A power management system for semiconductor memory

Abstract

The invention relates to the field of embedded system storage technology, and specifically discloses a power management system for semiconductor memory, which is used to solve the problem that existing vehicle-mounted memory screening devices often find it difficult to strike a balance between real-time performance and power consumption optimization, and the adjustment of screening strategies cannot respond to changing environmental requirements in a timely manner, which will lead to a lag in low-priority data processing, thereby causing waste of power consumption or delay in system response, and directly affecting the energy efficiency and driving safety of embedded vehicle-mounted systems. The invention comprises a data input module, an urban environment feature calculation module, a priority determination module, a screening module, a dynamic strategy adjustment module, a power management module and a feedback control module, wherein the data input module is connected to the urban environment feature calculation module and the priority determination module; and the real-time performance and power consumption of the vehicle-mounted embedded system in a complex driving environment are optimized by dynamically adjusting the memory power management, thereby extending the battery life.

Description

Technical Field

[0001] The present invention relates to the technical field of embedded system storage, and more particularly to a power management system for semiconductor storage. Background Art

[0002] In embedded systems, screening devices are components used to optimize memory access requests. According to predetermined strategies or real-time conditions, access requests entering the memory are screened and filtered to ensure that only those with priority levels that meet the set requirements can pass. This can significantly reduce the overall power consumption of the system and extend the battery life of the device. Especially in resource-limited embedded environments, screening devices can give priority to high-priority memory requests, reduce delays in critical tasks, improve the real-time and response speed of the system, and prevent excessive concurrent accesses from causing memory overload or crashes, thereby protecting the stability of the system and data integrity. In modern cars, on-board embedded systems need to process cameras in a short period of time to ensure the real-time and accuracy of functions such as autonomous driving, lane keeping assist, and collision warning. Processing these high-bandwidth video data will cause a sharp increase in system power consumption, which is inconsistent with the low power consumption requirements of embedded systems. The screening device needs to maintain sufficient real-time performance while doing its best to reduce power consumption, especially in urban driving scenarios. The screening device for semiconductor memory faces significant challenges in power management. The urban environment is complex and changeable. Vehicles need to process real-time video captured by cameras, radar data, and inputs from other auxiliary sensors while driving. The priority and processing requirements of this data are frequently adjusted as traffic conditions change, which will cause a significant increase in the system load. The screening device needs to have the ability to quickly adjust the screening strategy to adapt to this dynamic load, but existing screening devices often find it difficult to strike a balance between real-time performance and power consumption optimization. The adjustment of the screening strategy cannot respond to changing environmental requirements in a timely manner, which will cause low-priority data processing to lag, thereby causing power waste or system response delays, directly affecting the energy efficiency and driving safety of embedded vehicle systems. Summary of the invention

[0003] In order to overcome the above-mentioned defects of the prior art, the present invention provides a power management system for semiconductor memory, which optimizes the real-time performance and power consumption of the vehicle-mounted embedded system in a complex driving environment and prolongs the battery life by dynamically adjusting the memory power management.

[0004] In embedded systems, especially in-vehicle embedded systems, memory power management is critical. Such systems often need to process large amounts of real-time data, such as video streams captured by cameras, radar data, and inputs from other auxiliary sensors. The processing of this data must be completed in a very short time to support the normal operation of functions such as autonomous driving, lane keeping assistance, and collision warning. However, there is a contradiction between this high-bandwidth data processing requirement and the low power consumption requirements of embedded systems. How to minimize power consumption, extend the battery life of the device, and ensure system stability and responsiveness while ensuring system real-time performance through a reasonable power allocation vector is an urgent need at present, especially in urban traffic environments.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A power management system for a semiconductor memory comprises a data input module, an urban environment feature calculation module, a priority determination module, a screening module, a dynamic strategy adjustment module, a power management module and a feedback control module. The data input module is connected to the urban environment feature calculation module and the priority determination module. The urban environment feature calculation module is connected to the priority determination module and the dynamic strategy adjustment module. The priority determination module is connected to the screening module and the feedback control module. The screening module is connected to the power management module. The power management module receives a screening factor matrix from the screening module, and obtains a power vector allocated to the memory at the current moment based on a preset power allocation formula in combination with the basic power consumption and maximum power of the current system. The preset power allocation formula is:

[0007] ;

[0008] Where: For the current moment, For the current moment The power vector allocated to the memory of the in-vehicle embedded system, is the basic power consumption vector of the current vehicle, obtained through the vehicle embedded system, is the screening factor matrix at the current moment, obtained through the screening module, for The transposed matrix of is the maximum power matrix, obtained through the vehicle embedded system, A vector whose elements are all 1, used to filter factors Perform normalization processing, The dynamic adjustment coefficient is obtained dynamically by the dynamic strategy adjustment module considering the influence of system load, environmental characteristics, temperature influence and battery status on power management.

[0009] The operating environment of in-vehicle embedded systems is extremely dynamic, and the system load fluctuates frequently with changes in traffic conditions, road conditions, and driving scenarios. For example, when a vehicle is driving on city streets, the camera needs to handle a large number of pedestrian detection and traffic sign recognition tasks; while when driving on the highway, the system needs to pay more attention to lane keeping and distance monitoring of the vehicle in front. Such dynamic loads require the system to be able to flexibly adjust power allocation to ensure sufficient processing power under high load conditions and reduce power consumption under low load conditions. Power Allocation Vector The setting ensures that the system can give priority to high-priority memory requests. The system can dynamically adjust power allocation according to the importance of the current task based on the calculation and application of the current task. High-priority tasks, such as collision warning and autonomous driving control, can be given priority when system resources are limited, thereby reducing the delay of critical tasks and improving the real-time performance of the system. Embedded systems, especially vehicle systems, are usually limited by battery power. The setting of power allocation vectors enables the system to dynamically adjust the power consumption of each part and optimize energy efficiency according to the current system load and environmental conditions. By introducing a dynamic adjustment coefficient , the system can adjust power allocation under different environmental conditions (such as temperature changes, battery power reduction, etc.), ensuring that the system can minimize power consumption while ensuring key functions. It is calculated based on the processing time and priority of the current task. When a task with a short processing time has a higher priority, the system tends to allocate more power resources to it. This mechanism ensures that the system can complete as many high-priority tasks as possible within limited time and resources, optimizing the overall task processing efficiency. and a vector whose elements are all 1 After normalization, the formula can balance the global allocation of system resources. This process ensures that under high load, the power allocation between different modules is coordinated and consistent, and the overall stability of the system will not be affected by excessive allocation of local resources. The memory power allocation in the embedded system cannot exceed the design upper limit of the system, otherwise it will cause overheating or over-discharge, which will affect the stability and life of the system. By introducing the maximum power matrix The formula ensures that the power allocation always operates within the safe range of the system, avoiding potential hardware damage. Dynamic adjustment coefficient Generated by the dynamic policy adjustment module, it takes into account the comprehensive state of the current system (including load, environmental characteristics, temperature, and battery status). This coefficient allows the power allocation to be adjusted over time to adapt to changes in current environmental conditions and system requirements. For example, in a high temperature environment, the system needs to reduce power consumption to prevent overheating; while when the battery is fully charged, power consumption can be appropriately increased to improve system performance.

[0010] As a further solution of the present invention, the data input module acquires the raw data required by the system from the vehicle-mounted radar, camera, urban traffic monitoring system and map in real time, cleans, denoises and calibrates the collected data, and transmits the data to the urban environment feature calculation module and the priority determination module.

[0011] As a further solution of the present invention, in the urban environment feature calculation module, the speed and distance data of the vehicle are obtained through the vehicle-mounted radar, camera and urban monitoring system, and the traffic flow within a preset distance in front of the current vehicle is analyzed. The vehicle-mounted camera is used to detect pedestrians, and the deep learning model is used to identify and count them to obtain the pedestrian density. The map data and the vehicle-mounted laser radar data are combined to extract the road complexity within the preset distance in front of the vehicle. The weighted sum of traffic flow, pedestrian density and road complexity is used as the current urban environment feature. The formula for urban environment features is:

[0012] ;

[0013] Where: For the current urban environment characteristics, , , They are traffic flow, pedestrian density and road complexity, among which traffic flow is the weighted average of the set reference speed of the current road and the vehicle speed on the current road, and the pedestrian density is the pedestrian density within the field of view of the vehicle camera, and the road complexity is the weighted average of the complexity contribution factors, , , To adjust the importance weights of traffic flow, pedestrian density, and road complexity, they are adjusted according to the application scenario.

[0014] As a further solution of the present invention, in the priority determination module, based on environmental characteristics, real-time data of cameras, real-time data of vehicle-mounted radars, real-time data of urban traffic monitoring systems, historical system performance data and current load status data, a data input matrix is ​​constructed, and the defined weight vector and offset vector are used in combination with the input matrix to obtain the priority score vector according to the priority determination matrix formula. The priority determination matrix formula is:

[0015] ;

[0016] Where: is the current priority score vector, is the current data input matrix, , is the weight vector, , It is an offset vector, which is preset manually according to different scenarios.

[0017] As a further solution of the present invention, in the screening module, according to the output of the priority determination module, combined with the time-related dynamic adjustment factor, the vehicle embedded memory access request is screened and filtered, and the inverse element matrix of the Hadamard product of the current priority score vector and the dynamic adjustment factor is used as the dynamic screening factor matrix. The dynamic screening factor matrix formula is:

[0018] ;

[0019] Where: is the dynamic screening factor matrix, is the Hadamard product operator of two matrices, It is a dynamic adjustment factor obtained based on the current system load, environmental characteristics, temperature status and historical data analysis. is the inverse matrix of the matrix, and the new matrix obtained by taking the reciprocal of each element of the matrix is ​​the inverse matrix of the matrix.

[0020] As a further solution of the present invention, in the screening module, the process of obtaining the dynamic adjustment factor of the dynamic screening factor matrix formula is as follows:

[0021] Step 11, system load monitoring: obtain the current CPU usage, GPU usage, memory usage and memory access frequency of the system, normalize each load indicator, and obtain the weighted sum of the normalized values ​​of the current CPU usage, GPU usage, memory usage and memory access frequency, and the weight is set according to actual needs;

[0022] Step 12, environmental feature acquisition: obtaining the current traffic flow, pedestrian density and road complexity, and obtaining the current urban environmental features;

[0023] Step 13, temperature status analysis: obtain the device temperature and external temperature of the current memory through the embedded vehicle system, and normalize the temperature data to obtain the weighted sum, and the weight is set according to the demand;

[0024] Step 14, generating a dynamic adjustment factor: processing the outputs of step 11, step 12 and step 13 through a Sigmoid function to generate a dynamic adjustment factor matrix;

[0025] Step 15, time correlation adjustment: considering historical data and current system status, through the exponential shift method, using the past dynamic adjustment factor matrix and the current calculation results, the dynamic adjustment factor matrix at the current moment is smoothly adjusted through the preset dynamic adjustment factor correlation adjustment formula.

[0026] As a further solution of the present invention, in step 15 of the process of obtaining the dynamic adjustment factor of the screening module, the time correlation adjustment formula of the dynamic adjustment factor is:

[0027] ;

[0028] Where: is the new adjustment factor matrix calculated according to the current system state, is the adjustment factor matrix of the previous moment, is the smoothing factor, which is set according to actual needs.

[0029] As a further solution of the present invention, in the dynamic strategy adjustment module, the process of obtaining the dynamic adjustment coefficient is:

[0030] Step 21, data collection: collect the current urban environment characteristics, CPU and GPU usage, memory usage, memory access frequency, data bandwidth utilization, and external temperature and internal temperature of the vehicle system memory in real time, monitor the current power and discharge rate of the memory battery, and normalize the collected indicators;

[0031] Step 22, scoring of various influencing factors: the weighted sum of the normalized values ​​of CPU and GPU usage, memory usage, memory access frequency, and data bandwidth utilization is used as the system load monitoring score; the weighted sum of the external temperature and internal temperature of the vehicle system memory is used as the temperature environment score; the weighted sum of the current battery power and discharge rate is used as the current battery performance score;

[0032] Step 23, calculate the dynamic adjustment coefficient: use the hyperbolic sine function to process the linear adjustment value of the sum of the system load monitoring score, the temperature environment score, the battery performance score and the current urban environment characteristics, and use its value as the dynamic adjustment coefficient at the current moment to calculate and obtain the dynamic adjustment coefficient at the current moment.

[0033] As a further solution of the present invention, in step 23 of the power management module acquiring the dynamic adjustment coefficient, the formula of the dynamic adjustment coefficient is:

[0034] ;

[0035] Where: , , They are the current system load monitoring score, temperature environment score, and battery performance score. , for The linear adjustment coefficient can be set according to actual needs.

[0036] As a further solution of the present invention, the feedback control module monitors the power consumption of the current vehicle-mounted system, the system response time, processing speed, and task completion rate in real time, identifies the changing trends of system performance and power consumption based on the system's historical data, predicts future performance and power consumption changes based on the machine learning model, makes adjustments in advance, and adjusts the weight vector and offset vector in the priority determination module based on the feedback results, optimizes the priority scoring formula, and further optimizes the adjustment coefficient in the power management module based on the feedback power consumption and performance data, and feeds the analysis results back to the dynamic adjustment module to form a closed-loop control.

[0037] Compared with the prior art, the technical effect of the present invention is as follows: the present invention can flexibly respond to different loads and environmental changes in complex urban driving scenarios by dynamically adjusting the power allocation strategy, screening factor matrices and dynamic adjustment coefficients, ensuring that high-priority tasks are given priority when resources are limited, while minimizing system power consumption and extending the battery life of the device. By combining multi-dimensional environmental characteristics and system load analysis, the response speed and energy efficiency of the vehicle-mounted embedded system are significantly improved, so that it can achieve global optimization of power management while ensuring driving safety and system stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 The system block diagram of the system of the present invention. DETAILED DESCRIPTION

[0039] The following will be combined with the accompanying drawings in the present invention to clearly and completely describe the technical solution in the present invention. Obviously, the described technical solution is only a part of the present invention but not all of it. Based on the technical solution in the present invention, all other technical solutions obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0040] Compared with the prior art, the power management system for semiconductor memory proposed in the present invention can flexibly respond to different loads and environmental changes in complex urban driving scenarios by dynamically adjusting the power allocation strategy, screening factor matrix and dynamic adjustment coefficient, ensuring that high-priority tasks are given priority when resources are limited, while minimizing system power consumption and extending the battery life of the device. By combining multi-dimensional environmental characteristics and system load analysis, the response speed and energy efficiency of the vehicle-mounted embedded system are significantly improved, so that it can achieve global optimization of power management while ensuring driving safety and system stability.

[0041] like Figure 1 As shown, the system proposed in the present invention includes a data input module, an urban environment feature calculation module, a priority determination module, a screening module, a dynamic strategy adjustment module, a power management module and a feedback control module. The data input module is connected to the urban environment feature calculation module and the priority determination module. The urban environment feature calculation module is connected to the priority determination module and the dynamic strategy adjustment module. The priority determination module is connected to the screening module and the feedback control module. The screening module is connected to the power management module. The power management module receives the screening factor matrix from the screening module, and combines the basic power consumption and maximum power of the current system to obtain the power vector allocated to the memory at the current moment based on the preset power allocation formula. The preset power allocation formula is:

[0042] ;

[0043] Where: For the current moment, For the current moment The power vector allocated to the memory of the in-vehicle embedded system, is the basic power consumption vector of the current vehicle, obtained through the vehicle embedded system, is the screening factor matrix at the current moment, obtained through the screening module, for The transposed matrix of is the maximum power matrix, obtained through the vehicle embedded system, A vector whose elements are all 1, used to filter factors Perform normalization processing, The dynamic adjustment coefficient is obtained dynamically by the dynamic strategy adjustment module considering the influence of system load, environmental characteristics, temperature influence and battery status on power management.

[0044] The operating environment of in-vehicle embedded systems is extremely dynamic, and the system load fluctuates frequently with changes in traffic conditions, road conditions, and driving scenarios. For example, when a vehicle is driving on city streets, the camera needs to handle a large number of pedestrian detection and traffic sign recognition tasks; while when driving on the highway, the system needs to pay more attention to lane keeping and distance monitoring of the vehicle in front. Such dynamic loads require the system to be able to flexibly adjust power allocation to ensure sufficient processing power under high load conditions and reduce power consumption under low load conditions. Power Allocation Vector The setting ensures that the system can give priority to high-priority memory requests. The system can dynamically adjust power allocation according to the importance of the current task based on the calculation and application of the current task. High-priority tasks, such as collision warning and autonomous driving control, can be given priority when system resources are limited, thereby reducing the delay of critical tasks and improving the real-time performance of the system. Embedded systems, especially vehicle systems, are usually limited by battery power. The setting of power allocation vectors enables the system to dynamically adjust the power consumption of each part and optimize energy efficiency according to the current system load and environmental conditions. By introducing a dynamic adjustment coefficient , the system can adjust power allocation under different environmental conditions (such as temperature changes, battery power reduction, etc.), ensuring that the system can minimize power consumption while ensuring key functions. It is calculated based on the processing time and priority of the current task. When a task with a short processing time has a higher priority, the system tends to allocate more power resources to it. This mechanism ensures that the system can complete as many high-priority tasks as possible within limited time and resources, optimizing the overall task processing efficiency. and a vector whose elements are all 1 After normalization, the formula can balance the global allocation of system resources. This process ensures that under high load, the power allocation between different modules is coordinated and consistent, and the overall stability of the system will not be affected by excessive allocation of local resources. The memory power allocation in the embedded system cannot exceed the design upper limit of the system, otherwise it will cause overheating or over-discharge, which will affect the stability and life of the system. By introducing the maximum power matrix The formula ensures that the power allocation always operates within the safe range of the system, avoiding potential hardware damage. Dynamic adjustment coefficient Generated by the dynamic policy adjustment module, it takes into account the comprehensive state of the current system (including load, environmental characteristics, temperature, and battery status). This coefficient allows the power allocation to be adjusted over time to adapt to changes in current environmental conditions and system requirements. For example, in a high temperature environment, the system needs to reduce power consumption to prevent overheating; while when the battery is fully charged, power consumption can be appropriately increased to improve system performance.

[0045] It should be noted that the data input module obtains the raw data required by the system from the vehicle-mounted radar, camera, urban traffic monitoring system and map in real time, cleans, denoises and calibrates the collected data, and transmits the data to the urban environment feature calculation module and the priority determination module.

[0046] The data input module acquires and processes raw data from multiple sources in real time to ensure the accuracy and reliability of the data, and transmits it to the urban environment feature calculation module and the priority determination module, so that the system can adapt to complex driving environments more accurately, optimize real-time performance and power consumption management, thereby improving the system's response speed and overall stability.

[0047] It should be noted that in the urban environment feature calculation module, the vehicle speed and distance data are obtained through the vehicle-mounted radar, camera and urban monitoring system, and the traffic flow within the preset distance in front of the current vehicle is analyzed. The vehicle-mounted camera is used to detect pedestrians, and the deep learning model is used to identify and count them to obtain the pedestrian density. Combined with the map data and the vehicle-mounted lidar data, the road complexity within the preset distance in front of the vehicle is extracted, and the weighted sum of traffic flow, pedestrian density and road complexity is used as the current urban environment feature. The formula for urban environment features is:

[0048] ;

[0049] Where: For the current urban environment characteristics, , , They are traffic flow, pedestrian density and road complexity, among which traffic flow is the weighted average of the set reference speed of the current road and the vehicle speed on the current road, and the pedestrian density is the pedestrian density within the field of view of the vehicle camera, and the road complexity is the weighted average of the complexity contribution factors, , , To adjust the importance weights of traffic flow, pedestrian density, and road complexity, they are adjusted according to the application scenario.

[0050] By acquiring and analyzing key environmental characteristics in real time, the system can more accurately predict and respond to load demands in different driving scenarios. This precise load prediction helps to dynamically adjust the power allocation of the memory, so that high-priority tasks can still obtain sufficient resources in complex environments, ensuring the real-time and reliability of the system; according to changes in traffic flow, pedestrian density and road complexity, the system can intelligently adjust power consumption. For example, in scenes with dense traffic or dense pedestrians, the system can allocate more power to handle key tasks, while in simpler road conditions, it can reduce power consumption, thereby extending the battery life of the device; combined with multi-dimensional environmental data, it effectively reduces power consumption fluctuations caused by environmental complexity, allowing the system to maintain stable performance output in various urban driving scenarios, avoiding memory overload or resource waste, thereby improving the stability and durability of the entire system.

[0051] It should be noted that in the priority determination module, a data input matrix is ​​constructed based on environmental characteristics, real-time data from cameras, real-time data from vehicle-mounted radars, real-time data from urban traffic monitoring systems, historical system performance data, and current load status data. The defined weight vector and offset vector are combined with the input matrix to obtain the priority score vector according to the priority determination matrix formula. The priority determination matrix formula is:

[0052] ;

[0053] Where: is the current priority score vector, is the current data input matrix, , is the weight vector, , It is an offset vector, which is preset manually according to different scenarios.

[0054] By comprehensively considering environmental characteristics and the current state of the system, the priority determination module can assign an accurate priority score to each memory access request, dynamically adjust resource allocation, give priority to high-priority tasks, ensure timely processing of key data, and optimize the efficiency of memory resource use; using real-time data and historical performance data to determine priority, the system can quickly respond to changes in the environment and load, so that the system can adaptively adjust memory power allocation when facing complex and dynamic urban driving environments, thereby improving the real-time and response speed of the system; through the priority score vector, the system can identify low-priority or non-urgent memory requests and reduce their processing frequency or By postponing its processing time, this power management strategy can significantly reduce unnecessary power consumption, thereby extending the battery life of the device, which is particularly important in embedded systems with limited resources; by integrating multi-dimensional data for priority determination, the system can better balance the processing requirements of different tasks, avoid memory overload or system crash caused by excessive allocation of resources to low-priority tasks, thereby improving the overall stability of the system and data integrity; using offset vectors and weight vectors, the manually set parameters allow the system to flexibly adjust the priority determination criteria according to different scenarios (urban environment, road conditions, system load, etc.), thereby ensuring that efficient power management can still be maintained in changing scenarios.

[0055] It should be noted that in the screening module, according to the output of the priority determination module, combined with the time-related dynamic adjustment factor, the vehicle embedded memory access request is screened and filtered, and the inverse element matrix of the Hadamard product of the current priority score vector and the dynamic adjustment factor is used as the dynamic screening factor matrix. The dynamic screening factor matrix formula is:

[0056] ;

[0057] Where: is the dynamic screening factor matrix, is the Hadamard product operator of two matrices, It is a dynamic adjustment factor obtained based on the current system load, environmental characteristics, temperature status and historical data analysis. is the inverse matrix of the matrix, and the new matrix obtained by taking the reciprocal of each element of the matrix is ​​the inverse matrix of the matrix.

[0058] By calculating the dynamic screening factor matrix, the system can more accurately control the allocation of resources. The priority score is combined with the dynamic adjustment factor and adjusted through the inverse element matrix, so that tasks with fast processing speed can get priority resources, ensuring the real-time and response speed of the system; Adaptability and flexibility are enhanced: The dynamic adjustment factor takes into account the current load, environmental characteristics, temperature status and historical data analysis of the system, and can dynamically adjust the screening strategy according to the real-time status of the system. In this way, the system can flexibly adapt to the changing driving environment and load changes, ensuring the priority processing of key tasks, while avoiding excessive allocation and waste of resources; By accurately screening and filtering memory access requests, the system can effectively reduce the resource allocation of low-priority tasks and reduce unnecessary power consumption, which is particularly important for embedded systems with limited resources, and can significantly extend the battery life of the device and improve the energy efficiency of the system; Through the use of the dynamic screening factor matrix, the system can still maintain stable processing efficiency under high load conditions. The inverse element matrix allows tasks with shorter processing times to be processed first, avoiding system crashes due to overload and ensuring stable system operation. The introduction of the dynamic screening factor matrix ensures global optimization of power management. By comprehensively considering the priority and processing efficiency of each data request, the system can maximize power usage while ensuring that important tasks are processed first, thereby achieving global power management optimization. Combined with dynamic adjustment factors for screening, the system can process a variety of different data sources (such as camera data, radar data, etc.) in complex urban driving environments, and make timely adjustments based on real-time changes in the environment, thereby improving the system's processing capabilities in complex scenarios.

[0059] It should be noted that in the screening module, the process of obtaining the dynamic adjustment factor of the dynamic screening factor matrix formula is as follows:

[0060] Step 11, system load monitoring: obtain the current CPU usage, GPU usage, memory usage and memory access frequency of the system, normalize each load indicator, and obtain the weighted sum of the normalized values ​​of the current CPU usage, GPU usage, memory usage and memory access frequency, and the weight is set according to actual needs;

[0061] Step 12, environmental feature acquisition: obtaining the current traffic flow, pedestrian density and road complexity, and obtaining the current urban environmental features;

[0062] Step 13, temperature status analysis: obtain the device temperature and external temperature of the current memory through the embedded vehicle system, and normalize the temperature data to obtain the weighted sum, and the weight is set according to the demand;

[0063] Step 14, generating a dynamic adjustment factor: processing the outputs of step 11, step 12 and step 13 through a Sigmoid function to generate a dynamic adjustment factor matrix;

[0064] Step 15, time correlation adjustment: considering historical data and current system status, through the exponential shift method, using the past dynamic adjustment factor matrix and the current calculation results, the dynamic adjustment factor matrix at the current moment is smoothly adjusted through the preset dynamic adjustment factor correlation adjustment formula.

[0065] In step 11, by monitoring and normalizing the CPU, GPU, memory usage, and memory access frequency, the system can accurately reflect the current load status. The system comprehensive load score calculated by weighted sum enables the dynamic adjustment factor matrix to accurately reflect the current resource demand, thereby dynamically adjusting the power allocation strategy to ensure that system resources are reasonably allocated under high load and improve the efficiency of task processing; the environmental characteristics (traffic flow, pedestrian density, road complexity) obtained in step 12 enable the system to dynamically adjust the power management strategy of the memory according to the complexity of the real-time urban driving environment. This real-time environmental response capability enables the system to prioritize the processing of key mission data under busy traffic or complex road conditions to ensure the reliability and real-time performance of the system; in step 13, by monitoring and normalizing the device temperature and external temperature, the system can effectively control the operating temperature of the memory to prevent performance degradation or hardware damage caused by overheating. The temperature impact score calculated by weighted sum enables the system to actively reduce power consumption in high temperature environments to extend device life, while reasonably improving performance in low temperature environments to optimize overall power consumption management; in step 14, the comprehensive output of system load, environmental characteristics and temperature status is processed by the Sigmoid function to generate a dynamic The adjustment factor matrix can flexibly adjust the power distribution under different load and environmental conditions. This nonlinear processing method ensures that the system can respond and adjust in time under extreme conditions (such as high load or high temperature) to avoid resource waste or system overload, thereby achieving global power optimization. The time-related adjustment in step 15 uses the exponential shift method to smoothly adjust the current dynamic adjustment factor matrix by considering historical data and the current system state. This smoothing process avoids drastic changes in power distribution caused by drastic fluctuations in the environment or load, allowing the system to operate more stably and ensure long-term system performance and power consumption balance. Through the dynamic adjustment factor matrix generated by the above process, the system can maintain stable performance output under variable driving environments and load conditions, reduce memory overload or system crash caused by over-allocation of resources or ignoring environmental changes, thereby significantly improving the stability and reliability of the system. Through precise dynamic adjustment, the system can reasonably allocate power according to current needs and reduce unnecessary power consumption, which is particularly critical for vehicle-mounted embedded systems and can extend battery life to ensure that the system can still maintain efficient power management during long-term operation.

[0066] It should be noted that in step 15 of the process of obtaining the dynamic adjustment factor of the screening module, the time correlation adjustment formula of the dynamic adjustment factor is:

[0067] ;

[0068] Where: is the new adjustment factor matrix calculated according to the current system state, is the adjustment factor matrix of the previous moment, is the smoothing factor, which is set according to actual needs.

[0069] The smoothing factor plays a key role in the time-dependent adjustment of the dynamic adjustment factor matrix. It controls the current adjustment factor matrix In the new calculation, the adjustment factor matrix The adjustment factor matrix of the previous moment Specifically, the setting of the smoothing factor needs to be adjusted according to the dynamic changes in the actual application scenario. The following are several typical scenarios and the corresponding smoothing factor settings:

[0070] Rapidly changing scenarios (e.g. emergency driving situations): In emergency driving situations, such as sudden braking and obstacle avoidance, the system needs to respond quickly to ensure driving safety. At this time, the system load and environmental characteristics will change dramatically. In this case, the smoothing factor is 0.7 to 0.9, so that the system will be more inclined to use the calculation results at the current moment. In order to quickly adapt to the current emergency situation, adjust the power management strategy in time, ensure that high-priority tasks are given priority, and enable the system to more sensitively reflect current changes and respond quickly.

[0071] Stable scenario (e.g. high-speed cruising): When cruising steadily on the highway, the system load and environmental characteristics change little, the system state is relatively stable, the smoothing factor is 0.3 to 0.5, and the system will rely more on the previous adjustment factor matrix , ensure the stability of the power management strategy, avoid frequent adjustments due to slight fluctuations, which will lead to unstable system power consumption, and ensure that the system runs smoothly in a stable environment, extending battery life.

[0072] Gradual transition scenarios (such as entering complex urban road conditions): When the vehicle enters complex urban road conditions from the highway, the system needs to gradually adapt to the increased load and complex environmental characteristics. The smoothing factor is 0.5 to 0.7. The system can gradually transition, referring to both the current environmental changes and the previous adjustment factors, so that the system can smoothly adapt to the new driving environment and ensure that the system can smoothly transition to the new working state to avoid performance fluctuations caused by rapid changes.

[0073] Scenarios of dramatic environmental changes (e.g., sudden weather changes or traffic jams): Sudden weather changes (e.g., heavy rain, dense fog) or traffic jams require the system to quickly adjust power allocation to ensure the priority of key sensors and processing units. The smoothing factor is 0.7 to 0.9, which enables the system to quickly adapt to dramatic changes in the new environment, helping the system to quickly respond to dramatically changing environments and ensure driving safety.

[0074] Resource-constrained scenarios (such as low-battery states): When the battery is low, the system needs to maintain low-power operation while ensuring the continuity of key functions. The smoothing factor is 0.5 to 0.7, so that the system can respond to current power state changes in a timely manner without over-relying on new data to increase power consumption. This helps to maintain stable power consumption in low-battery states and extend battery life.

[0075] It should be noted that in the dynamic strategy adjustment module, the process of obtaining the dynamic adjustment coefficient is as follows:

[0076] Step 21, data collection: collect the current urban environment characteristics, CPU and GPU usage, memory usage, memory access frequency, data bandwidth utilization, and external temperature and internal temperature of the vehicle system memory in real time, monitor the current power and discharge rate of the memory battery, and normalize the collected indicators;

[0077] Step 22, scoring of various influencing factors: the weighted sum of the normalized values ​​of CPU and GPU usage, memory usage, memory access frequency, and data bandwidth utilization is used as the system load monitoring score; the weighted sum of the external temperature and internal temperature of the vehicle system memory is used as the temperature environment score; the weighted sum of the current battery power and discharge rate is used as the current battery performance score;

[0078] Step 23, calculate the dynamic adjustment coefficient: use the hyperbolic sine function to process the linear adjustment value of the sum of the system load monitoring score, the temperature environment score, the battery performance score and the current urban environment characteristics, and use its value as the dynamic adjustment coefficient at the current moment to calculate and obtain the dynamic adjustment coefficient at the current moment.

[0079] By real-time acquisition and normalization of various load indicators of the system (such as CPU / GPU usage, memory usage, memory access frequency, and data bandwidth utilization), the system can accurately score the current load status. Combined with the nonlinear processing of the hyperbolic sine function, the dynamic adjustment coefficient can timely reflect the changes in the system load, ensuring that the system can give priority to the power requirements of key tasks when the load is high, and improve the response speed and processing efficiency of the system; by monitoring the external and internal temperature of the vehicle system memory and weightedly calculating the temperature environment score, the system can accurately control the impact of temperature on power consumption, and the dynamic adjustment coefficient can adjust the power allocation of the memory according to the current temperature conditions to avoid hardware damage or performance degradation caused by excessive temperature, and reasonably improve system performance in low temperature environments to achieve efficient power consumption management; by monitoring the current battery power and discharge rate and calculating the battery performance score, the system can dynamically adjust the power allocation according to the battery status. The calculation of the dynamic adjustment coefficient ensures that when the power is sufficient, the system can make full use of power resources to improve performance, and when the power is low, the system will give priority to saving power. By incorporating urban environmental characteristics into the calculation of the dynamic adjustment coefficient, the system can dynamically adjust the power allocation strategy of the memory according to environmental changes such as real-time traffic flow, pedestrian density, and road complexity. This real-time response capability enables the system to prioritize high-priority tasks in complex urban driving scenarios to ensure driving safety and system stability. By using the hyperbolic sine function to process the system load, temperature environment, battery performance, and comprehensive scores of urban environmental characteristics, the system can perform global optimization. This nonlinear processing method makes the dynamic adjustment coefficient smoother, avoids drastic changes in power allocation caused by drastic fluctuations in the environment or load, ensures smooth operation of the system, and reduces the impact of sudden changes on system performance and power consumption. By comprehensively considering various influencing factors and dynamically adjusting the adjustment coefficient, the system can maintain a stable power management strategy under changing driving environments and load conditions. This dynamic and precise adjustment mechanism helps prevent memory overload, system crashes, or excessive power consumption, and improves the overall stability and reliability of the system.

[0080] It should be noted that in step 23 of the power management module acquiring the dynamic adjustment coefficient, the formula of the dynamic adjustment coefficient is:

[0081] ;

[0082] Where: , , They are the current system load monitoring score, temperature environment score, and battery performance score. , for The linear adjustment coefficient can be set according to actual needs.

[0083] It should be noted that the feedback control module monitors the power consumption of the current vehicle system, the system's response time, processing speed, and task completion rate in real time. Based on the system's historical data, it identifies the changing trends of system performance and power consumption, and predicts future performance and power consumption changes based on the machine learning model. Adjustments are made in advance, and based on the feedback results, the weight vector and offset vector in the priority determination module are adjusted, the priority scoring formula is optimized, and based on the feedback power consumption and performance data, the adjustment coefficient in the power management module is further optimized. The analysis results are fed back to the dynamic adjustment module to form a closed-loop control.

[0084] The feedback control module monitors power consumption, system response time, processing speed and task completion rate in real time, and combines historical data with machine learning models to identify changing trends in system performance and power consumption. This enables the system to adaptively optimize its own parameters, such as the weight vector and offset vector in the priority determination module, and the adjustment coefficient in the power management module. In this way, the system can continuously optimize performance under different workloads and environments, achieve adaptive adjustments, and ensure that the system always runs in the best state; by predicting future performance and power consumption changes, the feedback control module can make adjustments in advance to avoid performance degradation or hardware damage due to sudden changes in system load or surges in power consumption. This preventive adjustment helps maintain system stability and prevents problems such as overheating, overload or battery exhaustion, thereby improving system reliability and durability. The feedback control module optimizes the adjustment coefficient in the power management module based on real-time monitoring data and historical trends, making power allocation more accurate. By continuously adjusting the adjustment coefficient, the system can better balance performance and power consumption, ensuring that power consumption is minimized while ensuring system response and processing efficiency, thereby extending battery life. Closed-loop control enables the system to continuously apply real-time feedback data to the dynamic adjustment module to form a continuously optimized closed-loop system. This closed-loop control method ensures that the system can be adjusted based on the latest feedback data at any point in time to optimize the global power management strategy. By adjusting the weight vector and offset vector in the priority determination module, the system can more accurately evaluate and process task priorities, thereby improving the system's response speed and task completion rate. This optimization enables high-priority tasks to be processed faster, ensuring the real-time and reliability of the system in key scenarios. By using machine learning models to analyze historical data and predict future trends, the system can continuously learn and improve its power management strategy. This long-term learning ability enables the system to gradually improve performance and perform better and more efficiently in response to new loads and environmental changes. Through the application of closed-loop control and machine learning models, the system can autonomously identify and adjust optimization parameters, reducing reliance on manual intervention. The system's intelligence level is improved, enabling it to automatically adapt to changing usage environments, maintain efficient operation, and provide a better user experience.

[0085] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0086] Finally: The above description is only an implementation scheme of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A power management system for a semiconductor memory, characterized in that: The system includes a data input module, an urban environment characteristic calculation module, a priority determination module, a screening module, a dynamic strategy adjustment module, a power management module and a feedback control module. The data input module is connected to the urban environment characteristic calculation module and the priority determination module. The urban environment characteristic calculation module is connected to the screening module and the dynamic strategy adjustment module. The priority determination module is connected to the screening module and the feedback control module. The screening module is connected to the power management module. The power management module receives the screening factor matrix from the screening module, and obtains the power vector allocated to the memory at the current moment based on the preset power allocation formula in combination with the basic power consumption and maximum power of the current system. The preset power allocation formula is: Where: t is the current time, P mem (t) is the power vector allocated to the vehicle embedded system memory at the current time t, P base is the basic power consumption vector of the current vehicle, obtained through the vehicle embedded system, F(t) is the screening factor matrix at the current moment, obtained through the screening module, F(t) T is the transposed matrix of F(t), P max is the maximum power matrix obtained through the vehicle embedded system, S T is a vector whose elements are all 1, used to normalize the screening factor F(t), α(t) is the dynamic adjustment coefficient, and the dynamic strategy adjustment module dynamically obtains the influence of system load, environmental characteristics, temperature influence and battery status on power management; In the priority determination module, a data input matrix is ​​constructed based on environmental characteristics, real-time data from cameras, real-time data from vehicle-mounted radars, real-time data from urban traffic monitoring systems, historical system performance data, and current load status data. The defined weight vector and offset vector are combined with the input matrix to obtain the priority score vector according to the priority determination matrix formula. The priority determination matrix formula is: P nonlinear (t)=tanh[D(t)·ω+b]; Where: P nonlinear (t) is the current priority score vector, D(t) is the current data input matrix, ω is the weight vector, ω=[ω1,ω2,ω3], b is the offset vector, which is preset manually according to different scenarios; In the screening module, the access request to the vehicle embedded memory is screened and filtered according to the output of the priority determination module and the time-related dynamic adjustment factor. The inverse element matrix of the Hadamard product of the current priority score vector and the dynamic adjustment factor is used as the dynamic screening factor matrix. The dynamic screening factor matrix formula is: Where: F(t) is the dynamic screening factor matrix, is the Hadamard product operator of the two matrices, M(t) is the dynamic adjustment factor, which is obtained based on the current system load, environmental characteristics, temperature status and historical data analysis, T -1 is the inverse matrix of the matrix, and the new matrix obtained by taking the reciprocal of each element of the matrix is ​​the inverse matrix of the matrix; In the screening module, the process of obtaining the dynamic adjustment factor of the dynamic screening factor matrix formula is as follows: Step 11, system load monitoring: obtain the current CPU usage, GPU usage, memory usage and memory access frequency of the system, normalize each load indicator, and obtain the weighted sum of the normalized values ​​of the current CPU usage, GPU usage, memory usage and memory access frequency, and the weight is set according to actual needs; Step 12, environmental feature acquisition: obtaining the current traffic flow, pedestrian density and road complexity, and obtaining the current urban environmental features; Step 13, temperature status analysis: obtain the device temperature and external temperature of the current memory through the embedded vehicle system, and normalize the temperature data to obtain the weighted sum, and the weight is set according to the demand; Step 14, generating a dynamic adjustment factor: processing the outputs of step 11, step 12 and step 13 through a Sigmoid function to generate a dynamic adjustment factor matrix; Step 15, time correlation adjustment: considering historical data and current system status, using the exponential shift method, using the past dynamic adjustment factor matrix and the current calculation results, and smoothly adjusting the dynamic adjustment factor matrix at the current moment through the preset dynamic adjustment factor correlation adjustment formula; The feedback control module monitors the power consumption of the current vehicle system, the system's response time, processing speed, and task completion rate in real time. Based on the system's historical data, it identifies the changing trends of system performance and power consumption, predicts future performance and power consumption changes based on the machine learning model, makes adjustments in advance, and adjusts the weight vector and offset vector in the priority determination module based on the feedback results, optimizes the priority scoring formula, and further optimizes the adjustment coefficient in the power management module based on the feedback power consumption and performance data. The analysis results are fed back to the dynamic adjustment module to form a closed-loop control.

2. A power management system for a semiconductor memory according to claim 1, characterized in that: The data input module obtains the raw data required by the system from the vehicle-mounted radar, camera, urban traffic monitoring system and map in real time, cleans, denoises and calibrates the collected data, and transmits the data to the urban environment feature calculation module and priority determination module.

3. A power management system for a semiconductor memory according to claim 2, characterized in that: In the urban environment feature calculation module, the vehicle speed and distance data are obtained through the vehicle-mounted radar, camera and urban monitoring system, and the traffic flow within the preset distance in front of the current vehicle is analyzed. The vehicle-mounted camera is used to detect pedestrians, and the deep learning model is used to identify and count them to obtain the pedestrian density. Combined with the map data and the vehicle-mounted laser radar data, the road complexity within the preset distance in front of the vehicle is extracted, and the weighted sum of traffic flow, pedestrian density and road complexity is used as the current urban environment feature. The formula for urban environment features is: E1=β1T f +β2P d +β3R c ; Where: E1 is the current urban environment characteristics, T f , P d , R c They are traffic flow, pedestrian density and road complexity, where traffic flow T f is the weighted average of the set reference speed of the current road and the vehicle speed on the current road, and the pedestrian density P d is the pedestrian density within the field of view of the vehicle camera, and the road complexity R c is the weighted average of the complexity contribution factors, β1, β2, and β3 are the importance weights for adjusting traffic flow, pedestrian density, and road complexity, which are adjusted according to the application scenario.

4. A power management system for a semiconductor memory according to claim 3, characterized in that: In step 15 of the process of obtaining the dynamic adjustment factor of the screening module, the time correlation adjustment formula of the dynamic adjustment factor is: M(t)=γM new (t)+(1-γ)M(t-1); Where: M new (t) is the new adjustment factor matrix calculated according to the current system state, M(t-1) is the adjustment factor matrix at the previous moment, and γ is the smoothing factor, which is set according to actual needs.

5. A power management system for a semiconductor memory according to claim 4, characterized in that: In the dynamic strategy adjustment module, the process of obtaining the dynamic adjustment coefficient is as follows: Step 21, data collection: collect the current urban environment characteristics, CPU and GPU usage, memory usage, memory access frequency, data bandwidth utilization, and external temperature and internal temperature of the vehicle system memory in real time, monitor the current power and discharge rate of the memory battery, and normalize the collected indicators; Step 22, scoring of various influencing factors: the weighted sum of the normalized values ​​of CPU and GPU usage, memory usage, memory access frequency, and data bandwidth utilization is used as the system load monitoring score; the weighted sum of the external temperature and internal temperature of the vehicle system memory is used as the temperature environment score; the weighted sum of the current battery power and discharge rate is used as the current battery performance score; Step 23, calculate the dynamic adjustment coefficient: use the hyperbolic sine function to process the linear adjustment value of the sum of the system load monitoring score, the temperature environment score, the battery performance score and the current urban environment characteristics, and use its value as the dynamic adjustment coefficient at the current moment to calculate and obtain the dynamic adjustment coefficient at the current moment.

6. A power management system for a semiconductor memory according to claim 5, characterized in that: In step 23 of obtaining the dynamic adjustment coefficient by the power management module, the formula of the dynamic adjustment coefficient is: α(t)=tanh[μ+ρ(θ1+θ2+θ3+E1)]; Where: θ1, θ2, θ3 are the current system load monitoring score, temperature environment score, and battery performance score, respectively; ρ and μ are the linear adjustment coefficients of θ1+θ2+θ3+E1, which are set according to actual needs.

Images