7results about How to "Avoid overestimation" patented technology

The invention relates to an intelligent control method for a high-speed aircraft, and the method comprises the steps: building an aircraft control architecture, carrying out the decision-making of a neural network model, training a decision-making neural network, and carrying out the real-time compensation of a control instruction according to the flight state of the aircraft, so as to achieve the control of a variable-configuration high-speed aircraft. An aircraft controller adopts a model + learning double-layer architecture, a decision neural network is designed to be of a Double Network structure, overload errors of pitching, yawing and rolling channels are considered in a reward function design, training of a decision neural network model is based on a double-delay deterministic strategy gradient algorithm, a training process is optimized by adopting a prior experience playback technology and a target strategy smoothing mechanism, and a decision neural network model is obtained. The control system can achieve real-time intelligent adjustment of control instructions, overload quantities and attitude angles of three channels of the aircraft and corresponding rudder deflection instructions output by the controller at the last moment are obtained from the environment, the intelligent body outputs control instructions of pitching, yawing and rolling channels, and therefore intelligent control over the variable-configuration high-speed aircraft is achieved.

The invention provides a dual-arm robot cooperative control method based on multi-agent reinforcement learning. The dual-arm robot cooperative control method comprises the steps that system control of a dual-arm robot is converted into a Markov decision problem; establishing a double-arm robot simulation environment, and designing rounds and tasks according to reinforcement learning problems; and constructing a multi-agent reinforcement learning model, and training the simulation model through the multi-agent reinforcement learning model. According to the method, the two mechanical arms are regarded as independent agents for multi-agent reinforcement learning training, the actual movement condition is better met, and the success rate of a cooperative task is effectively increased; when the actions are selected, the actions are scored through the Critic network, then the scores of the two actions are compared, the action with the higher score is selected for execution, the problem of local optimum of the actions occurring in the training process and the problem of over-estimation or underestimation occurring when the Critic network judges the value of the actions are solved, and the stability and accuracy of cooperative control tasks of the two-arm robot are improved.

The application relates to the technical field of internet big data analysis, and discloses an article reach person-time estimation algorithm based on a propagation path, which comprises the following steps: step 1, collecting a propagation event log; step 2, data cleaning and filtering; step 3, constructing a propagation path tree and obtaining node attributes; step 4, estimating independent exposure person-time brought by sharing; step 5, merging and calculating total article reach person-time; and step 6, correction and periodic parameter updating. The article reach person-time estimation algorithm based on the propagation path sets effective exposure rate parameters for different sharing channels, for example, based on platform sampling statistics, the effective exposure rate of a friend circle share is 0.12, the effective exposure rate of a microblog share is 0.18, and the effective exposure rate of other channels is 0.10, so that the differences in actual visible probability of information flow of various channels can be reflected, the single sharer basic exposure amount is calculated on this basis, rough processing methods of adopting the same exposure proportion for different channels are avoided, and the exposure estimation is more in line with the actual distribution mechanism of the platform.

This invention provides a user incentive and points management method and system for carbon emission reduction behavior, relating to the field of carbon emission reduction behavior management technology. The method includes: acquiring the user's original trajectory information in the target travel behavior, and simultaneously collecting corresponding environmental feature data and equipment motion status data; performing segmentation processing; performing anomaly identification processing on each trajectory segment; performing segment consistency reconstruction processing on trajectory segments containing anomaly segment markers; performing travel mileage calculation processing to obtain the user's actual travel mileage in the target travel behavior, and generating corresponding carbon emission reduction data; performing points generation processing based on the carbon emission reduction data, converting the carbon emission reduction data into user points records; and mapping user points to a preset incentive resource pool to generate incentive resource availability data. This invention improves the autonomy and accuracy of carbon emission reduction behavior management.

The application provides a rice field evapotranspiration prediction method and device based on public weather forecast and crop coefficient method, and relates to the technical field of agricultural information processing. The method comprises the following steps: obtaining the public weather forecast information of a target area, calculating a reference crop evapotranspiration dynamic prediction value; constructing an average crop coefficient model coupled with a density coefficient and a plant growth temperature constraint, and calculating an average crop coefficient, wherein the density coefficient is calculated by a canopy coverage and a crop height based on a nonlinear regression model of cumulative heat time; and calculating an actual evapotranspiration based on the reference crop evapotranspiration dynamic prediction value and the average crop coefficient. The method can break through the technical bottleneck of precise evapotranspiration prediction under the condition of limited meteorological information, realize daily dynamic evapotranspiration estimation based on easily obtained data such as air temperature and weather type, provide a scientific basis for efficient use and intelligent decision of farmland water resources, and promote the intelligent transformation of agriculture and water conservancy.

The present application belongs to the technical field of computer aided engineering simulation, and relates to a kind of offshore floating photovoltaic support structure wind resistance stability simulation analysis method, comprising: according to the relative wind direction angle row determination shelter, determine effective wind area;With the height difference between pressure center and resistance center as independent force arm, combined with effective wind area, get each row independent wind inclination torque;Sum up the wind inclination torque of all support rows, photovoltaic equipment and floating body structure to get total wind inclination torque;With the deviation of initial inclination angle restoring torque and total wind inclination torque driving iteration, update the shelter interval boundary and pressure center height varying with inclination angle, execute shelter determination and calculation cyclically, until the deviation converges or the iteration number reaches the upper limit, and output the final inclination angle as the wind resistance stability equilibrium floating state.The present application fully considers the dynamic change of aerodynamic shelter caused by floating body inclination, realizes wind inclination and stability coupling iteration, truly reflects the equilibrium floating state, and improves the simulation accuracy.

The application provides a large model intelligent quality inspection method, device and equipment and medium, relating to the technical field of artificial intelligence, which comprises: performing a preprocessing operation on a multi-modal original data set formed by collecting voice data and text data around the same quality inspection object to generate structured conversation data, then performing rule matching processing, and forming a corresponding rule determination result when the rule is triggered. When the rule determination result is not formed, the structured conversation data is subjected to emotion state recognition and sensitive expression detection to output corresponding detection results and convert them into a structured semantic feature vector; the structured conversation data and the structured semantic feature vector are subjected to semantic analysis to output a semantic determination result, and the rule determination result and the detection result are combined to perform fusion decision processing and generate a quality inspection result through dynamic weighted calculation. The application can effectively fuse multi-source data and output stable and consistent quality inspection results, thereby improving the accuracy of quality inspection.