385results about How to "Avoid divergence" patented technology

The invention is a charge state evaluation method and system of a power lithium ion battery. According to the method, step one is to establish a circuit model of an equivalent battery. Charging and discharging and standing experiments are performed on the battery, and timing sampling is performed so that a voltage time curve is obtained. Model parameters are identified via a formula so that a non-linear relation between an open-circuit voltage OCV and an SoC is obtained. Step two is to obtain an optical estimation value of the SoC by matrixes of state prediction, prediction error variance, filtering gain, state estimation, estimation error variance, etc., according to Kalman algorithm. According to the system, an analog / digital converter, a program storage device, a programmable storage device, a timer and a displayer are respectively connected with a microprocessor. A current sensor and a voltage sensor are respectively connected in a circuit formed by connecting the battery to be tested and a load, and outputs of the current sensor and the voltage sensor are accessed into the analog / digital converter. The programmable storage device stores battery model parameters obtained by the experiments. The program storage device stores estimation program of the method. Estimation precision of the SoC can reach 1%, and the charge state evaluation method and system is more stable; besides, the system provides estimation values of the SoC in real time.

The invention provides a distributed parallel training method and system for a neural network acoustic model. The distributed parallel training method comprises the steps that 101) each group of training data is inputted into a client respectively; 102) each of the client receives the inputted training data, and transmits parameters between a plurality of GPU and a first CPU, which are arranged in the client, by adopting a bus, wherein the parameters comprise the model weight and the gradient; each GPU calculates the gradient based on the inputted model weight parameter and inputs the calculated gradient into the first CPU; the first CPU updates a model copy in the corresponding client by using the gradient uploaded by the GPUs, transmits an updated weight parameter back to each GPU so as to be used for carry out gradient calculation again, and meanwhile, the first CPU accumulates the gradient inputted by each GPU and updates a model in a parameter server according to a accumulation result; and 103) latest gradient information acquired by accumulation carried out by the CPU of each client is inputted into the server by adopting network resources, and then a neural network acoustic model stored in the server is updated.

The invention discloses a multi-system dynamic PPP resolving method based on robust self-adaption Kalman smoothing. The method includes the steps that receiving machine outline coordinates and receiving machine clock bias of all systems are solved through selecting-weight-iteration pseudo-range single-point positioning, and accordingly all positioning error correction values are calculated according to an error correction model in combination with the satellite precise ephemeris and satellite precise clock bias; strict data quality control is conducted on observation data. Due to the fact that dynamic PPP accuracy is easily affected by undetected small cycle slips or the gross error and the like, an observation equation weight matrix is adjusted according to the observation value residual vectors, and the undetected small cycle slips or the gross error and other influence factors are removed; self-adaption factors are determined according to the state predictive information, and thus the influence on parameter estimation of the predictive information is controlled. By means of the method, when multi-system dynamic PPP is conducted through a single receiving machine, the feature that the number of multi-system satellites is increased greatly, on the basis that the stability of the satellite structure is guaranteed, the influence of the gross error is weaken effectively, the dynamic noise abnormity in dynamic positioning is improved, and finally the high-precision and high-stability multi-system dynamic PPP result is achieved.

The invention provides a GNSS / INS / vehicle integrated navigation method for agricultural machinery operation in the technical field of navigation. The method comprises the following steps: S1, inertial navigation calculation is carried out after correction of zero offset of an inertial measurement unit (IMU), and a state transition matrix and a system covariance matrix are calculated through Kalman filtering; S2, whether GNSS information is updated is judged; if the GNSS information is updated, the GNSS information is extrapolated, filtering estimation is carried out by taking the position and speed differences between INS and GNSS as observations, the dynamic condition of the carrier is judged, correction platform error and IMU zero offset error are fed back, and the method returns to S1 for loop execution; or, the current information is the position, speed and attitude information of the carrier; and S3, whether vehicle information is updated is judged; if the vehicle information is updated, the vehicle information is extrapolated, filtering estimation is carried out by taking the speed difference between INS and the vehicle as an observation, and correction platform error and IMU zero offset error are fed back; or, the current information is the position, speed and attitude information of the carrier. The method has high navigation accuracy, and can achieve high navigation accuracy with a low-cost IMU.

The invention discloses a real-time navigation system and a real-time navigation method for an underwater structure detection robot. The navigation system comprises a magnetic compass, a gyroscope, an accelerometer, a depth meter and a navigation microprocessor, wherein the magnetic compass, the gyroscope, the accelerometer and the depth meter are used for respectively collecting magnetic field intensity, an angular speed, a linear speed and submerged depth data and transmitting the magnetic field intensity, the angular speed, the linear speed and the submerged depth data to the navigation microprocessor; the navigation microprocessor is used for calculating attitude and position of the underwater robot according to the collected data. The navigation method comprises an attitude algorithm, a speed algorithm and a depth algorithm; according to the attitude algorithm, a complementary filtering method, a quaternion gradient descent method and a Kalman algorithm are combined for obtaining an attitude matrix and an attitude angle; the speed algorithm is used for calculating the speed and the position of the robot by using a three-order upwind scheme with rotary compensation; the depth algorithm is used for processing the data of the depth meter by using a moving average filter algorithm so as to obtain the submerged depth. By virtue of the real-time navigation system for the underwater structure detection robot and the method thereof, the navigation cost is reduced and a relatively good navigation precision is achieved.

A method for active noise cancellation in a volume generates a mathematical model of a noise process to be cancelled. Using the mathematical model, a noise signal in a next sample period is predicted from a measured noise signal. The predicted signal is inverted and applied to the volume. Destructive interference of the noise signal and the inverted signal cancels the noise in the volume.

A latitude unknown self-aligning method of a strapdown inertial navigation system under the dynamic interference condition includes the steps that firstly, a geometrical analytic formula is built by means of the characteristic that the projection of gravity acceleration is unchanged in an inertial coordinate system, the gravity acceleration is subjected to integration to obtain speed information, and the latitude value of the position where a carrier is located is calculated according to the speed information; secondly, on the basis of the double-vector altitude determination principle and by means of the characteristic that the gravity acceleration of an inertial system includes north orientation information, rough solution of an initial attitude matrix under the inertial system is achieved; finally, on the basis that coarse alignment is finished, a precise alignment error model under the latitude unknown dynamic interference condition is built according to a speed error equation, a misalignment angle equation and a latitude error equation, the latitude error angle and the misalignment angle of the carrier are calculated by means of the self-adaptive filtering method based on information, the latitude value is compensated with the latitude error angle, the strapdown altitude matrix is corrected according to the misalignment angle, and high-precision quick self-aligning of the strapdown inertial navigation system is achieved.

The invention discloses a CKF filtering-based vehicle dynamic model auxiliary inertial navigation combined navigation method. The CKF filtering-based vehicle dynamic model auxiliary inertial navigation combined navigation method comprises the following steps: calculating posture, speed and position of a vehicle according to angle increment and specific force output by a micro-inertia device and by an inertial navigation numerical value updating algorithm; establishing a three-degree-of-freedom vehicle dynamic model, and calculating the speed of a carrier by taking a steering wheel angle and a longitudinal force as control input quantity and by a fourth order Ronge-Kutta method in real time; designing a CKF filter by taking an inertial navigation equation as a state equation and speed difference between a dynamical model and inertial navigation calculation to perform state estimation on a combined navigation system; performing output correction on strapdown inertial navigation calculation result by the position the speed and the posture error obtained by CKF estimation, and performing feedback correction on the inertial navigation through peg-top and adding error. The method aims at the problems that the inertial navigation error is accumulated along with time and navigation precision cannot be maintained for a long time, and the accuracy and the reliability of a vehicle navigation system can be improved.

The invention discloses an internal temperature estimation method for a lithium battery based on a discrete sliding-mode observer. The method comprises the following steps that a lithium battery discrete state space model for internal temperature estimation is established; a working voltage, a working current, a surface temperature and an external environment temperature of the lithium battery under a working condition are collected in real time; parameters of the lithium battery discrete state space model are identified online; and the discrete sliding-mode observer is constructed to estimate the internal temperature of the lithium battery in real time. The method has a good temperature estimation effect, can ensure the convergence of the observer strictly, and is highly robust for modeling errors, model parameter perturbation and external disturbance of the lithium battery thermal model.

The invention discloses a strapdown inertial navigation system(SINS) / global navigation satellite system (GNSS) combined navigation filter system comprising a primary binding module, an earth parameter calculation module, an initial alignment module, a strapdown inertial navigation calculation module, an improved filter algorithm parameter calculation module, a normal-mode combined navigation filtering module, a fault-mode combined navigation filtering module and a feedback correction output module. The improved filter algorithm parameter calculation module calculates a sate transition matrix, a system drive noise covariance matrix, and a measurement noise covariance matrix that are needed by an improved kalman filtering algorithem based on data provided by an SINS and a GNSS and transmits the information to the normal-mode combined navigation filtering module and the fault-mode combined navigation filtering module to carry out filtering calculation according to the result and a status flag. According to the invention, the combined navigation filtering algorithm is modularized, thereby obviously improving precision of the combined navigation and accelerating the convergence speed of the combined navigation filtering algorithm; and the kalman filtering divergence can be effectively inhibited.

The invention relates to an echo cancellation method and system for a speech recognition front end. The echo cancellation method mainly comprises the steps of time-frequency domain signal transformation, echo signal and residual signal estimation, sub-band signal power spectrum and cross correlation coefficient calculation, nonlinear processing based on a gain function of cross correlation coefficients, frequency-time domain transformation and the like to finally output signals without echoes. The single-channel echo cancellation method is adopted for processing frequency-domain echoes under avehicle-mounted condition or in other application scenes, and has the advantages of small calculation amount and high convergence speed.

The ion implanting apparatus according to this invention includes: an ion source for producing the ion beam 20 including desired ion species and being shaped in a sheet with a width longer than a narrow width of a substrate 82, a mass separating magnet 36 for selectively deriving the desired ion species by bending the ion beam in a direction perpendicular to a sheet face thereof, a separating slit 72 for selectively making the desired ion species pass through by cooperating with the mass separating magnet 36, and a substrate drive device 86 for reciprocatedly driving the substrate 82 in a direction substantially perpendicular to the sheet face 20 s of the ion beam 20 within an irradiating area of the ion beam 20 which has passed through a separating slit 72.

The invention provides a multi-medium complex-environment high-precision vision / inertia combination navigation method which is suitable for combination navigation of an underwater vehicle under a complex electromagnetic environment in a transparent medium. By means of the underwater vehicle for simulating rendezvous and docking of a spaceship and a satellite in a space micro-gravity environment, important test data and test verification are provided for the rendezvous and docking. The whole experiment system is used on ground. By means of buoyancy of water and adjustment of the direction of the ground electromagnetic field and the intensity of the electromagnetic field, the underwater vehicle is in a neutral buoyancy level, thereby satisfying an equivalent test condition of coincidence of space and ground. The multi-medium complex-environment high-precision vision / inertia combination navigation method can be used for controlling the underwater vehicle for simulating rendezvous and docking in space, so that the attitude, the speed, the position and the acceleration information of the underwater vehicle can be measured accurately.

The invention relates to a strong-tracking self-adaptive-square-root-quadrature-Kalman-filter (SQKF)-based state of charge (SOC) estimation method of an emergency lamp battery. The method comprises: establishing a second-order RC equivalent model of an emergency lamp battery; on the basis of the battery equivalent model, establishing a corresponding discrete state space model equation; and carrying out filtering estimation on the SOC value of the battery by the strong-tracking self-adaptive SQKF method. Compared with the existing SOC estimation method under the Kalman filtering framework, the provided method has high estimation precision. Meanwhile, with introduction of a time-varying fading factor and an on-line estimation system noise variance, a filter divergence problem caused by unknown time varying of the noise statistical characteristic during battery system modeling can be effectively suppressed.

The present invention relates to a color filter using surface plasmons, a liquid crystal display (LCD) device and a method for fabricating the same. Through forming a transmissive pattern which is composed of a plurality of sub-wavelength holes having a period on a metal layer for selectively transmitting the light with a certain wavelength, the transmittance of a LC panel can be improved and a whole process can be simplified. There is provided a color filter using surface plasmons, comprising: the metal layer; the transmissive pattern formed in the metal layer, and consisting of a plurality of sub-wavelength holes having a period; and a plurality of non-penetrating grooves formed at a region encompassing the plurality of holes, wherein a desired color is implemented by selectively transmitting light of a specific wavelength by using surface plasmons. A plurality of grooves with the period is formed on an area with a plurality of holes thereby realizing the light beam with extraordinarily narrow radiation angle. Thus the correlation between a wavelength of a main peak of the transmitted light and a visual angle can be improved. Furthermore the transmittance and calibration are improved.

The invention relates to a vector search iterative matching method based on inertia / gravity matching integrated navigation, the vector search iterative matching method is characterized by comprising the following steps of estimating a matched search scope and a current matching performance by an inertial navigation output location; taking a correlative extremal function as a target search to determine an optimal matching transformation; performing constant weighted iteration on the result of matching position to obtain an optimal matching position finally. The vector search iterative matching method provided by the invention serves as a hub for connecting component parts of a gravity matching system; a calculating process from gravity measurement information in real time to a carrier position can be completed; an inertial navigation error range is taken as a matching scope; a global optimum solution can be obtained, and divergence can be restrained; the requirement of an initial position error is low; the error and divergence problems caused by database linear processing of gravity data can be effectively avoided; the efficiency and the property of an algorithm are improved, and the vector search iterative matching method has a strong robustness on an gravity measurement random error and an inertial navigation system variable error.

A method to increase low-energy beam current according to this invention is applied to an irradiation system with ion beam comprising a beam generation source, a mass analysis device, a beam transformer, a deflector for scanning, a beam parallelizing device, an acceleration / deceleration device, and an energy filtering device. The beam transformer comprises a vertically focusing DC quadrupole electromagnet QD and a longitudinally focusing DC quadrupole electromagnet QF. The beam transformer transforms a beam having a circular cross-section or an elliptical or oval cross-section to the beam has an elliptical or oval cross-section that is long in the scan direction in all the region of a beam line after deflection for scanning.

A method of plasma particle simulation capable of preventing solution divergence. A space within a housing chamber of a plasma processing apparatus is divided into a plurality of cells. A weighting factor corresponding to the number of plasma particles represented by a superparticle is set in each of the divided cells. Superparticles are set in each of the divided cells using plasma particles contained in the divided cell and the set weighting factor. The behavior of the superparticles in each of the divided cells is calculated. The weighting factor becomes smaller as the divided cell is located closer to a solid wall surface of the housing chamber.

Provided are method and apparatus for recognizing space according to the movement of an input device.A method of recognizing space according to the movement of an input device, the method comprising: measuring angular velocity data using an angular velocity sensor; measuring acceleration data using an accelerometer; estimating a bias of the angular velocity sensor using the acceleration data; calculating Euler angles between a reference navigational frame and a body frame using the angular velocity data and the acceleration data; and identifying position information of the input device according to the movement of the input device by using the calculated Euler angles.

The invention relates to the field of resisting magnetic interference, in particular to a method for resisting magnetic interference. The method includes that magnetic-field strength changing, magnetic inclination changing and magnetic-field component changing of a geographic system are used at the same to judge whether magnetic-field interference exists or not, so that whether the magnetic-field interference exists or not can be judged more accurately; when a magnetic field has the magnetic-field interference and the magnetic-field interference lasts for a long time, course angle divergence caused by solely using a gyroscope for integral computation of a course angle for a long time is avoided; under this circumstance, magnetometer data are recorded and utilized to correct the course angle, so that the course angle divergence is avoided; when the magnetic field has the magnetic-field interference and the magnetic-field interference lasts for a long time, due to the course angle correcting process, interference of the magnetic-field interference on course angle measuring accuracy can be effectively weakened and controlled, and the measuring accuracy of the course angle is guaranteed.

The invention discloses a carrier tracking loop based on a self-adaptive second-order Kalman filter and a filtering method of the carrier tracking loop. The carrier tracking loop based on the self-adaptive second-order Kalman filter comprises a local carrier generator, an integral clearer, a carrier loop phase discriminator, a loop filter and the self-adaptive second-order Kalman filter. The invention overcomes the defect of tracking loop lock losing caused by divergence of the traditional Kalman filter when priori information is inaccurate. When a system model is incomplete or the priori information is inaccurate, a forecast error covariance matrix is self-adaptively regulated through fading factors, thus the purpose of regulating a gain matrix in real time is achieved, the filter approaches the optimal, and the divergence of the filter can be effectively inhibited. Meanwhile, the carrier tracking method has the advantages of simpleness in calculation of a method of estimating the fading factors, small calculated amount, strong adaptability and reliability improvement of a filtering algorithm.

The embodiment of the present invention provides a monocular vision attitude determination method and system. By using the world coordinates and pixel coordinates of each feature unit as initial information, based on the EPNP algorithm, the initial posture information of the cooperative stereo target is determined, so that the obtained initial posture information can be closer to the obtained real-time posture information. At this time, that initial posture information is taken as an iterative initial value, the real-time attitude information of the cooperative stereo target is determined by SoftPOSIT algorithm, thereby being able to make the real-time posture information more accurate, effectively avoid the SoftPOSIT algorithm diverging or converging to the wrong posture, improve the convergence performance and accuracy of attitude solution, achieve real-time posture measurement, and have a good application prospect.

The invention discloses a robot positioning method and device based on SLAM (Simultaneous Localization And Mapping) of covariance intersection fusion. The method comprises the steps of collecting an image of a surrounding environment through a binocular camera of the robot and correcting the image to obtain the corrected image; processing the image, and further constructing a corresponding map point to obtain a state space model of the robot and an observation model of the robot on a road sign; predicting a predicted value of the robot posture at the current moment according to an optimized estimation value of the robot posture at the previous moment; obtaining an observed quantity at the current moment according to the observation model of the robot on the road sign; performing correctionupdate by use of a CI fusion-based filter method in combination with the observed quantity and the predicted value at the current moment to obtain the optimized estimation value of the robot postureat the current moment. According to the robot positioning method and device based on SLAM of covariance intersection fusion, the covariance intersection fusion algorithm is applied to the SLAM, thereby enhancing the robustness in the linear error, improving the filter precision and improving the accuracy of positioning of the robot in different practical application environments.

The invention belongs to a POS post-processing method, and specifically relates to a POS post-processing method with the combination of closed-loop forward filtering and reverse-smoothing. The invention aims at providing a high-precision position and attitude post-processing method. The method comprises a Kalman filtering step and a reverse R-T-S smoothing step. Horizontal velocity error and height error estimated by forward Kalman filtering is fed back to an input end of an inertial navigation solver in real time through feedback controlling parameters; inertial velocity and position divergences are inhibited, and error model linearity is improved; and reverse R-T-S smoothing is carried out on the basis, such that a POS high-precision position and attitude post-processing result is obtained. The method has the advantages that: POS application requirements are satisfied; and high-precision position and attitude information is provided for a related image processing system (such as radar or a camera) and the like.

The present invention provides a deodorizing type closet pan comprising a pedestal pan and a squatting pan, which arranges bad smelly air exhausting passage to connect with air exhausting pipe and exhausting fan outside of the urinal in the urinal or cushion plate of the pedestal plate, the bad smelly air exhausting passage comprises several air exhausting holes in the urinal or surround of the cushion plate, and air collecting passages surrounding the urinal inside and cushion plate inside; the air exhausting hole and the air collecting passage can be combined together with water flushing hole and water exhausting passage previously in the urinal, which not only exhausts water, but also sucks air. The invention has advantages that negative pressure can be produced to exhaust bad smelly air along the air exhausting passage by activating the air exhausting fan, such that smelly air is prevented to transpire from the urinal.

The vehicle includes a vehicle body having a steering section and a drive section connected together; a steerable wheel which steers the vehicle body; a drive wheel which drives the vehicle body; an inclination actuator apparatus for inclining the steering section or the drive section in a turning direction; a plurality of sensors which detect lateral accelerations; yaw angular velocity detection means; vehicle speed detection means; and a control apparatus which controls the inclination of the vehicle body by controlling the inclination actuator apparatus. The control apparatus performs feedback control on the basis of the lateral accelerations, calculates a link angular velocity predicted value from the derivative value of the yaw angular velocity and the vehicle speed, and controls the inclination of the vehicle body by performing feedforward control while using the calculated link angular velocity predicted value.

The invention discloses a nuclear thermal coupling computing method for a nuclear reactor. A mathematical model for nuclear thermal coupling computation is constructed, and a fixed point equation of power distribution is solved, so that nuclear thermal coupling power distribution is obtained. The computing method comprises the following specific ways: initial power distribution is set to be average distribution; an iterated function containing a relaxing factor lambda is constructed, the power distribution is iterated twice successively, the power change rate of the successively iterated function is computed, and if the power change rate of N grids is smaller than 1, the power distribution can be solved in a successive iteration method; and otherwise, the power distribution is solved with a partially iterated function. A computational formula of a maximum relaxing factor in a partial iteration method is given, and sinusoidal distribution is recommended to be used as the initial power distribution. The nuclear thermal coupling computing method provides convenience for the design and analysis of the reactor.

When a new theme is proposed and a thread is created, a category applicable to the thread is specified. In a menu, the theme and the category are displayed together. The thread is displayed in a form in accordance with the specified category. By specifying the category in accordance with the proposed theme, it is possible to suppress the divergence of comment details made on the thread in advance and limit the direction of comments. Further, on the thread, it becomes easy to display the contents proper for the nature of the proposed theme and efficiently perform an output display. This allows more efficient decision making and information sharing by using a bulletin board system. With this operation, it is possible to provide a bulletin board system in which the direction of comments can be limited in accordance with the category of the theme, such as discussion or question, and a summary of conclusion based on the comment details and an output in accordance with the theme can be efficiently performed.

The invention relates to a high precision real-time contour error estimation method, belongs to the precise high efficiency numerical control processing technology field and particularly relates to the high precision real-time contour error estimation method on the basis of an initial value regeneration Newton iterative method in a numerical control parameter curve interpolation processing process. The method comprises steps that firstly, before each step of iterative calculation, according to present interpolation point parameters and actual cutter location points, an iteration parameter initial value on a parameter curve for Newton iterative calculation perpendicular foot points is calculated on the basis of a first order Taylor series expansion method; secondly, a Newton iterative method is utilized, a single step iteration parameter final value is calculated according to the iteration initial value; thirdly, for avoiding greatly increasing the algorithm calculation time, iteration is ended through double constraints of an iteration termination condition and a largest iteration frequency condition, a perpendicular foot point parameter estimate is acquired, and a contour error vector estimate is further calculated. Through the method, iteration divergence can be effectively avoided, contour error estimation precision is improved, and the method has great signification for guaranteeing contour control precision during parameter curve interpolation processing.