptive cooperation capability of the robot in a dynamic environment is improved. The method provided by the invention is finally applied to a case of assembling the engine through man-machine cooperation, and the effectiveness of the method is verified.

The invention relates to the technical field of power equipment monitoring, in particular to a multi-mode sensor power equipment state monitoring system based on AI smell. The data acquisition module is used for acquiring multi-dimensional sensing data around the electrical equipment; the AI olfaction processing module is used for converting original sensor data into standardized odor feature fingerprints by performing spatial-temporal feature extraction and multi-modal feature fusion on the multi-dimensional perception data; the overheating state analysis module is used for generating an overheating comprehensive index and an equipment state evaluation result based on the smell characteristic fingerprints; the multi-stage early warning and response module is used for carrying out early warning grade judgment based on the overheating comprehensive index and the equipment state evaluation result and generating graded early warning information; and the storage module is used for storing historical monitoring data and early warning event records. According to the invention, through system integration of the multi-mode sensor array of the bionic olfaction mechanism, AI feature fusion processing and intelligent analysis decision, active sensing and accurate identification of the overheating state of the power equipment are realized.

The invention provides a patient rehabilitation tracking system, a rehabilitation training regulation and control method based on active perception and environment self-adaption, equipment and a medium, and the method comprises the steps: determining a cooperative motion state of a nerve-muscle-bone system according to a multi-modal physiological data flow of rehabilitation training tracking of a target patient; performing intention-muscle force cooperative identification based on the cooperative motion state to obtain motion intention and muscle force contribution distribution of the target patient; determining a dynamic risk map of the target patient during rehabilitation training; variable universe fuzzy reasoning is carried out on the motion intention, the muscle force contribution distribution and the dynamic risk map, and then a rehabilitation adjustment instruction matched with the rehabilitation state of the target patient is obtained; the rehabilitation progress of the target patient is continuously tracked, and the rehabilitation training robot is controlled to execute self-adaptive power-assisted compensation according to the rehabilitation adjustment instruction. By means of the scheme, motion intention decoding and muscle force contribution analysis can be conducted on rehabilitation training of the patient, and rehabilitation training regulation and control are conducted in combination with the dynamic environment risk.

The invention relates to the technical field of Internet of Vehicles and big data processing, and discloses an Internet of Vehicles user perception method and system based on multi-modal data fusion and dynamic portray.The method comprises the following steps that a vehicle terminal collects operation, interaction and environment data and sends the data to a cloud; the cloud access module cleans and vectorizes data; the real-time calculation module generates a fused context feature by using sliding window aggregation and spatial association; the offline module outputs long-term preference and attribute correction characteristics; the dynamic portrait module combines the above characteristics, and constructs a user portrait comprising basic attributes, real-time states and intention labels by using a rule discrimination and deep learning double-track architecture and an adaptive weighting algorithm; and generating and issuing a control instruction according to the matching rule. According to the invention, through spatio-temporal data alignment, static attribute dynamic correction and intention real-time prediction, the problems of portrait lag and prediction delay are solved, and the accuracy and active perception ability of the Internet of Vehicles service are improved.

The invention discloses a vehicle-mounted interaction system and method for adjusting the emotion of a driver, a vehicle and a medium, and relates to the technical field of intelligent driving. The vehicle-mounted interaction system comprises an emotion sensing module, a data processing module and a feedback control module which are connected with one another, the emotion sensing module is used for acquiring multi-modal information of a driver, and the multi-modal information comprises facial data, physiological data and voice data; the data processing module is used for fusing the multi-modal information and determining an emotional state level of the driver; and the feedback control module is used for controlling operation parameters of the multi-channel actuator according to the emotional state level, and the multi-channel actuator comprises an auditory channel, a visual channel, an olfactory channel and a tactile channel. By adopting the vehicle-mounted interaction system, the emotion of the driver can be actively sensed and cooperatively adjusted, and the driving safety and comfort are effectively improved.

The invention provides a 3D target detection tracking method based on visual image and radar tensor sparse proposal fusion, which comprises the following steps: firstly, carrying out generalization extraction on color texture information of a visual image, and establishing multi-scale semantic high-dimensional features; secondly, extracting multi-scale space high-dimensional features of a radar tensor by using SCAN, respectively mapping a learnable sensing probe to radar and visual high-dimensional feature spaces, and performing generalization sparseness on different modal features by means of multi-head deformable attention to form radar and visual proposal features; sparse proposal fusion of radar and visual proposal features is carried out, and 3D target detection is completed; and finally, carrying out mixed multi-feature cascade matching and batch track management on a detection result, and feeding back generated track time sequence information to a front-end learnable sensing probe to realize an active target detection and tracking integrated circulating progressive network based on time sequence information guidance. According to the scheme of the invention, an integrated active sensing framework of single-frame target detection and time sequence target tracking is established, and the reliability of environment target sensing by a multi-source sensor in automatic driving is enhanced.

The invention relates to the technical field of civil engineering construction, and discloses an underneath pass existing station horizontal freezing construction effect data prediction and intelligent control method, which comprises the following steps: constructing a three-dimensional physical model, and collecting multi-source monitoring data in real time; based on simulation and real-time data, an AI agent model capable of predicting the construction effect in real time is generated and synchronously calibrated, and a digital twinning environment is formed; generating a control instruction set for performing differential adjustment on each freezing pipe through a hierarchical reinforcement learning method; quantizing the strategy uncertainty during decision making of the high-level intelligent agent, and generating an active sensing instruction when the strategy uncertainty is greater than a preset threshold value; and executing a control and sensing instruction, performing closed-loop control on physical construction, and optimizing a data acquisition process. According to the technical scheme, the AI agent model is adopted to deduce the construction effect in real time, the reinforcement learning agent generates the prospective control instruction, decision making can be conducted based on prediction of the future state, and active intervention on the freezing construction process is achieved.

The invention relates to the technical field of image data processing, and discloses an intelligent drawing auditing method and device based on a vector atlas and an active perception closed loop, and the method comprises the steps: obtaining a to-be-audited engineering drawing file and a compliance specification text; analyzing vector data of the engineering drawing file, and converting the vector data into an attribute graph used for representing a topological structure and semantics of the drawing; generating an audit instruction through a multi-modal large language model planner based on the attribute graph and the compliance specification text; executing the auditing instruction on the attribute graph through a graph signal processing verifier so as to carry out iterative compliance verification and diagnostic analysis on the attribute graph, and generating a corresponding compliance verification result and a diagnostic result; and generating a compliance audit report containing the traceable evidence based on the compliance verification result and the diagnosis result. According to the method, the accuracy and the reliability of the auditing result can be improved while automatic auditing of the engineering drawing is realized.

The invention discloses a load twinborn modeling and prediction method and system based on multi-source heterogeneous feature fusion, and the method comprises the steps: collecting original data from a user side, a power grid side and an external environment side, and carrying out the preprocessing, and obtaining multi-source heterogeneous feature data; extracting a structural spectrum index and a structural disturbance embedded spectrum entropy index of the multi-source heterogeneous feature data based on a random matrix theory; training a load prediction model based on the multi-source heterogeneous feature data, the structural spectrum index and the structural disturbance embedded spectrum entropy index; and constructing a virtual load twinborn model based on the trained load prediction model, the user static parameters and the dynamic variables, and performing multi-granularity load prediction. According to the method, the prediction precision and generalization ability can be remarkably improved, the interpretability and prediction precision of user load behaviors are improved, active perception and intelligent regulation and control of power system operation are supported, and the method has good engineering expandability and is suitable for dynamic management and control requirements of park-level, regional-level and distributed load resources.

The invention discloses a train control system, method, equipment and medium based on active sensing and autonomous localization, and the system comprises a multi-source autonomous sensing system which collects and outputs the information of the environment in front of a train by fusing the sensing data of a plurality of sensors; the vehicle-mounted standby system is in communication connection with the multi-source autonomous sensing system, accesses a pre-stored digital track map and is used for dynamically generating route topology and movement authorization and generating a train control instruction according to the train front environment information, the train real-time position and the line data of the digital track map; a dual-mode communication switching unit is arranged in the vehicle-mounted standby system and used for monitoring the communication state of the main train control system, and when the main train control system breaks down, the train control right is automatically switched to the standby system. Compared with the prior art, after the main train control system loses efficacy, the train can be made to be independent of a central control system and a fixed electronic map, the real-time state of a line is automatically adapted, and the safe driving instruction is generated.

The invention belongs to the technical field of robot autonomous operation and reinforcement learning, and discloses an intelligent active perception and behavior interaction method in an unknown shielding scene. According to the method, a viewpoint planning strategy of an active sensing agent and a pushing strategy of a behavior interaction agent are closely fused, a 2.5 D occupancy height map is combined, and a pushing point screening mechanism of geometric features and task heuristic weights and a dense reward function are fused; the problems of insufficient perception information, low action efficiency and unstable strategy training in an unknown shielding environment are solved. According to the method, the pushing success rate, the environment entropy reduction efficiency and the strategy generalization ability are remarkably superior to those of a traditional method, efficient environment exploration can be achieved only through minimum invasive pushing of the scene, a good foundation is laid for tasks such as downstream grabbing and cleaning, and the method is suitable for robot autonomous operation of disordered shielding scenes such as bookshelves and storage cabinets.

The invention relates to the technical field of gas detection, and discloses an underground coal mine all-optical cooperative multi-parameter gas detection anti-interference noise suppression system and method.The method comprises the steps that original optical characteristic signals are collected in an underground coal mine area to be detected, and an interference characteristic parameter set and an environment compensation vector are constructed; calculating a real-time estimation noise variance and a real-time credibility weight of each detection channel based on the interference characteristic parameters; performing inversion on the original optical characteristic signal to obtain an original concentration, performing regression compensation by using an environment compensation vector, performing weighted fusion on the compensated concentration based on a real-time credibility weight, and outputting a final fusion concentration; and finally outputting a composite decision in combination with the interference characteristic parameters. According to the method, the quantitative mapping of the interference parameter and the noise variance is established, so that active sensing and adaptive weighted fusion of the health state of the detection channel are realized, the problem of measurement distortion caused by sensor performance degradation is effectively solved, and the robustness and accuracy of the system are remarkably improved.

The invention discloses a target navigation method and device based on active three-dimensional Gaussian spatter and visual language model reasoning, and the method comprises the steps: obtaining an RGB-D image and a pose in an unknown environment, and constructing an incremental three-dimensional Gaussian spatter map as persistent memory through active perception; generating an exploration map based on the constructed 3DGS map, extracting a leading edge point and carrying out space structure adaptive clustering; generating a guidance track, carrying out free viewpoint optimization based on the track, and rendering a leading edge point first-person view angle image containing rich information; constructing a structured visual prompt, combining with a thinking chain prompt, inputting a visual language model to carry out reasoning planning, and selecting an optimal navigation target; in the navigation process, a real-time target detector is used for screening potential targets, and a new view angle is rendered in a 3DGS space through an action decision VLM for target re-verification. The navigation success rate and efficiency are improved.

The invention provides a cooperative identification method and system for a power line communication signal and an arc fault, and the method comprises the steps: firstly, constructing and training a double-flow communication fault sensor comprising a sensing early-warning sub-module and an active sensing sub-module, then carrying out the communication fault sensing based on the double-flow communication fault sensor, generating a communication fault sensing data set, and transmitting the communication fault sensing data set to a server; then, a graph neural network is adopted to carry out fault mode recognition on the communication fault sensing data set, a communication fault mode graph is generated, then, communication fault positioning is carried out based on the communication fault mode graph, a communication fault positioning data set is generated, and finally, actual feedback data of the communication fault positioning data set is obtained. And the system is updated according to the actual feedback data, so that the fault diagnosis rate and accuracy are improved.

The invention relates to the technical field of navigation system error correction, and discloses a multi-path error correction method in a Beidou navigation system, which comprises the following steps of: scanning through an environment perception sensor integrated with a navigation receiver to obtain three-dimensional space data of a surrounding environment of the navigation receiver; generating a set comprising a plurality of candidate paths according to the space reflection source model; establishing a sparse optimization model taking an observation matrix as a core through sparsity priori of a plurality of candidate path signals; solving the sparse optimization model, and identifying a real reflection path which forms dominant influence on the pseudo-range observation value; and a corrected pseudo-range observation value is obtained, and a final positioning result is calculated based on the corrected pseudo-range observation value and is fed back. According to the method, potential reflecting surfaces around a receiver are actively sensed and parameterized through an environment modeling and geometric path prediction mechanism based on three-dimensional space data, and the additional delays of all candidate reflecting paths are calculated according to the optical reflection law.

The invention relates to the technical field of natural resource monitoring and evaluation, and discloses a natural resource dynamic monitoring and evaluation system based on multi-source spatio-temporal data fusion, comprising a semantic mapping module for analyzing a business document to generate a constraint rule and constructing a basic spatio-temporal tensor; the perception strategy module monitors state node evolution and generates an acquisition strategy instruction to drive an external sensor to obtain enhanced perception data; the state evolution module deduces an expected legal state sequence and reconstructs an actual physical state sequence in combination with enhanced perception data; the energy quality fingerprint module establishes nonlinear mapping by using the time sequence energy quality data, and calculates a fingerprint consistency index to verify the authenticity of a physical state; and the deviation risk module calculates a space-time semantic distance and a propagation risk value based on the topological network in combination with the consistency index, and generates a resource audit report. According to the invention, through event-driven active perception and cross-modal energy quality implicit verification, the problems of insufficient monitoring data space-time precision and difficult camouflage compliance identification are effectively solved.

The invention discloses a target navigation method and system for active perception and rule-guided reinforcement learning, and relates to the technical field of artificial intelligence, and the method comprises the following steps: constructing an environment semantic map by using an environment image and position information obtained by a robot on a global map; in the global exploration stage, a heuristic exploration rule is utilized to guide a reinforcement learning strategy, a robot is driven to carry out adaptive global exploration on the environment, and a coarse-grained target area is determined; in the tail end positioning stage, target semantic detection feedback information of a target object is obtained by using an open vocabulary semantic detection model and an arbitrary segmentation model; inputting target semantic detection feedback information and an environment semantic map into a tail end reinforcement learning model, and dynamically adjusting the position and orientation of the robot by using a reinforcement learning strategy; according to the target navigation method and system, the global exploration efficiency and the tail end positioning precision of the robot in an open vocabulary complex environment are remarkably improved, and the navigation success rate and the optimization of the operation posture are both considered.

The embodiment of the invention provides a multi-modal large model agent collaborative water quality inversion method based on active perception, which belongs to the technical field of data processing, and specifically comprises the following steps: step 1, a decision agent issues a water quality monitoring task; 2, the data intelligent agent collects multi-source remote sensing and ground monitoring data, image preprocessing and multi-modal enhancement are carried out, and a standardized input sample is generated; step 3, analyzing intelligent agent fusion image, time sequence and text features, and performing water quality level prediction and natural language interpretation through a multi-modal large language model; 4, performing task scheduling and resource allocation by the decision-making agent based on a water quality grade prediction result, generating a standard operation process, and triggering sampling and early warning; and 5, applying the execution result and the reward feedback to strategy optimization and parameter updating of the decision-making agent, and continuing to publish a new task to form a self-adaptive closed-loop iteration mechanism. Through the scheme of the invention, the prediction accuracy and interpretability are improved.

The invention discloses a robot indoor person searching method and system based on active perception. According to the method, the distribution and orientation information of the characters in the scene is analyzed in real time based on an active perception strategy, the minimum observation view is dynamically planned, repeated invalid search is effectively avoided, and more efficient character search is realized by filtering irrelevant characters and similar characters in advance for combined observation. According to the method, a multi-feature fusion recognition mechanism is realized through character color features and facial features. The invention provides a multi-person head posture estimation method based on fuzzy reasoning, and deployment and real-time reasoning can be carried out on a mobile robot platform.

The present invention relates to an embodied intelligence agent based on physical intelligence and active perception, and a control method therefor. The method controls the embodied intelligence agent on the basis of a multimodal embodiment large model, and the multimodal embodiment large model comprises a data processing module, a perception and planning module, a physical intelligence module, an information transmission module, an environment interaction module, and a motion control execution module. The method comprises: the perception and planning module receives data processed by the data processing module and, according to the data, outputs perception information and a task planning instruction; the physical intelligence module receives the perception information and the task planning instruction, and outputs a motion action planning instruction; the environment interaction module collects environment information in real time and compares same with the perception information; and the motion control module receives the motion action planning instruction and incorporates the comparison result to perform action correction so as to implement control of the embodied intelligence agent. Compared with the prior art, the present invention has the advantages of high generalisation performance and high task execution accuracy.

The invention relates to the field of body intelligence, and provides an autonomous decision-making method and system for a body intelligent robot. The autonomous decision-making method comprises the following steps of: acquiring environment and interaction data in real time by adopting an active perception mechanism, and converting the environment and interaction data into a scene and an own semantic code fused with the scene and the own; inputting the subject semantic code into a shunting decision-making layer to output a structured semantic abstract and an execution action, wherein the structured semantic abstract comprises scene parameters, user requirements and confidence; in response to the fact that the confidence is located in the middle confidence interval, determining and executing a non-invasive tentative action instruction used for verifying the user demand based on the execution action and the scene parameters; collecting user feedback data when the non-invasive heuristic action instruction is executed to determine a user feedback result; and determining a decision result based on a user feedback result, and adjusting the range of the dynamic confidence interval and the association weight between the execution action in the shunting decision layer and the user demand.

The invention discloses a bicycle driving risk early warning method and system based on visual identification, and relates to the technical field of visual identification. The method comprises the following steps: acquiring driving sensing data through a sensing module carried on a bicycle body; identifying a plurality of dynamic traffic participants based on the visual perception data, and predicting future behavior intentions and motion trails; carrying out risk assessment by fusing the predicted behavior intention, the motion trail and the running state data of the bicycle, and constructing a risk perception grid which comprises a traffic event, a participant, a risk level and a movement tracking state; and based on the risk perception grid, identifying the traffic event with the risk level or the mobile tracking state reaching an early warning threshold value and the positioning space, and generating early warning feedback information. The technical problems of insufficient bicycle driving risk identification and untimely early warning in the prior art are solved, and the technical effects of driving risk active perception and timely early warning based on visual prediction are achieved.

The invention discloses an autonomous operation robot and method for a power frequency line parameter tester in a high induced electricity environment, and belongs to the technical field of robots. The autonomous operation robot comprises a physical layer, a middleware layer, a cognition and control layer and a monitoring and training layer, the physical layer comprises a humanoid robot body, a high-precision multi-degree-of-freedom dexterous hand, a data acquisition module and a working frequency line parameter tester serving as an operation target; the middleware layer realizes real-time and safe data exchange and communication among modules of the physical layer and the cognitive and control layer; the cognition and control layer is used for realizing active perception, autonomously outputting a control instruction and driving the robot body or the head to adjust the position so as to re-observe a target and form closed-loop perception; and the monitoring and training layer is used for training, verifying and iteratively optimizing a control strategy. The robot is trained in a field randomization environment through an active sensing strategy, and environment disturbance of a real operation site can be effectively coped with.

The present invention proposes a dual-arm robot wiring control method and system based on active perception, which relates to the field of robot control technology. It includes constructing a high-level primitive selection imitation learning network, identifying the primitive that needs to be executed currently based on the local image, the global image and the position of the end effector of the operating robot arm; inputting the local image and the position of the end effector of the operating robot arm into the low-level primitive imitation learning network model, obtaining the next action of the operating robot arm, and updating the position of the end effector of the operating robot arm; inputting the global image, the position of the end effector of the perceived robot arm and the cable point cloud data into the active perception network model, obtaining the next action of the perceived robot arm, and updating the position of the end effector of the perceived robot arm; and looping the above steps based on the updated information to complete the entire wiring process. The present invention adopts an active perception method to reduce the self-occlusion rate of the robot operating deformable objects and improve the robot's autonomous wiring capability.

The application discloses an AI teaching feedback and self-evolution method and system based on active sensing and mixed thinking mechanism. The method comprises the following steps: generating student behavior characteristics, emotion characteristics and teacher-student voice characteristics; generating student state-teaching content structured information through a semantic analysis model; obtaining behavior emotion analysis information through a thinking scheduling module core model; generating visual perception module optimization parameters, semantic analysis module optimization parameters and thinking scheduling module optimization parameters according to all historical high-quality classroom video annotated efficient interaction segments and behavior emotion analysis information of each period; and updating the visual perception type model, the semantic analysis model and the thinking scheduling module core model. The application constructs a self-evolution system, which can continuously improve the sensing accuracy, analysis accuracy and scene adaptability without manual intervention, and completely solves the problems of static and no evolution ability of the traditional teaching analysis system.

The invention discloses an automatic medical record generation method and system based on multi-modal data and active perception, and belongs to the technical field of data processing.The method specifically comprises the steps that multi-modal original data of a patient and a doctor are collected, a field evidence graph is constructed, a constraint set is generated according to a clinical ontology and a time sequence rule, and configuring a priority and a dynamic threshold for each medical record field, generating an additional collection action based on information gain, and performing iterative update on the field evidence graph, the iterative update being to control the additional collection action of a target field until the target field reaches the dynamic threshold or meets a constraint set. Performing constraint decoding on medical record fields in the iteratively updated field evidence graph to obtain each target field value, and outputting a medical record with an evidence fingerprint chain; according to the method, the multi-modal information required by the medical record can be comprehensively covered, the information gap can be dynamically found and actively supplemented in the acquisition process, and the logic consistency and the time sequence reasonability between the fields are ensured.

The present invention provides a method and system for self-evolution of heterogeneous systems based on active perception and intelligent prediction, including: generating a heterogeneous system operating environment, tracking and monitoring each request, formatting and extracting the perceived sensitive information, and performing consistency judgment; locating defects and reproducing attack scenarios in defective heterogeneous system operating environments, performing defect processing, extracting corresponding features to construct an attack knowledge base and a defect knowledge base; pre-perceiving system component defects for subsequent attack requests and sensitive information generated in the system, and dynamically changing the system and its operating environment before the attack chain is completed; performing robust execution judgment for suspected malicious requests, and selecting a processing result that meets preset conditions as the request response. The present invention improves the security strength of the system and its operating environment by intelligently predicting threats, proactively and pre-emptively performing system self-evolution, and eliminating exploitable component defects.

The application discloses a kind of multi-machine cooperation SLAM methods based on active deep reinforcement learning.The method comprises the following steps: running ORB-SLAM2 program to robot, and the initial motion trajectory pose graph of multiple machines is obtained by pose estimation of image acquisition through camera;Based on the obtained robot motion trajectory pose graph, more accurate pose is obtained by using deep reinforcement learning TD3 algorithm training to optimize trajectory;On the basis of reinforcement learning algorithm, active perception strategy is introduced to optimize the pose of multiple machines simultaneously, and according to the real-time SLAM estimation probability value P, the corresponding robot is selected to optimize the pose information by TD3 algorithm;The pose information and actual distance information of each robot are transmitted between robots, and the back-end optimization of SLAM trajectory is carried out using TD3 algorithm, to eliminate the cumulative error effect.The application can effectively eliminate the error accumulation in SLAM system, improve the positioning and mapping accuracy of SLAM, and there is no loop restriction, which increases the robustness of SLAM system.