6results about How to "Avoid processing delays" patented technology

The present disclosure provides an agent-based data flow real-time scheduling method, which comprises: according to the factory production environment, building an intelligent scheduling framework driven by reinforcement learning, which comprises an environment, an agent and a reward function; converting data flow characteristics and system state into feature vectors to construct a multi-dimensional feature reinforcement learning state space; the agent perceives the reinforcement learning state space and calls a prediction model to predict the data flow characteristic change trend and the system state change trend in the future preset time period, and generates a resource pre-allocation scheme based on the prediction result; the agent generates at least one scheduling action based on a multi-objective scheduling optimization algorithm to form a scheduling strategy according to the current production target, the resource pre-allocation scheme and the reinforcement learning state space. The present disclosure effectively avoids resource waste and data processing delay, significantly improves resource utilization, and ensures stable and efficient production of the intelligent factory production environment.

This invention discloses a distributed message processing system based on dynamic sliding windows and intelligent scheduling, aiming to solve the problems of processing latency, resource waste, and insufficient scalability caused by fixed grouping and polling strategies in traditional message processing technologies. It is suitable for dynamic message scheduling and load balancing in high-concurrency, high-throughput scenarios. A four-layer collaborative architecture is adopted, with a distributed message queue layer, a sliding grouping module, an intelligent scheduling module, and a cluster of processing units working together to form a closed-loop control through a state feedback mechanism. The system designs a hybrid sliding window and dynamic triggering grouping mechanism, a two-way feedback intelligent scheduling system, and a multi-dimensional elastic control strategy, which can dynamically optimize resource allocation based on message traffic fluctuations and processing unit load status, balancing processing timeliness and transaction consistency. It can be widely applied to scenarios such as financial transaction clearing, e-commerce flash sale traffic smoothing, and IoT data processing, achieving adaptive throughput improvement and efficient resource utilization, and possesses strong practicality and promotional value.

The invention relates to a multi-screen playing method and device, computer equipment, a storage medium and a computer program product. The method comprises the following steps: acquiring video source data of a target video, and demultiplexing the video source data to obtain video coding data; determining a video timestamp of each video frame in the target video according to the playing speed in the video coding data; packaging the video coding data to obtain a real-time transmission data packet, the real-time transmission data packet at least comprising the video coding data and each video timestamp; and sending the real-time transmission data packet to a receiving end device, so that the receiving end device decodes the video coding data to obtain video decoding data, and performs picture rendering on the video decoding data based on the video timestamps to synchronously play video pictures of the target video with the sending end device based on the video timestamps. By adopting the method, the playing time delay of picture synchronous playing on different screen interfaces can be reduced.

The application discloses an end-edge collaborative adaptive enhancement and coding method for low-light video analysis, comprising: a terminal device collects a low-light video stream in real time and synchronously monitors environmental light intensity, edge computing power and network bandwidth; an edge side generates a collaborative enhancement strategy based on multi-modal perception data, dynamically selects an enhancement level and a coding resolution, and delivers them to the terminal; the terminal executes brightness correction and contrast stretching through a lightweight spatial enhancement network, then encodes and transmits to the edge side according to the strategy resolution; after decoding the video stream, the edge side optimizes cross-frame motion compensation and noise suppression through a timing enhancement network; finally, a target detection model is input for inference, and the result is fed back to optimize the strategy. Through dynamic resource scheduling, division of labor and cooperation of neural enhancement networks and closed-loop feedback mechanism, the performance of low-light video analysis is significantly improved; through deep reinforcement learning, the enhancement level and resolution are optimized in real time, and the accuracy, delay and resource efficiency are balanced.

This disclosure relates to transposing a neural network matrix in hardware. The disclosure relates to methods, systems, and apparatus for transposing a neural network matrix in hardware, including computer programs encoded on a computer storage medium. In one aspect, a method includes the actions of: receiving a request on a hardware circuit having a matrix computation unit to perform computation on a neural network, the request specifying a transpose operation to be performed on a first neural network matrix; and generating instructions, when executed by the hardware circuit, to transpose the first neural network matrix by performing a first operation, wherein the first operation includes repeatedly performing the following second operations: updating the first neural network matrix by swapping the upper right quadrant and lower left quadrant of each current submatrix for a current subdivision of the first neural network matrix that divides it into one or more current submatrixes; and updating the current subdivision by subdividing each current submatrix into a corresponding new submatrix.

The application discloses a kind of task processing systems based on intelligent internet of things equipment.The system includes system initialization and task table dynamic construction unit, hardware, protocol stack initialization, create event table etc.and predefine core task;Hierarchical event polling and priority processing unit, event is handled according to priority;Adaptive low-power management unit, adjust system power consumption;Dynamic task extension and hot plug support unit, identify new equipment and allocate task resources.Wherein, hierarchical event polling unit can distinguish key and ordinary event processing.The application realizes the efficient processing of internet of things equipment task by the cooperation of each unit, with dynamic expansion capacity, can also adaptively adjust power consumption, improve system performance and energy utilization efficiency.