141results about How to "Improve search speed" patented technology

The invention relates to the technical field of building design, and discloses a GAN and GA-based low-carbon building design decision-making method in a hot-summer and warm-winter region, and the method comprises the steps: setting a generator in a GAN model to generate external design parameters and internal design parameters of a building in a hierarchical manner, introducing a reinforcement learning algorithm, and setting an adaptive discriminator, thereby achieving the design of a low-carbon building in a hot-summer and warm-winter region. And the genetic algorithm GA is combined with particle swarm optimization PSO, so that the balance between the energy efficiency and the building material cost can be achieved while the low-carbon emission requirement is met, and the intelligence and dynamic response capability of the building design are fully realized.

The invention provides an industrial defect image detection method based on dynamic regular entropy differential architecture search, and relates to the technical field of image analysis. Industrial image data to be detected and classified are acquired, an industrial image data set is constructed, and a training set and a verification set are divided; the method comprises the following steps: constructing a hypernet model capable of differential search for a search architecture, initializing a network weight omega and an architecture parameter alpha, and calculating a loss function and a loss model and an optimizer for gradient update; multi-round training is carried out, the training set and the verification set are respectively used in each batch of each round to calculate loss, and omega and alpha are updated; and after training is completed, discretizing the alpha to obtain a neural network architecture for final search, retraining the architecture on the whole data set, taking the architecture as a classification model for detecting abnormal defects in the industrial image, and inputting an actual industrial image to be detected and classified to obtain a defect detection result. According to the method, performance collapse does not occur under long-time search, and a framework with excellent detection performance can be searched.

The invention provides a method for calculating a flexural capacity characteristic value of a steel and UHPC (Ultra High Performance Concrete) composite structural beam. Different constitutive models are adopted according to performance characteristics of UHPC, a T-shaped steel beam and a steel bar in the composite structural beam, and the contribution of material strain hardening performance to the flexural capacity is fully considered; compared with an ideal elastic-plastic model, the real performance of the structure can be better obtained; meanwhile, the cross section curvature, the curvature iterative calculation increment and the neutral axis position iterative calculation increment are given for initial values of conventional bridge calculation, and the neutral axis position and the cross section curvature are searched by adopting an influence coefficient method, so that the search speed is increased, and the problem that convergence cannot be realized due to manual setting of the iterative increment is solved; and finally, the condition that the UHPC top strain reaches the UHPC compression limit strain or the T-shaped steel beam bottom strain reaches the steel limit strain is taken as a judgment condition, and two damage modes of UHPC crushing and the condition that the steel reaches the tensile ultimate strength can be considered at the same time.

The invention relates to the technical field of fire prediction, in particular to an improved GWO-based neural network fire trend prediction method, which comprises the steps of collecting fire detection parameters; the detection parameters are preprocessed; establishing a fitness function according to a mean square error of a predicted value and an actual measurement value of the detection parameter; initializing a hyper-parameter and a fitness function of the deep learning network model, and optimizing the hyper-parameter of the deep learning network model by using an improved GWO algorithm; and inputting fire detection parameters collected in real time into the improved deep learning network model to predict the fire. The problems that an existing fire treatment method is low in treatment speed and poor in accuracy are solved.

The invention discloses a mobile terminal path planning method and system based on a distance field, and the method comprises the steps: firstly building an ESDF map, and calculating a gradient value of each grid point in the ESDF map; establishing a state vector and a discretization state transition equation of the mobile terminal based on the vehicle model; and finally, calculating the next state of the mobile terminal moving from the current node by using an improved # imgabs0 # algorithm until an optimal path from the starting node to the terminal node is formed. Wherein the ESDF cost is calculated by using the ESDF value and the gradient value of the node, and the cost function of the improved # imgabs1 # algorithm is the sum of the actual cost, the heuristic estimation cost and the ESDF cost. According to the distance field and gradient information of the ESDF map, collision detection and expansion cost calculation in the # imgabs2 algorithm are combined, and the path planning efficiency is improved while the path safety is guaranteed.

The invention provides a cross-platform live broadcast visitor data fusion and routing load balancing method and system, and relates to the technical field of load balancing, and the method comprises the steps: filtering repeated visitor identifiers through a Bloom filter array, calculating the number of independent visitors through a logarithm radix counting algorithm, and dividing the priority of the visitors based on a data flow entropy value; constructing an annular routing structure by using a consistent Hash algorithm, mapping visitors to virtual nodes, and performing fragmentation routing based on priorities; the response time of the virtual nodes is recorded in real time, abnormal nodes are identified, associated visitors are migrated to standby nodes, and dynamic load balancing is achieved; the visitor data processing efficiency and the system stability of the live broadcast platform are effectively improved, the load pressure of the server is reduced, and the fluency of the live broadcast service is guaranteed.

The invention provides an unmanned vehicle obstacle avoidance path planning method, and relates to an unmanned vehicle path planning method, which comprises the following steps of: adding an improved gravitational potential field to a target point, adding an improved repulsive potential field to an obstacle, meanwhile, a road boundary potential field and a speed repulsion potential field are introduced to ensure that the unmanned vehicle does not collide with a road boundary and can avoid static and dynamic obstacles; when the path nodes are beyond the maximum influence distance of the obstacle, a traditional gravitational potential field is added to the two adjacent path nodes in the bat position updating formula, and the path length is shortened; in the updating process of the local new solution of the bat algorithm, two strategies of Gaussian distribution and Cauchy distribution are used for searching to generate the local new solution, so that the searching speed and the searching precision are improved. Based on the improvement, static and dynamic obstacle avoidance can be realized while the path is ensured to be as short as possible in the driving process of the unmanned vehicle, and a smooth path is generated.

The invention discloses a parallel power supply path searching method and system based on a graph database, and the method comprises the steps: building a power grid power supply association graph model based on structural model connection files and equipment coordinate information of all transformer substations, power transmission lines, feeder lines, loads and distributed power supplies in a region; updating the power grid power supply association graph model according to the latest minute-level section data based on the power grid power supply association graph model; calculating an optimized power supply path by using mixed integer linear programming based on the updated power grid power supply association graph model; based on the optimized power supply path, if a power grid fault is detected, using a Dijkstra algorithm to quickly find an alternative path bypassing a fault area; based on the alternative path bypassing the fault area, deep analysis is carried out through a station power supply source searching algorithm, and a power supply path is refined; a graph parallel computing technology is utilized to distribute computing tasks to a plurality of nodes for parallel execution, and a path search result is obtained; and determining a power supply path from the source side to the target station according to the path search result and the power supply state, marking the power supply path on the geographic coordinates by using a graph database visualization component, and simultaneously performing minute-level dynamic updating to obtain visual and complete path information. Under the conditions that a large number of distributed power supplies are accessed and the topological structure of the power grid is complex and changeable, rapid and accurate power supply path information can be provided, and valuable support is provided for regulation and control and optimization decision making of a power system.

The invention discloses a wheelchair path planning method based on a grid map and an eye-controlled wheelchair navigation system. Belongs to the technical field of intelligent path planning and particularly relates to the technical field of wheelchair path planning based on a grid map. The method overcomes the defects that according to an existing common path planning method, when the road condition is complex, the path planning quality is low, and the path is too close to an obstacle. An original cost function is formed by respectively weighting existing path cost, heuristic cost and safety cost, the method provides a dynamic weighting strategy, and comprises the following steps: firstly, weighting a w2 heuristic cost weight according to a current distance; specifically, in the initial stage of searching, we hope to converge to the position of a target point as quickly as possible, namely, increase the weight value of w2, and decrease the value of w2 as the w2 approaches to the target point, and at the moment, better path quality is required to reach the target point. In addition, the value of w3 is dynamically adjusted according to the number of obstacles in a certain range where the current node is located, so that a safer path is exchanged in a complex environment.

The invention relates to the technical field of path planning, in particular to a global path planning method for a robot in a complex static environment, and the method comprises the steps: constructing a static environment model; setting parameters; initializing a population position; calculating the fitness; adjusting the position of the sparrow to avoid static obstacles; after each iteration, carrying out local optimization by using cubic spline interpolation; and repeating the steps until the maximum number of iterations or path convergence is achieved, and outputting the optimal path. Compared with the prior art, the method has the advantages that the nonlinear attenuation factor mu is introduced into the discoverer algorithm, so that the discoverer can perform wide exploration in the early stage and can perform accurate exploration on the developed area in the later stage, the early-stage search speed of the discoverer is increased, and the convergence speed is increased; and different search paths are determined according to the alarm value, and large-range search is performed when the alarm value is low, so that the optimal path is quickly searched, and the convergence speed is further improved.

The invention discloses a multi-unmanned aerial vehicle power transmission line inspection autonomous path planning method. The method comprises the steps of performing digital processing on a power transmission line map; constructing a multi-objective optimization path planning model, and considering a plurality of optimization objectives of inspection time, inspection efficiency and unmanned aerial vehicle load balance while minimizing a total motion path; smoothing the initial path output by the path planning model by adopting an improved Dubins curve, and generating a continuous smooth path conforming to the flight ability of the unmanned aerial vehicle through track point sorting and dynamic adjustment of the maximum curvature of the Dubins curve; and performing unmanned aerial vehicle deployment optimization on the smoothed path by using an improved bait bird foraging algorithm. According to the method, the efficiency and the accuracy of multi-unmanned aerial vehicle path planning are remarkably improved, the smoothness and the stability of the path are optimized, the cooperative inspection efficiency of the multi-unmanned aerial vehicle is enhanced, the energy consumption and the cost are reduced, and safe and stable operation of the power transmission line is guaranteed.

The invention relates to the technical field of artificial intelligence and natural languages, in particular to a long document question and answer generation method based on a large language model.The method comprises the steps that approximate nearest neighbor search is conducted through an HNSW index, N most relevant triads are retrieved, and for each position in the retrieved triads, a fixed-size window such as [-M, M] is obtained; all (k, v) pairs within a range of 1, 2, 3, 4, 5, 6, 7, 7, 8, 8, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, a new K'matrix and a new V 'matrix are formed by splicing (k, v) pairs obtained by retrieval and expansion with K and V matrixes of a current input sequence, attention is calculated by using a standard zoom dot product attention mechanism, and only vectors corresponding to an input problem sequence are reserved, so that the global view is ensured, unnecessary calculation is avoided, and the accuracy of the attention calculation is improved. Therefore, the serious problems of calculation efficiency and memory use during long sequence processing are solved.

The invention discloses a memory management method for reducing fragmentation and an operating system, and belongs to the technical field of computer storage. Memory management is performed through a memory pool, a logic space is created for mapping with the memory pool, a page table is utilized to record a mapping relation between memory pages of the memory pool and memory pages of the logic space, and a memory recovery table records memory information supporting automatic recovery and the total memory capacity in the memory recovery table according to an application sequence. Memory allocation, memory access, memory release and memory recovery are all carried out through a logic space. According to the invention, the rapid allocation of the memory is realized, and the fragmentation of the memory is reduced.

The invention discloses a demand side flexible resource participation multi-objective optimization modeling method and a related device. The method comprises the following steps: determining initial data including wind power generation, thermal power generating units, electric vehicle disordered charging, distributed electric heating and user side energy storage; an objective function of the multi-objective optimization model is determined, boundary conditions of the multi-objective optimization model are designed according to the objective function and the basic data, and the multi-objective optimization model is constructed; designing a conventional thermal power adjustment scene and a typical demand side flexible resource adjustment scene based on the auxiliary service market; based on a conventional thermal power adjustment scene and a typical demand side flexible resource adjustment scene, the obtained initial data are utilized, an improved particle swarm optimization algorithm is adopted to solve a multi-target optimization model, and an optimization solving result is output. According to the method, the regulation capability of the power grid is optimized, the wind curtailment rate and the wind curtailment cost are effectively reduced, the absorption capability of new energy is improved, the problem of power curtailment caused by blockage of a delivery section is reduced, and the method has important significance in promoting sustainable development of the new energy.

The invention relates to the technical field of unmanned aerial vehicles, and particularly provides a cluster unmanned aerial vehicle region traversal search method, which utilizes an unmanned aerial vehicle as a communication relay node and other unmanned aerial vehicles as search machines to improve the task deployment and search speed of a cluster, and adopts distributed mode prediction control to make an online decision on the search process of the unmanned aerial vehicles. According to the method, the search decision of the unmanned aerial vehicle better meets task requirements, the unmanned aerial vehicle quickly discovers a target, deals with various constraints and unforeseen circumstances and makes an adaptive decision, and the robustness of a cluster in executing a search task in a complex, dynamic and unknown environment is improved.

The invention discloses a robot safety monitoring system which comprises a sensor module, a sensing module, a positioning module, a hardware monitoring module, a software monitoring module, a planning control module, an early warning module, a record feedback module and a robot control system. Identifying and analyzing objects and obstacles in the working environment; and the planning control module is used for generating a driving path and performing real-time adjustment according to the sensing and positioning information. According to the robot safety monitoring system provided by the invention, the laser radar, the camera and other multi-element sensors are fused, a high-precision environment map is constructed, and autonomous navigation of the robot can be realized in combination with the positioning module; the early warning module performs graded early warning according to abnormal conditions to ensure operation safety; the hardware monitoring module and the software monitoring module work cooperatively to guarantee stable operation of the system.

In the extremely cold object storage system, the access method and the management method, an extremely cold object index layer comprises a storage device, and the storage device is used for storing an extremely cold object index; the cold object data layer comprises a storage device C1 used for storing extremely cold object data; the extremely cold object metadata layer comprises a storage device D1 used for storing extremely cold object metadata; the extremely cold object index comprises the name or number of the extremely cold object and the position information of the storage device D1 stored by the extremely cold object metadata; the extremely cold object metadata comprises extremely cold object names or serial numbers and position information of storage equipment C1 of extremely cold object data; the storage device C1 is a tape storage device, and the storage device D1 is a tape storage device; finding the position information of the storage device D1 stored by the extremely cold object metadata in the extremely cold object index according to the name or the number; after obtaining the extremely cold object metadata, obtaining position information of a storage device D1 of the extremely cold object data; and extremely cold object data is obtained from the storage device D1.

The invention provides a key value storage system. The key value storage system comprises a data pre-reading module, a cache management module and an index adjustment module. And the data pre-reading module is used for predicting the data blocks which can be accessed in the preset time period and loading the data blocks to the cache in advance so as to reduce reading delay and disk I / O (Input / Output) operation, so that the reading efficiency is improved. The cache management module is used for monitoring the running state of the key value storage system in real time and dynamically adjusting the cache space based on the running state, it is ensured that the cache meets the high hit rate, memory use can be reasonably controlled, resource waste is avoided, and the reading performance and the response speed are further improved. The index adjusting module is used for dynamically adjusting an index structure according to the access frequency of the key value pair, reducing disk I / O operation during reading, improving the searching speed and reducing the searching cost. Therefore, according to the key value storage system, the negative influence caused by read amplification can be reduced, the response speed of the system is improved, and faster and more efficient data access experience is provided for a user.

The embodiments of the present application disclose a path planning method, device, equipment and storage medium to solve the problem that the traditional path planning method cannot consider the influence of special areas and traverses many nodes and turning points during path planning. The present application improves the A* algorithm based on the regular hexagonal grid map modeling, optimizes the calculation method of the mobile cost, so that the environmental information can be fully considered during the search, the mobile cost of the node can be flexibly set, and the optimality of the planned path can be maintained. By introducing an adaptive heuristic, the number of traversed nodes is reduced and the search speed is increased; by designing a turning penalty measure, the number of turning points in the planned path is reduced to avoid unnecessary turns. Therefore, the present application can not only ensure the accuracy of path planning in complex scenarios, but also realize the flexible setting of the mobile cost, improve the search efficiency, reduce the number of turns, and meet the requirements of high-quality path planning of robots.

The application discloses a multimedia resource search method, related device and vehicle, and relates to the field of human-computer interaction. When a user performs resource search, it is first judged whether there is a running multimedia application. When it is judged that there is a running multimedia application, it is further judged whether to perform search in the multimedia application according to the obtained search voice content. When it is judged that search is performed in the multimedia application, the multimedia application is directly called to perform in-application search according to the search voice content, so that the searched multimedia resource is more consistent with the user's expectation, the user's use experience is improved, the search speed is improved due to the direct search in the application, and the user's use experience is further improved. When it is judged that there is no running multimedia application, search can be performed in a media resource library outside the multimedia application, so that the user's resource search demand is met, and the user experience is ensured.

Embodiments of the invention provide a map matching method and apparatus, and a storage medium. The method comprises the steps of obtaining a to-be-matched map; inputting a to-be-matched map into the image segmentation model to obtain a plurality of image segmentation mask images; determining a plurality of target mask images greater than a threshold value from the plurality of image segmentation mask images; performing positive and negative binarization processing on each target mask image to obtain a positive and negative binarized mask image; aiming at each standard map in a preset standard map library, obtaining a standard map mask map and the size of the standard map mask map; generating a new positive and negative binarization mask picture of each target mask picture according to the standard map mask picture and the size; based on an image matching model, respectively matching the generated new positive and negative binarization mask images with a standard map mask image, and respectively outputting collocation point pairs of each target mask image and the standard map mask image; and determining a standard map matched with the to-be-matched map according to the maximum matching point pairs. According to the invention, intelligent map matching can be realized.

The invention relates to a log association analysis method based on an FP-Growth algorithm, and the method is characterized in that the method comprises the steps: S1, obtaining an original log, and generating an event; s2, based on a given division standard, dividing log events, and constructing a transaction data set; s3, for the transaction data set, mining a frequent item set by adopting an FP-Growth algorithm in a frequent event mode of the FP-Growth algorithm; and S4, generating a strong association event rule. According to the method, the FP-Growth algorithm is adopted to mine the association relationship of the log events, and the FP-tree is constructed to store the frequent item set information, so that the memory occupation can be reduced, and the mining process of the frequent pattern is simplified. The invention further provides a device for implementing the log association analysis method based on the FP-Growth algorithm and a storage medium.

The present invention discloses a search and scanning control method for an optoelectronic turret to control a target area identified by a radar, and belongs to automatic control technology. The scanning search mode and the search range are automatically calculated according to the field of view size of the turret sensor and the possible target range given by the radar, so as to realize fully automatic search and scanning. During the regional search and scanning process, the optoelectronic turret will switch the search and scanning mode according to the target area range given by the radar and the current sensor field of view angle information range, select a suitable scanning mode from four scanning modes according to the size of the target area range and the field of view range, and calculate the scanning parameters, so as to ensure that the entire target area range can be scanned even when the target area is constantly changing. Actual experiments have proved that this search and scanning control method realizes fully automatic search and scanning without human intervention, and can find the target even when the range given by the radar is not accurate enough, which greatly improves efficiency and reduces the probability of missing the target.

This invention discloses a data storage method and system based on a scalable multidimensional learning index structure, belonging to the field of computer data storage. This system fully utilizes storage characteristics to organize data into memory and disk, significantly reducing disk I / O with minimal memory usage, accelerating data search speed, and supporting data expansion. This invention can be applied to multidimensional data indexing, typically represented by geospatial data, meeting the need for rapid multidimensional key searching in the high-dimensional key space of such data.

The application discloses a model management method and device, the method comprises the following steps: obtaining a first model, editing the first model based on a first input, and saving the edited first model as a first template, the first input is used for adding running dependency files corresponding to N first modules respectively, the running dependency files store parameter information of at least one simulink model, at least one second model is created according to the first template, and nodes corresponding to modules at all levels of the second model are displayed on a model configuration interface of the second model, in response to a second input to a node corresponding to a target module in the N first modules, the target module is updated based on parameter information of a target simulink model, and the target simulink model is a simulink model matched with the target module. According to the embodiment of the application, the difficulty of finding the to-be-replaced module in the model can be reduced, and the finding and replacing efficiency can be improved.

The application provides a battery group arrangement equalization strategy optimization method based on a QTA-TLBO double-layer multi-objective algorithm, which comprises the following steps: establishing an upper-layer battery group optimization model comprising an upper-layer optimization objective function and an upper-layer optimization constraint condition, and obtaining an optimal single cell arrangement structure of the battery group; establishing a lower-layer battery group optimization energy scheduling model comprising a lower-layer optimization objective function and a lower-layer optimization constraint condition, and obtaining an optimal equalization strategy of the battery group; establishing a coupling optimization model between the upper-layer battery group single cell arrangement structure and the lower-layer battery group equalization strategy; and using the proposed quantum tunneling annealing-teaching and learning double-layer multi-objective optimization algorithm to optimize and solve the model, so as to obtain an optimal grouping scheme of the battery group and an optimal equalization strategy of each cycle. Through the application, the reasonable arrangement of the single cells of the battery group can be obtained, the long-term reliable operation of the battery group can be realized, and the energy scheduling strategy of the battery group between the single cells can be planned, so that the maximization of battery energy utilization is realized.

The invention discloses a space station gyroscope orthogonality judgment method and device, and belongs to the technical field of data processing. The method comprises the following steps: pre-constructing a plurality of configuration sets; determining LTU components contained in each configuration set to form a plurality of component sets in one-to-one correspondence with the plurality of configuration sets; when gyro orthogonality judgment is carried out on the current configuration, determining a target configuration set to which the current configuration belongs according to the plurality of configuration sets; determining all LTU components contained in a corresponding target component set according to the target configuration set; whether a target LTU assembly exists in all LTU assemblies contained in the target assembly set or not is determined, three gyroscopes in the target LTU assembly participate in fault diagnosis and have no faults, and if yes, it is determined that the three gyroscopes in the target LTU assembly have orthogonality. According to the invention, the operand can be reduced and the searching speed can be improved.

The present application discloses a robot positioning method, a chip and a laser robot based on a multi-layer grid map. The robot positioning method comprises: the robot uses a laser sensor to collect laser point data; the robot obtains the multi-layer grid map layer by layer in order of resolution from low to high; the robot traverses each candidate solution in the current layer grid map; when the robot determines that the currently traversed candidate solution is a feasible solution, the robot controls the laser point data to stop adding the multiple occupancy probability values ​​corresponding to the currently traversed candidate solution, and then sets the determined feasible solution as the candidate solution of the next layer grid map based on the ratio between the resolution of the next layer grid map and the resolution of the current layer grid map, and then recursively determines the feasible solution in the candidate solutions of each layer grid map in order of resolution from low to high, until the optimal feasible solution is determined, and the optimal feasible solution is set as the repositioning result of the robot.

The invention discloses an efficient Hash algorithm supporting progressive re-Hash, in order to make full use of a first-level cache, a plurality of key values with the same Hash value are placed in a cache line, access to a memory is reduced, the utilization rate of a CPU (central processing unit) is improved, the progressive re-Hash algorithm does not perform re-Hash operation at one time any more, and the Hash efficiency is improved. Instead, the re-hash work is dispersed into a plurality of data insertion steps, and the re-hash work of only one hash bucket is performed each time, so that the one-time re-hash work is smoothed, the whole process can be smoothed by adopting a progressive hash algorithm, the delay of a transaction system is controlled within an acceptable range, an alarm is not triggered, and the transaction efficiency is improved. The method has no obvious influence on the stable operation of the whole transaction system, improves the search speed and insertion speed of the whole transaction system, reduces the delay of each order, efficiently utilizes the memory, and provides better use experience for users.