43results about How to "Increase computing load" patented technology

The temperature of a stator coil is measured by a temperature sensor (14), amplified by a stator coil temperature amplifier (21), and transmitted to a vehicle control section (23). Further, motor cooling oil (17) for cooling the outer periphery of the stator cools the stator coil (16) along an end coil section of the stator coil (16). The temperature of the motor cooling oil raised by the stator coil (16) is measured by a temperature sensor (15) and transmitted also to the vehicle control section (23) via a motor cooling oil temperature amplifier (22). The vehicle control section (23) estimates the temperature of a rotor magnet based on a thermal model (relationship between temperature, a heat production amount, and heat resistance) of the motor cooling oil, the stator coil, and the rotor magnet by using the motor cooling oil temperature and the stator coil temperature as input values and sends a control instruction to a motor control section (24).

A feedback cancellation assembly for an electroacoustic communication apparatus may include a signal transmission path for generation and emission of an outgoing sound signal to an external environment through an electrodynamic loudspeaker and a signal reception path comprising a microphone for generation of a microphone input signal corresponding to sound received from the external environment. The signal reception path may generate a digital microphone signal. The outgoing sound signal may be acoustically coupled to the microphone. An electronic feedback cancellation path may be coupled between a tapping node and a summing node to produce a feedback cancellation signal to the summing node.

The invention discloses a spatial non-cooperative target relative pose estimation method based on deep learning, and the method comprises the steps: making a data set of a non-cooperative target through software, and expanding the data set; designing a target detection network based on a convolutional neural network, judging whether the target is a non-cooperative target or not, and detecting a 2Dbounding box around the non-cooperative target; segmenting the non-cooperative target around the 2D bounding box, searching the center of the non-cooperative target based on Kirchhoff, estimating thedepth from the center of the non-cooperative target to the camera, and converting 2D pixel coordinates into 3D coordinates in combination with a depth value; designing a key point extraction network,extracting key points such as corner points and three-axis end points of the non-cooperative target, and performing regression on the key points to obtain a rotation quaternion to represent rotation;and finely tuning the estimated relative pose through iterative optimization. According to the method, the conditions of severe shielding, sensor noise, low estimation precision during object symmetry and the like can be processed, and meanwhile, the real-time speed requirement can be met.

The invention provides a dummy metal filling method. According to the invention, a plurality of dummy metals are filled in an area required to be filled by the dummy metals including first dummy metals and second dummy metals; the filling area of each first dummy metal is larger than that of each second dummy metal; and the second dummy metals are positioned between interconnection lines of the first dummy metals and the area where the dummy metals are positioned. Accordingly, the invention further provides an integrated circuit layout structure. Through the dummy metal filling method provided by the invention, the thickness consistency of an integrated circuit layout is realized, and the capacitance increment between the interconnection lines of an integrated circuit is reduced at the same time, the function of the chip is prevented from being damaged due to the introduction of the dummy metals, the dummy metals in different areas can be filled in the area required to be filled by the dummy metals, the filling number of the dummy metals is reduced, and other calculation data quality is reduced.

The invention provides an active damper which improves the stability of the inverter grid-connected system and can be adjusted in power, the active damper is connected in parallel with a plurality ofinverter grid-connected systems at a common coupling point, comprises a power output circuit and a control circuit, A power output circuit outputs a three-phase grid-connected current, the control circuit detects the three-phase PCC voltage, The harmonic content in the target frequency range is extracted, and the active damper is made to simulate the resistance characteristics in the control bandwidth by the feed-forward compensation unit in the static coordinate system, so as to damp the oscillation caused by the instability in the grid-connected inverter system and improve the stability of the whole system. At the same time, the active power and reactive power of the grid-connected current are controlled directly in the synchronous coordinate system. The active damper can stabilize the system quickly and effectively when the grid-connected system is unstable while the active power and reactive power are controlled, and it also has good damping effect for the case of multiple resonance points simultaneously.

A method for auto-labeling a training image to be used for learning a neural network is provided for achieving high precision. The method includes steps of: an auto-labeling device (a) instructing a meta ROI detection network to generate a feature map and to acquire n current meta ROIs, on the specific training image, grouped according to each of locations of each of the objects; and (b) generating n manipulated images by cropping regions, corresponding to the n current meta ROIs, on the specific training image, instructing an object detection network to output each of n labeled manipulated images having each of bounding boxes for each of the n manipulated images, and generating a labeled specific training image by merging the n labeled manipulated images. The method can be performed by using an online learning, a continual learning, a hyperparameter learning, and a reinforcement learning with policy gradient algorithms.

A method for learning parameters of an object detector based on a CNN adaptable to customer's requirements such as KPI by using a target object estimating network and a target object merging network is provided. The CNN can be redesigned when scales of objects change as a focal length or a resolution changes depending on the KPI. The method includes steps of: a learning device instructing convolutional layers to generate a k-th feature map by applying convolution operations to a k-th manipulated image which corresponds to the (k−1)-th target region on an image; and instructing the target object merging network to merge a first to an n-th object detection information, outputted from an FC layer, and backpropagating losses generated by referring to merged object detection information and its corresponding GT. The method can be useful for multi-camera, SVM (surround view monitor), and the like, as accuracy of 2D bounding boxes improves.

A vehicle control device includes a storage device and a processor. The storage device is configured to store relationship prescription data that prescribe a relationship between a state of a vehicle and an action variable that is a variable related to an operation of an electronic device in the vehicle. The processor is configured to calculate a reward corresponding to the operation of the electronic device. The processor is configured to update the relationship prescription data using, as inputs to updated mapping determined in advance, the state of the vehicle that is based on a detection value that is acquired, a value of the action variable that is used to operate the electronic device, and the reward corresponding to the operation of the electronic device when a computation load on the processor is equal to or less than a predetermined load.

The invention discloses an elevator group control system and method based on a multi-agent competition mode. Each elevator control system is an autonomous agent; each agent which acquires a call signal generates bidding information according to a preset protocol and sends the bidding information to other agents in a collaborative range, each agent carries out bidding for the bidding information, and the agent with the optimal bid provides services for passengers; each agent comprises a user interface unit, a communication unit, a bidding unit and a control unit, wherein the user interface unitis used for carrying out interaction with the passengers; the communication unit is used for carrying out information transfer with other agents in the collaborative range; the bidding unit is used for generating all information in the bidding process and carrying out bidding with other agents in the collaborative range; and the control unit is used for operating a plan and executing the servicesaccording to a call form after bidding. According to the invention, a nonconcentrated control mode which does not depend on additional group control equipment is adopted; by a bidding mechanism, uniform scheduling of multiple elevators is changed into collaborative serving of multiple elevators; reliability of the elevator group control system is promoted; the requirements for software and hardware are reduced.

A processor system having a processor core configured to control an external apparatus in accordance with a control algorithm; a signal acquisition managing device configured to receive state signals provided by the apparatus, perform corresponding actions, and generate corresponding synchronized command signals; and a peripheral module structured to receive the synchronized command signals and generate output signals to be processed by the processor core in accordance with the control algorithm.

A method for learning an object detector based on an R-CNN by using a first to an n-th filter blocks respectively generating a first to an n-th feature maps through convolution operations in sequence, and a k-th to a first upsampling blocks respectively coupled with the first to the n-th filter blocks is provided. The method includes steps of: a learning device instructing the k-th upsampling block to the first upsampling block to generate a (k−1)-st pyramidic feature map to the first pyramidic feature map respectively; instructing an RPN to generate each ROI corresponding to each candidate region, and instructing a pooling layer to generate a feature vector; and learning parameters of the FC layer, the k-th to the first upsampling blocks, and the first to the n-th filter blocks by backpropagating a first loss generated by referring to object class information, object regression information, and their corresponding GTs.

The invention relates to a road slope estimation method, in particular to a slope estimation method based on a Kalman filtering algorithm. The objective of the invention is to solve the problem that when acquiring acceleration sensor data or acquiring slope signals through a GPS and the like, more sensors are used, or when estimating position parameters by acquiring data from an automobile CAN bus, the estimation of the road slope is much more complicate due to own parameters of the automobile, difficulty in decoupling of external resistance change and the time varying features of the road, and the method comprises the following steps that 1) R and Q in Kalman filtering parameters are calibrated offline; 2) acquiring a slope estimation formula; and 3) establishing a slope estimation model.According to the method, only the vehicle speed signal and the acceleration signal in vehicle motion are involved and are easy to obtain, only the acceleration sensor needs to be additionally arranged, the price of the acceleration sensor is low, and cost reduction is facilitated.

The invention discloses a dynamic skip reentry vehicle aiming point adjustment method. The method includes: calculating dimensionless forward speed U1 according to secondary reentry point speed and path angle obtained by forecast calculation; calculating maximum range R1 according to aircraft lift-drag ratio and minimum heeling angle data; calculating minimum range R2 according to overload design parameters in the secondary reentry phase; calculating secondary reentry range capacity level R according to corridor design parameters; calculating dynamic adjustment amount deltaR according to distance to go RTG from the secondary reentry point to the parachute-opening point and the secondary reentry range capacity level R. The secondary reentry range capacity is quickly predicated by means of analysis, linking of the primary reentry phase and the secondary reentry phase is realized, calculated amount for numerical prediction is reduced, and parachute-opening point control accuracy is improved. The method can be conveniently applied to guidance of reentry flight of the aircrafts with low lift-drag ratio at the second cosmic velocity.

The invention discloses an airborne UWB positioning method for unmanned aerial vehicles. A cluster system and a positioning method are included. The method is characterized in that the cluster systemcomprises a host unmanned aerial vehicle, a plurality of slave unmanned aerial vehicles, a host positioning module and slave positioning modules; the host positioning module is composed of three distance measuring modules, the layout of the distance measuring modules is an isosceles right triangle, and the three distance measuring modules are located on the three vertexes of the triangle respectively; and each slave positioning module is composed of one distance measuring module and located in the geometric center of the corresponding unmanned aerial vehicle. According to the invention, the positioning of the relative positions between the host and the slaves is realized by adopting anchor points directly mounted on the host unmanned aerial vehicle and the labels on the slaves, the whole cluster takes the host as a reference, the position of the host determines the positions of the slaves under the condition that the formation is fixed, the accuracy of an operation result is ensured, anchor points do not need to be arranged in front of the cluster, mutual positioning between the host and slave unmanned aerial vehicles can be realized, the anchor point power supply problem is solved, and the positionable range can move along with movement of the host.

The invention discloses a KNL-based testing method. The KNL-based testing method comprises the following steps: acquiring a testing task set by a main node, and distributing a testing task subset in the testing task set into a KNL secondary node; and when the KNL secondary node receives the testing task subset, distributing the testing tasks in the testing task subsets into a progress set, and calculating the testing task in each progress by an ant colony algorithm to obtain the testing result, wherein the progress number in the progress set is less than or equal to the task number of the testing task subset. Thus, after the main node acquires the testing task set, the testing tasks are distributed to a KNL secondary node set, the KNL secondary node in the KNL secondary node set receives the testing task subset, the testing task in each progress is calculated by the ant colony algorithm to obtain the counting result, the testing tasks are processed in parallel through a plurality of KNL servers, the calculation load is improved, and calculation is conducted by the ant colony algorithm, in conclusion, the processing speed of the testing tasks is increased.

The present invention provides a high-dynamic fast / precise acquisition method for an NSCC signal. The method comprises a step of performing a Zoom-FFT operation on each subcarrier in an NSCC digital intermediate frequency signal according to a selected frequency band to be analyzed and obtaining refined spectral information, a step of searching for a frequency domain peak point after the Zoom-FFToperation, reading a peak point corresponding frequency, and determining a frequency range to be updated and monitored with the peak point corresponding frequency as a center according to a Doppler dynamic change range, a step of performing a sliding FFT operation on the frequency range to be updated and monitored to accurately estimate the frequency information of each subcarrier of the NSCC in real time in a sliding recursive mode to complete the update / monitoring of each subcarrier peak of the NSCC, a step of carrying out interpolation estimation based on a sinc function by using the peak point and adjacent peak information, and a step of obtaining the real-time precise Doppler shift value of each subcarrier of the NSCC based on an interpolation result to achieve high-dynamic precise acquisition. According to the method, the accurate estimation of the Doppler shift of each carrier of the NSCC during an acquisition phase can be achieved with a small computational cost.

The invention discloses a road surface obstacle detection method based on detection radar, a vehicle and a storage medium, the method is applied to the vehicle, the vehicle is provided with a plurality of radar antennas forming a preset included angle, and the road surface obstacle detection method comprises the following steps: according to a preset time division scanning strategy, periodically transmitting detection beams in different scanning directions in sequence in the driving direction of the vehicle based on the plurality of radar antennas; and receiving reflection beams reflected back in different scanning directions, and determining the obstacle type of the road obstacle in the driving direction according to the reflection beams. According to the invention, the driving safety of the vehicle using the Doppler radar to carry out road obstacle detection is improved.

A travel control device 70 of a vehicle provided by the present invention is provided with a turning control unit 72 that executes a turning assist control for causing a turning state amount to follow a target turning state amount by causing at least one of a front wheel steering angle adjustment device 10, a rear wheel steering angle adjustment device 30 and a brake device 40 to be driven; and a steering instruction unit 73 that, when steering is not being performed during the turning assist control, instructs a steering device 20 to derive holding torque and apply the holding torque to a steering wheel 21. In the turning assist control when steering is not being performed, the turning control unit 72 causes at least one of the rear wheel steering angle adjustment device 30 and the brake device 40 to be driven, but causes the front wheel steering angle adjustment device 10 not to be driven. In the turning assist control when steering is being performed, the turning control unit 72 causes the front wheel steering angle adjustment device 10 to be driven.

When a judgment result of a cut point judgment unit (16) at a cut point detection unit (1) indicates a cut point, a shot length calculation unit (2) for calculating the shot length of the shot starting at a cut point preceding the aforementioned cut shot by one. By using the shot length calculated by the shot length calculation unit (2) as a judgment material, it is judged whether the shot starting at the cut point preceding by one is an important shot.

The invention belongs to the technical field of geographic information systems, and discloses a shortest time path planning method capable of predicting a dynamic space-time environment, which comprises the following steps of: 1, quantifying various environmental factors influencing the advancing speed; 2, comprehensively evaluating the influence of various environment factors on the advancing speed, and constructing a dynamic space-time environment influence rule expression model capable of expressing the whole space-time environment according to the environment influence expression rule of any space at any moment; and 3, on the basis of the constructed dynamic space-time environment influence rule expression model, evaluating the consumed time of linear movement between any two adjacentpoints, and on the basis, planning a continuous movement track from a starting point to a target point according to the principle of minimum cost by adopting a path search algorithm and taking the time as cost. According to the invention, the path planning result can reflect the dynamic influence in the continuous change process of the environment, and the operation burden is not significantly increased.

Provided is an operating method of a memory controller which comprises receiving first decision data of M bits from a memory device, where M is a natural number; converting the M-bit first decision data into second decision data of N bits, where N is a natural number less than M; and attempting a first decoding using the second decision data.

The invention discloses a computing method applied to a voltage included angle of a power distribution terminal. Voltages are acquired through an acquisition circuit in the power distribution terminal, relevant processing is performed on the voltages subjected to acquisition processing through an arithmetic device in the power distribution terminal to obtain a corresponding operation result, an arc cosine table which is made according to the operation result in advance is stored in a storage of the power distribution terminal, and the voltage included angle corresponding to the operation result is found through automatic retrieval. According to actual measurement, the computing method can work out the required voltage included angle accurately to meet actual application under the premise of enough sample data, the operation is not affected by signal frequency changes, the floating point operation is avoided, and the operation efficiency is improved.

The invention relates to the technical field of train fault detection, in particular to a train fault analysis method and system based on a fusion distance method. The train fault analysis system based on the fusion distance method comprises a working sample acquisition module, a correlation coefficient matrix construction module, a calculation module, a fault alarm threshold establishment module and a fault confirmation detection module, and due to the complexity of train faults, the correlation or independence between sample data sets is fuzzy and uncertain. The independence and the correlation are required to be considered in the calculation process; and the advantage of independence of the Euclidean distance and the advantage of correlation of the Mahalanobis distance can be considered by utilizing the distance fusion algorithm, so that the fault diagnosis is more accurate and reasonable.

The invention discloses an elevator group control system and method based on a multi-agent competition mode. Each elevator control system is an autonomous agent; each agent which acquires a call signal generates bidding information according to a preset protocol and sends the bidding information to other agents in a collaborative range, each agent carries out bidding for the bidding information, and the agent with the optimal bid provides services for passengers; each agent comprises a user interface unit, a communication unit, a bidding unit and a control unit, wherein the user interface unitis used for carrying out interaction with the passengers; the communication unit is used for carrying out information transfer with other agents in the collaborative range; the bidding unit is used for generating all information in the bidding process and carrying out bidding with other agents in the collaborative range; and the control unit is used for operating a plan and executing the servicesaccording to a call form after bidding. According to the invention, a nonconcentrated control mode which does not depend on additional group control equipment is adopted; by a bidding mechanism, uniform scheduling of multiple elevators is changed into collaborative serving of multiple elevators; reliability of the elevator group control system is promoted; the requirements for software and hardware are reduced.

The invention discloses a computing method applied to a voltage included angle of a power distribution terminal. Voltages are acquired through an acquisition circuit in the power distribution terminal, relevant processing is performed on the voltages subjected to acquisition processing through an arithmetic device in the power distribution terminal to obtain a corresponding operation result, an arc cosine table which is made according to the operation result in advance is stored in a storage of the power distribution terminal, and the voltage included angle corresponding to the operation result is found through automatic retrieval. According to actual measurement, the computing method can work out the required voltage included angle accurately to meet actual application under the premise of enough sample data, the operation is not affected by signal frequency changes, the floating point operation is avoided, and the operation efficiency is improved.

The invention discloses a control method of a three-level frequency converter, and belongs to a control method of a frequency converter. The control method comprises the steps that a. instant per-unit values of three-phase reference voltage used for controlling output voltage of the three-level frequency converter are determined, and the reference voltage is corrected according to a requirement of the three-level frequency converter on midpoint level balance control or not and according to a midpoint level deviation value; b. three-phase pulse sequence switching flags are calculated according to the instant per-unit values of the three-phase reference voltage respectively; c. the action time of three-phase virtual voltage space vectors is calculated; d. a three-level virtual voltage space vector diagram is superimposed by the three-phase virtual voltage space vectors, and a finally output pulse sequence is recombined and sorted according to the action time of the three-phase virtual voltage space vectors and by combining with the three-phase pulse sequence switching flags; and e. a frequency converter control pulse is generated by hardware, the calculated action time is loaded into a hardware unit, and the hardware unit automatically generates the control pulse as required. The control method has the advantages that a PWM (pulse-width modulation) method of the voltage space vectors is simplified.