A battery management method for an intelligent cleaning robot

Abstract

The application relates to the technical field of battery management, and particularly relates to a battery management method for an intelligent cleaning robot, which comprises collecting environmental temperature, task load, ground material and user habit data to construct an initial data set; using a support vector machine to identify battery loss acceleration characteristics under high load; combining charge and discharge logs to calculate loss acceleration indexes; generating a predicted attenuation curve through a random forest algorithm; adjusting dynamic adaptation parameters based on attenuation reasons to form optimized management rules; then fusing user habits and ground materials through a gradient boosting algorithm to generate an individualized management scheme; real-time monitoring of battery states and activation of a strategy adjustment module when high load scenarios are matched to dynamically correct operation parameters; and finally planning an efficiency improvement path by fusing environmental temperature. The application can accurately predict battery attenuation trends, adaptively optimize power management strategies, effectively delay capacity attenuation, and improve operation safety and equipment durability.

Description

Technical Field

[0001] This invention relates to the field of battery management technology, and more specifically to a battery management method for an intelligent cleaning robot. Background Technology

[0002] With the rapid development of service robot technology, intelligent cleaning robots are increasingly being used in home and commercial settings, becoming a key tool for improving quality of life and work efficiency. As their core energy unit, the performance and lifespan of the battery system directly affect the device's range and maintenance costs. Against the backdrop of continuously evolving technology, efficient and intelligent battery management systems have become a core technological pillar for ensuring the long-term stable operation of robots.

[0003] Real-time battery management in dynamic environments is a key focus of the industry. This technology aims to accurately sense energy consumption, thermal state, and task requirements during robot operation through multi-sensor fusion. Its fundamental goal is to achieve refined control of battery status through closed-loop analysis of operational data to adapt to power supply demands under different operating conditions.

[0004] However, existing battery management technologies, when dealing with complex operating environments, often focus on static current and voltage protection, failing to deeply analyze the coupling relationship between ambient temperature, workload, and ground material. This results in insufficient accuracy in identifying accelerated battery wear characteristics under high-load conditions. Furthermore, traditional solutions generally ignore the correlation between personalized user operating habits and long-term battery performance degradation, lacking effective predictive mechanisms to address capacity decline under varying usage frequencies and intensities. In addition, the feedback adjustment process of existing management rules is relatively lagging, unable to dynamically correct operating parameters based on real-time perception of complex environments. This makes batteries highly susceptible to irreversible performance damage under high-temperature or continuous high-power output conditions. Summary of the Invention

[0005] The purpose of this invention is to overcome the above-mentioned shortcomings and provide a battery management method for an intelligent cleaning robot.

[0006] To achieve the above objectives, the specific solution of the present invention is as follows: A battery management method for an intelligent cleaning robot, comprising the following steps: S1. Obtain battery operation data, extract ambient temperature, task load and ground material from the sensors of the intelligent cleaning robot, and combine them with user habits to obtain an initial dataset; S2. Based on the initial dataset, the support vector machine algorithm is used to classify the degree of influence of the ambient temperature on the task load, determine the battery wear acceleration characteristics under high load scenarios, and obtain the classification results. S3. Based on the classification results, obtain the battery charge and discharge cycle log. If the ambient temperature exceeds a preset threshold, calculate the interaction effect between the task load and the ground material to obtain the wear acceleration index. S4. For the aforementioned degradation acceleration index, the random forest algorithm is used to predict the long-term impact of user habits on battery performance, determine the personalized degradation model, and obtain the predicted degradation curve. S5. Extract the cause of attenuation based on the predicted attenuation curve. If the task load continues to be higher than the average level, adjust the dynamic adaptation parameters to obtain the optimized battery management rules. S6. Based on the optimized battery management rules, the gradient boosting algorithm is used to integrate the user habits and the ground material to generate a personalized management solution for complex environments. S7. Monitor the battery status in real time through the personalized management scheme. If the cause of degradation matches the high load scenario, activate the strategy adjustment module to obtain the updated operating parameters. S8. From the updated operating parameters, combined with ambient temperature, determine the path to improve equipment efficiency and obtain a control mechanism for stable service life.

[0007] The present invention is further configured such that, in obtaining battery operation data, the initial dataset is obtained by extracting ambient temperature, task load, and ground material from the sensors of the intelligent cleaning robot and combining this with user habits. The ambient temperature is collected in real time by multiple temperature sensors arranged in the battery pack. The suction motor's drive current is collected by a current sensor to characterize the dust collection load in the task load; the output power of the walking motor is calculated by combining the suction motor's drive current with the voltage data collected by a voltage sensor to characterize the movement load in the task load. The ground material can be identified by capturing the attenuation characteristics of infrared reflected signals through photoelectric sensors or by identifying torque changes fed back by suction motor torque sensors. The user habits include the set of path coordinates of the intelligent cleaning robot in historical cleaning tasks, the distribution of time points when the cleaning task is started, and the power output mode selected in different cleaning areas. The initial dataset is constructed into a multi-dimensional time series matrix by associating and mapping the ambient temperature, task load, and ground material with the user habits along a unified time axis.

[0008] The present invention is further configured such that, based on the initial dataset, a support vector machine algorithm is used to classify the degree of influence of the ambient temperature on the task load, determine the battery wear acceleration characteristics under high load scenarios, and obtain classification results including: During the training phase of the support vector machine algorithm, a historical high-load scenario battery aging dataset stored in the database is used for supervised learning, wherein the kernel function of the support vector machine algorithm is selected as a radial basis function; The nonlinear feature data in the initial dataset is mapped to a high-dimensional feature space using the support vector machine algorithm, and an optimal classification hyperplane is constructed in the high-dimensional feature space. The optimal classification hyperplane is used to identify the degree of nonlinear influence of the task load on the battery voltage drop rate and internal resistance change when the ambient temperature is in different temperature ranges. The classification results include an identifier that characterizes whether the current operating condition is in a risk zone of accelerated battery wear and the corresponding confidence score; the characteristics of accelerated battery wear include the nonlinear rate of drop in battery terminal voltage, the temperature rise of the battery surface per unit time, and the increase in the voltage difference between individual cells in the battery pack.

[0009] The present invention is further configured such that, based on the classification results, a battery charge / discharge cycle log is obtained; if the ambient temperature exceeds a preset threshold, the interaction effect between the task load and the ground material is calculated to obtain a wear acceleration index, including: The battery charge / discharge cycle log is read from the local memory. The battery charge / discharge cycle log records the number of cycles, depth of discharge, and changes in available capacity corresponding to a single cleaning task. The depth of discharge is determined by the ratio of the cumulative amount of electricity consumed in a single cleaning task to the battery's rated total capacity. Obtain the current intensity corresponding to the task load and the friction resistance coefficient corresponding to the ground material. Multiply the current intensity and the friction resistance coefficient to obtain the load resistance product. Obtain the thermal index correction factor corresponding to the current ambient temperature. Then, perform a weighted summation of the load resistance product and the thermal index correction factor to obtain the loss acceleration index. The loss acceleration index is used to quantitatively characterize the combined severity of polarization and heat accumulation inside the battery under the coupled effects of high load and high temperature.

[0010] The present invention is further configured such that, for the aforementioned degradation acceleration index, the random forest algorithm is used to predict the long-term impact of user habits on battery performance, determine a personalized degradation model, and obtain a predicted degradation curve, including: Configure multiple independent decision trees, and train each decision tree in parallel based on randomly selected user habits and loss acceleration indicators; The loss acceleration index, the user habits, and the historical capacity decay sequence in the initial dataset are used as input feature vectors and input into the random forest algorithm. The personalized degradation model outputs a predicted degradation trend of the battery capacity relative to the cumulative number of cleaning tasks, thereby generating the predicted degradation curve. In the predicted degradation curve, the inflection point where the slope changes abruptly is identified to determine the range of rapid capacity decline dominated by user habits; based on the distribution of the inflection point on the time axis, the contribution rate of different influencing factors to battery life degradation is calculated, and then the weight ranking of the causes of degradation is determined.

[0011] The present invention is further configured such that, in extracting the cause of degradation based on the predicted degradation curve, and if the task load remains higher than the average level, the dynamic adaptation parameters are adjusted to obtain the optimized battery management rules, including: Based on the weighted sorting, temperature interaction dominant factor, load duration dominant factor or ground material dominant factor are identified from the causes of attenuation. The process of adjusting the dynamic adaptation parameters includes: if the weight of the load dominance factor is found to exceed a preset ratio, the upper limit of power output is lowered by modifying the pulse width modulation drive limit of the power control system. If the weight of the temperature interaction dominant factor is found to exceed a preset ratio, the charging current limit value in the charging process is reduced. The dynamic adaptation parameters also include an intermittent hibernation trigger threshold set based on the difference between the battery surface temperature and the internal estimated temperature. The optimized battery management rules differentiate and combine the power output upper limit, the intermittent sleep trigger threshold, and the charging current limit based on the weight of the cause of degradation, forming a logic control sequence with preventive protection function.

[0012] The present invention is further configured such that, based on the optimized battery management rules, the step of integrating the user habits and the ground material using a gradient boosting algorithm to generate a personalized management solution for complex environments includes: In each iteration of the gradient boosting algorithm, the residual between the actual monitored battery state parameters and the predicted values ​​of the personalized degradation model is calculated. A new weak learner is trained using the residuals and then weighted and accumulated into the current battery management rules to achieve dynamic adjustment of the weight coefficients. The personalized management solution presets differentiated power allocation strategies for different types of floor materials. Specifically, when the floor material is identified as carpet, the current allocation of the suction motor is increased, and a temperature compensation strategy is activated simultaneously. The temperature compensation strategy linearly adjusts the proportional coefficient of the motor drive current based on the deviation between the real-time ambient temperature and the historical average ambient temperature, ensuring that the battery temperature rise curve is maintained within the preset safety envelope range. The personalized management scheme also incorporates the cleaning frequency in the user's habits to adaptively update the execution cycle of the power allocation strategy.

[0013] The present invention is further configured such that, in real-time monitoring of battery status through the personalized management scheme, if the cause of degradation matches the high-load scenario, the strategy adjustment module is activated to obtain updated operating parameters including: The ambient temperature and task load of the intelligent cleaning robot are collected in real time during operation, and the current temperature rise slope and current integral value are extracted. The temperature rise slope and the current integral value are matched with the risk characteristic interval defined in the predicted decay curve; When the matching degree exceeds a preset threshold and the cause of the attenuation is determined to be the high-load scenario, the strategy adjustment module is triggered to perform the following actions: The duty cycle of the suction motor is reduced to decrease output power, and the pause time interval in the cleaning path is increased to promote heat dissipation; the updated operating parameters include the charging cut-off voltage and the discharge depth limit value corrected according to the capacity inflection point position in the predicted decay curve; wherein the charging cut-off voltage is gradually reduced as the capacity decay increases, and the discharge depth limit value is reduced by increasing the battery low voltage protection threshold to prevent the battery from entering a deep discharge state.

[0014] The present invention is further configured such that the control mechanism for determining the equipment efficiency improvement path and obtaining a stable service life by integrating the updated operating parameters with the ambient temperature includes: Based on the coupling relationship between the updated operating parameters and the current ambient temperature, an evaluation model for cleaning efficiency and temperature rise ratio is established; the equipment efficiency improvement path optimizes the operating trajectory with the lowest total energy consumption by fine-tuning the moving speed and turning radius of the walking motor while meeting the cleaning coverage requirements. The stable service life control mechanism monitors the duration of continuous high-load operation per unit time and forcibly triggers an intermittent sleep process when the duration reaches a preset limit. During the intermittent hibernation process, the intelligent cleaning robot adopts a progressive power reduction mode, first reducing the speed of the suction motor, then reducing the walking speed, until it stops running for active cooling. If the battery charge / discharge cycle log shows that high-load tasks are executed multiple times in a row, the control mechanism will automatically reduce the starting current of subsequent charging cycles to suppress battery polarization and maintain the stability of the internal chemical reaction of the battery.

[0015] The beneficial effects of this invention are as follows: By integrating multi-dimensional data such as environmental perception, task characteristics, ground attributes, and user behavior, this invention constructs a battery degradation acceleration identification mechanism for high-load and high-temperature scenarios, achieving accurate prediction of battery performance degradation trends. Based on this, by introducing a personalized degradation model and a dynamic adaptive parameter system, the battery management strategy can adaptively evolve according to the user's actual usage habits. Furthermore, by continuously optimizing the management rules through a gradient boosting algorithm and combining it with real-time monitoring and strategy adjustment modules, a closed-loop feedback battery life guarantee mechanism is formed. This method significantly improves the battery operation stability of intelligent cleaning robots in complex environments, effectively slows down the capacity degradation rate, reduces safety risks caused by thermal runaway or overload, and simultaneously balances cleaning efficiency and equipment durability, providing users with a more reliable and durable user experience. Attached Figure Description

[0016] The invention will be further illustrated with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the invention. For those skilled in the art, other drawings can be obtained based on the following drawings without any creative effort.

[0017] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation Example 1

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0019] In a battery management method for an intelligent cleaning robot according to the present invention, step S1 is first executed, which involves acquiring battery operating data. This involves extracting ambient temperature, task load, and floor material from the intelligent cleaning robot's sensors and combining this data with user habits to obtain an initial dataset. Specifically, the sensors include multiple temperature sensors, current sensors, voltage sensors, and photoelectric sensors or suction motor torque sensors arranged in the battery pack to identify floor characteristics. During the cleaning robot's task execution, the sensors collect the ambient temperature, task load, and floor material in real time. The ambient temperature is collected at a frequency of once per second to capture subtle thermal fluctuations; the task load is characterized by real-time acquisition of the suction motor's drive current and the walking motor's output power; and the floor material is categorized into different levels such as hard flooring, short-pile carpet, and long-pile carpet based on feedback changes in motor torque or the attenuation characteristics of infrared reflection signals.

[0020] The user habits include the intelligent cleaning robot's historical cleaning paths, cleaning time distribution, and power output modes. Specifically, the historical cleaning paths record the areas frequently visited by the robot and the cleaning duration for each area; the cleaning time distribution records the specific times when the user habitually starts the task, such as early morning, noon, or late at night; and the power output modes record the suction levels the user tends to use in different areas. The initial dataset stores the ambient temperature, task load, floor material, and user habits in association with timestamps. This means that each set of environmental and load parameters corresponds to a precise time coordinate and matches the floor material and user operation instructions at that time, forming a multi-dimensional time series matrix, providing a structured data foundation for subsequent algorithm analysis.

[0021] After constructing the initial dataset, step S2 is executed: based on the initial dataset, a support vector machine (SVM) algorithm is used to classify the degree of influence of ambient temperature on the task load, determine the battery wear acceleration characteristics under high-load scenarios, and obtain the classification results. The SVM algorithm uses historical high-load scenario datasets for supervised learning during the training phase, and its kernel function is a radial basis function. The logic of this algorithm lies in mapping low-dimensional, nonlinearly separable data to a high-dimensional feature space to find an optimal hyperplane to distinguish the nonlinear influence of task load on battery wear under different temperature ranges.

[0022] Specifically, the support vector machine algorithm identifies whether the impact of the same workload (such as maximum suction mode) on battery voltage drop and internal resistance changes exceeds a preset normal fluctuation range under a specific high-temperature threshold (e.g., above 35 degrees Celsius). The classification result is used to identify whether the current operating condition is in a risk zone of accelerated battery degradation. If the classifier determines that the current condition is in a risk zone of accelerated degradation, a high-risk identifier is generated. The characteristics of accelerated battery degradation under high-load scenarios include, but are not limited to: the nonlinear rate of drop in battery terminal voltage, the temperature rise of the battery surface per unit time, and the increase in single-cell pressure difference caused by uneven chemical reaction rates within the battery pack. These characteristics are defined by the classification hyperplane of the support vector machine, ensuring that the key environmental load combinations leading to battery degradation can be accurately extracted under complex operating conditions.

[0023] Next, step S3 is executed: Based on the classification results, the battery charge / discharge cycle log is obtained. If the ambient temperature exceeds a preset threshold, the interaction effect between the task load and the ground material is calculated to obtain the wear acceleration index. The battery charge / discharge cycle log records the number of charge / discharge cycles, discharge depth, and capacity change for each cleaning task. Discharge depth is defined as the percentage of electricity consumed in a single cleaning task relative to the total battery capacity. When the classification results indicate that the current area is a high-risk zone, and the real-time monitored ambient temperature exceeds a preset threshold (e.g., 40 degrees Celsius), the system initiates the interaction effect calculation logic.

[0024] The interaction effect is calculated by multiplying the task load and the ground material to obtain the wear acceleration index. Specifically, the current value of the task load is multiplied by the friction resistance coefficient corresponding to the ground material, and a weighted sum is performed incorporating an ambient temperature correction factor. The wear acceleration index characterizes the combined impact of increased battery internal resistance and heat accumulation under high-load scenarios. For example, when performing high-intensity cleaning tasks on a long-pile carpet, the risk of motor stalling increases, leading to frequent current peaks. If the ambient temperature is high at this time, the wear acceleration index generated by the interaction effect will increase significantly, indicating that severe polarization will occur inside the battery.

[0025] Then, in step S4: For the aforementioned accelerated wear index, a random forest algorithm is used to predict the long-term impact of user habits on battery performance, determining a personalized degradation model and obtaining the predicted degradation curve. The random forest algorithm is constructed using multiple decision trees, each trained on a different subset of user behavior to enhance the model's generalization ability to individual usage differences. The algorithm uses the accelerated wear index, the user habits, and historical degradation data from the initial dataset as input features.

[0026] The personalized degradation model outputs a predicted trend of battery capacity change with the cumulative number of cleaning tasks. The time dimension of the predicted degradation curve is based on the cumulative number of cleaning tasks, rather than absolute time, to more accurately reflect the impact of actual usage intensity on battery life. Specifically, the curve can mark the rapid capacity decline intervals and inflection points dominated by user habits. For example, if a user habitually activates the deep cleaning mode at 2 PM, when the daily temperature is highest, the random forest model will identify the contribution of this regular behavior to the loss of active material in the battery's negative electrode and reflect this in the predicted curve as a risk of premature battery capacity degradation under this usage mode. Based on the inflection points in the predicted degradation curve, the system further determines the weight ranking of the degradation causes, thereby clarifying whether high temperature factors or frequent high-load ground operations are dominant.

[0027] After obtaining the predicted degradation curve, step S5 is executed: The causes of degradation are extracted based on the predicted degradation curve. If the task load remains consistently higher than the average level, the dynamic adaptation parameters are adjusted to obtain optimized battery management rules. The causes of degradation include temperature-driven factors, load-driven factors, and ground material-driven factors. The system checks the average power over the past 10 consecutive cleaning tasks. If this value remains consistently higher than the average level over the entire lifespan, dynamic adaptation is triggered.

[0028] The dynamic adaptation parameters include a power output upper limit, an intermittent sleep trigger threshold, and a charging current limit. The optimized battery management rules differentiate the dynamic adaptation parameters based on the weight of the degradation causes. Specifically, if temperature interaction has the highest weight, the system will significantly lower the charging current limit to reduce the thermal effect during charging; if load continuity has the highest weight, the system will set the power output upper limit to 85% of the rated maximum power and forcibly reduce the load by limiting the maximum duty cycle of the suction motor. The optimized management rules are not merely simple parameter adjustments, but a set of preventative protection logic based on predicting degradation trends.

[0029] Next, step S6 is executed: Based on the optimized battery management rules, a gradient boosting algorithm is used to integrate the user habits and the ground material to generate a personalized management plan for complex environments. In each policy update cycle, the gradient boosting algorithm uses the user habits and the ground material as the main input variables, and iteratively calculates the negative gradient direction of the loss function to continuously correct the weight coefficients in the optimized battery management rules.

[0030] The personalized management solution includes power allocation and temperature compensation strategies for different floor material types. On hard floors, the solution tends to allocate a lower motor drive current to extend battery life per charge. When a carpeted area is detected, the solution automatically activates the corresponding power allocation strategy, allowing for short periods of high current output, but simultaneously initiating a strict temperature compensation strategy. This temperature compensation strategy dynamically adjusts the upper limit of the motor drive current based on the deviation between the current ambient temperature and the historical average temperature, ensuring that the battery surface temperature rise remains within a safe envelope. The personalized management solution periodically updates these strategies based on the user's cleaning frequency and duration, typically every 7 days, to adapt to changes in the user's lifestyle.

[0031] After generating the solution, step S7 is executed: The battery status is monitored in real time using the personalized management solution. If the cause of degradation matches the high-load scenario, the strategy adjustment module is activated to obtain updated operating parameters. Real-time monitoring of the battery status involves continuously collecting the ambient temperature and task load, and comparing them in real time with the predicted degradation curve. If the current temperature rise slope or current integral value highly matches the "accelerated degradation" interval defined in the predicted curve, the cause of degradation is determined to match the high-load scenario.

[0032] After activation, the strategy adjustment module prioritizes reducing the output power of the suction motor and simultaneously extends the pause intervals in the cleaning path to reduce the overall heat load. The module dynamically adjusts the charging cut-off voltage and discharge depth in the operating parameters based on the matched high-load scenario. Specifically, if the battery shows significant signs of accelerated degradation, the system will reduce the charging cut-off voltage from 4.2 volts to 4.15 volts to slow the structural collapse of the positive electrode material; simultaneously, the discharge depth limit will be reduced from 95% to 85% to prevent the battery from entering an overstressed state of deep discharge. Furthermore, the strategy adjustment module will push suggestive prompts to the user terminal, guiding the user to avoid performing high-load cleaning tasks during high-temperature periods. The updated operating parameters include the adjusted maximum output power and temperature protection threshold, which are written into the RAM of the electronic control system in real time for execution.

[0033] Finally, step S8 is executed: From the updated operating parameters, combined with the ambient temperature, a path to improve equipment efficiency is determined, resulting in a control mechanism for stable service life. This equipment efficiency improvement path is dynamically planned based on the matching relationship between the updated operating parameters and the current ambient temperature. Specifically, the system calculates an optimal "cleaning efficiency-temperature rise ratio," and, while ensuring cleaning effectiveness, minimizes the total energy consumption of the equipment when completing the same cleaning area by fine-tuning the speed of the walking motor and the turning radius.

[0034] The stable lifespan control mechanism limits the maximum continuous high-load operation time per unit time and automatically triggers an intermittent hibernation process when a heat accumulation trend is detected. Specifically, the intermittent hibernation trigger threshold is dynamically set based on the difference between the current battery surface temperature and the estimated internal temperature. When the difference exceeds a certain range (e.g., more than 8 degrees Celsius), it indicates that there is thermal hysteresis inside the battery, meaning that the internal temperature is much higher than the surface heat dissipation rate. At this time, the system automatically shortens the continuous working cycle and extends the cooling interval. The control mechanism also includes intervention in the charging process. The charging current limit value is automatically lowered after detecting multiple consecutive high-load tasks and gradually restored during subsequent low-load tasks to avoid exacerbating the battery polarization effect due to frequent high-current charging. The charging cutoff voltage is also adaptively adjusted according to the capacity inflection point position in the predicted decay curve. Through this closed-loop feedback system, the stability of the internal chemical reaction of the battery can be maintained, thereby achieving a stable extension of battery life under complex usage environments.

[0035] In a further refined implementation of step S1, multi-stage filtering was employed for sensor data acquisition. For ambient temperature data, median filtering was used to remove isolated points caused by instantaneous sensor disturbances. Subsequently, a first-order hysteresis filtering algorithm was applied to smooth the data curve, ensuring high confidence in the temperature information of the initial input dataset. For task load data, the current sensor sampled at a high speed of 1000 Hz, and the effective value calculation circuit converted it into a quasi-static signal reflecting the actual power consumption of the motor. Ground material recognition combined with the frequency analysis of suction motor speed fluctuations. When the robot moves on ground with different friction forces, the back electromotive force of the motor generates characteristic harmonic components. By analyzing the power spectral density of these harmonic components, the system can achieve sub-centimeter-level differentiation between hardwood floors, marble, synthetic fiber carpets, and pure cotton carpets.

[0036] In step S2, the training set for the Support Vector Machine (SVM) algorithm consists of thousands of battery aging experiment data sets under different operating conditions. These data cover the entire temperature range from 0 degrees Celsius to 50 degrees Celsius. During operation, the algorithm continuously extracts feature vectors related to ambient temperature and workload. The determination of the classification hyperplane employs the Lagrange multiplier method to solve the dual problem. By maximizing the classification margin, the algorithm ensures a rapid response to wear risks under complex dynamic operating conditions. The classification result is not only a simple binary label but also includes a confidence score, which reflects the distance of the current operating condition from the safe operating range, providing an input reference for the weighted calculation of the wear acceleration index in subsequent step S3.

[0037] For step S3, the battery charge / discharge cycle log is persistently recorded. Each time the robot returns to the charging dock, the system automatically summarizes the electrochemical parameters of the current task. The estimation of capacity change employs a combined algorithm of open-circuit voltage method and ampere-hour integral method, and is calibrated in real time based on the voltage plateau characteristics during discharge. When the ambient temperature exceeds a preset threshold, the calculation of the interaction effect considers not only the current load but also the thermal exponential function. The thermal exponential function is a value that increases exponentially with increasing temperature, multiplied by the ground material drag coefficient and the task load intensity. For example, at a high temperature of 45 degrees Celsius, the loss acceleration index produced by the same power output is 3.5 times that at 25 degrees Celsius. This index directly reveals the physical risks of electrolyte decomposition and solid electrolyte interface film thickening inside the battery under the coupling effect of high temperature and high load.

[0038] Regarding the random forest prediction process in step S4, the system pre-configures 500 independent decision trees. During the growth of each tree, approximately two-thirds of the feature vectors are randomly selected for branch selection. This mechanism effectively prevents overfitting caused by a single strongly correlated feature (such as considering only ambient temperature). The generation of the predicted degradation curve is an iterative extrapolation process. The model simulates the battery state evolution for the next 500 tasks under the current user's habits, based on the current cumulative cleaning count. Inflection points in the curve typically correspond to critical moments when increased battery capacity causes the available capacity to drop below 80% of the initial value. By determining the weighted ranking of degradation causes, the system can identify, for example, "a continuous cleaning task lasting 2 hours on a Saturday afternoon" as the most critical factor leading to shortened battery life, thus providing a precise target for parameter adjustment in step S5.

[0039] In the dynamic adaptation parameter adjustment in step S5, the power output upper limit is adjusted by modifying the PWM drive limit of the electronic control system. If the cause is determined to be continuous load dominance, the system will limit the maximum duty cycle of the suction motor to a safe range. The dynamic setting of the intermittent sleep trigger threshold introduces a thermal effect model of the battery's internal impedance. When the internal heat generation power of the battery exceeds 1.2 times the heat dissipation power of the casing, the system will forcibly reduce the walking speed, or even stop for 30 seconds for active cooling. The reduction of the charging current limit is to cooperate with this protection strategy, giving the battery structure a chance to recover by reducing the ohmic heat and chemical reaction heat during the charging phase.

[0040] In step S6, the introduction of the gradient boosting algorithm enables self-learning of the control strategy. After each cleaning task, the algorithm calculates the residual between the actual observed temperature rise and the model's predicted value, and uses this residual to train a new weak learner, which is then weighted and accumulated into the existing management rules. This approach allows the personalized management scheme to continuously evolve with changes in the environment (such as changes in indoor temperature and humidity due to seasonal changes) and the natural aging of the battery itself. The personalized scheme for complex environments not only affects the battery itself but also works in conjunction with the robot's path planning algorithm. For example, at midday in summer, the management scheme sends instructions to the path planning module to prioritize cleaning shady areas and postpone cleaning tasks in high-load areas until the evening.

[0041] During the execution of step S7, the real-time monitoring module possesses sub-second-level abnormal interruption response capability. If the voltage drop slope detected in real-time suddenly becomes steeper and matches a high-risk characteristic in the predicted decay curve, the strategy adjustment module will intervene immediately without waiting for the task to end. In the updated operating parameters, the temperature protection threshold will be dynamically tightened based on the battery's historical thermal stress accumulation. For example, if the battery has undergone multiple high-temperature operations, its protection trigger threshold may be gradually reduced from 60 degrees Celsius to 55 degrees Celsius. This progressive protection mechanism can effectively prevent the battery from entering the pre-irreversible thermal runaway stage.

[0042] The stable lifespan control mechanism described in step S8 essentially seeks a dynamic balance between battery performance and lifespan. The dynamic planning of the device efficiency improvement path utilizes a Markov decision process model to find the shortest cleaning path under specific power constraints. When a heat accumulation trend is detected, the triggered intermittent hibernation process employs a gradual power reduction mode rather than a sudden shutdown. First, the suction motor speed is gradually reduced; if the temperature rise is not contained, the walking speed is then reduced. By limiting the maximum continuous high-load operation time per unit time, the system ensures that the diffusion rate of ions inside the battery can keep up with the charge transfer rate, thereby maintaining the uniformity and stability of the internal chemical reactions. Example 2

[0043] Based on Example 1, this example further discloses the application of the battery management method of the intelligent cleaning robot in multi-machine collaboration and cloud interaction in a large-scale commercial environment.

[0044] In step S1, acquiring battery operation data is no longer limited to single-machine collection. The initial dataset also includes environmental data sharing from other cleaning robots in the same area, acquired via the cloud. For example, in a large shopping mall, multiple robots work collaboratively. If one robot detects that the underfloor heating system in a certain area is causing an abnormal increase in local ground temperature, this information will be uploaded to the cloud in real time and synchronized with other robots entering that area as a preset environmental variable in the initial dataset.

[0045] In step S2, the classification results of the support vector machine algorithm are not only used for local control but also uploaded to the cloud for global feature aggregation. The cloud server uses data from tens of thousands of robots to optimize the hyperplane parameters of the classification algorithm online and then distributes the optimized model parameters to individual robots. This swarm intelligence learning method significantly improves the accuracy of identification in extremely rare and high-load scenarios.

[0046] In step S3, an inconsistency factor between individual battery cells is introduced when calculating the accelerated loss index. The battery charge / discharge cycle log records not only the overall voltage but also the voltage consistency of each cell string. If it is found that the voltage drop rate of a certain cell string at high temperature is significantly faster than that of other cells, the weight of the interaction effect calculation will be tilted towards the weaker cell. The resulting accelerated loss index will directly reflect the health status of the worst-performing cell, thereby implementing more stringent protection.

[0047] In step S4, the personalized degradation model predicted by the random forest algorithm incorporates battery production batch information. Batteries from the same batch often exhibit similar internal resistance growth characteristics. The system identifies the batch based on the battery barcode and retrieves the standard aging model for that batch from the cloud as a foundation, then performs incremental training based on user habits. The generated predicted degradation curve can distinguish between degradation caused by battery manufacturing tolerances and accelerated wear due to improper user operation.

[0048] In step S5, the adjustment of dynamic adaptation parameters adds linkage with the charging station. The optimized battery management rules not only adjust the robot's own operating parameters but also send instructions to the charging dock to adjust its output power and fan speed. If the predicted degradation curve shows that the battery is in a rapidly decreasing capacity range, the charging dock will initiate a special "equalization repair" pulse charging mode to attempt to eliminate polarization through controlled current pulses.

[0049] In step S6, the dataset integrated by the gradient boosting algorithm is expanded to include multi-dimensional environmental data, including air humidity and air pressure. Humidity affects the motor's heat dissipation efficiency, while air pressure affects the heat exchange coefficient within the airflow channel. The personalized management solution includes a temperature rise compensation coefficient for high-humidity environments. Through continuous iteration of the gradient boosting algorithm, the solution can adapt to the battery management needs of special geographical environments such as coastal areas or high altitudes.

[0050] In step S7, the real-time monitoring module introduces a state observer based on Kalman filtering. By observing the residuals of voltage and current, the internal state of the battery, such as state of charge and state of health, is estimated in real time. When the estimated internal temperature output by the observer deviates from the measured value of the physical sensor, the strategy adjustment module determines that the sensor may have temperature drift and automatically switches to model-based safety protection parameters.

[0051] In step S8, the stable lifespan control mechanism is remotely controlled via the cloud. When a large-scale risk of premature battery lifespan termination is predicted, the cloud management platform issues a unified "maintenance mode" command, forcing all robots to reduce their operating intensity and increase automatic cooling cycles. This mechanism is particularly important in commercial leasing models, as it can effectively extend the asset's lifespan.

[0052] Furthermore, in this embodiment, a wavelet transform-based noise reduction technique is employed for the data extraction process of the sensor. The raw electrical signals acquired by the sensor contain electromagnetic interference noise generated by the motor operation. Through multi-scale wavelet decomposition, the system can accurately separate the low-frequency signals reflecting real temperature changes and load fluctuations, while filtering out high-frequency interference signals. This preprocessing step ensures that the initial dataset obtained in step S1 has an extremely high signal-to-noise ratio.

[0053] In the classification process of step S2, to improve the execution efficiency of the support vector machine algorithm, this embodiment uses the sequential minimum optimization algorithm for training. This method decomposes a large optimization problem into multiple sub-problems, enabling complex nonlinear classification calculations to be completed quickly even on the embedded processor of the cleaning robot. The classification results also include a probability prediction of entering a high-load loss zone within the next 15 minutes.

[0054] In step S3, the judgment logic for the ambient temperature exceeding a preset threshold incorporates thermal inertia compensation. Because the battery pack has a large heat capacity, the surface temperature sensed by the sensor often lags behind the internal temperature rise. The system calculates the cell center temperature in real time by establishing a second-order equivalent thermal circuit model. If the calculated center temperature exceeds the threshold, the calculation of the interaction effect is immediately triggered even if the surface temperature has not yet reached the threshold, achieving proactive protection.

[0055] Regarding the random forest model in step S4, this embodiment adds feature extraction for the frequency of switching between specific floor materials. If a user is accustomed to frequently switching between hard floors and dark plush carpets in a short period of time, the starting and braking current of the motor will generate frequent pulse impacts. The random forest model will identify this "frequent oscillating load" feature and reflect its long-term negative impact on battery mechanical stress and microcracks in the predicted decay curve.

[0056] The optimized battery management rules in step S5 incorporate adjustments to the dynamic response time constant. In extremely high-temperature environments, the rules shorten the sampling and feedback times of all control loops, ensuring millisecond-level response speeds to load fluctuations. When the ambient temperature is suitable, the rules extend the sampling period to reduce the power consumption of the control system itself, thereby optimizing the overall energy efficiency ratio.

[0057] In step S6, each iteration of the gradient boosting algorithm considers the monotonicity constraint of battery capacity decay. This means that the predicted health status will not exhibit irrational rebounds, ensuring the logical consistency of the management plan. The personalized management plan also includes a "user habit optimization suggestion" engine, which provides feedback to users through a mobile application, such as informing users that "it is recommended to perform carpet cleaning tasks after the indoor air conditioning is turned on, which can extend battery life by approximately 20%."

[0058] The updated operating parameters in step S7 include a temperature-sensitive adjustment of the discharge cut-off voltage. In low-temperature environments, the battery's internal resistance increases, and the cut-off voltage will appropriately decrease to release more charge. Conversely, in the high-temperature, high-load scenario addressed in step S7, the cut-off voltage will be increased to prevent irreversible chemical reactions in the low-voltage range. This bidirectional adjustment strategy, combined with the adjusted maximum output power, creates a comprehensive protection system.

[0059] The stable lifespan control mechanism in step S8 also incorporates battery self-discharge management. When the cleaning robot is detected to be idle for an extended period and the ambient temperature is high, the mechanism automatically initiates an "active power reduction" process to deplete the battery and maintain it at an optimal storage level (e.g., around 50%), thus avoiding electrochemical degradation caused by high-temperature, high-power storage. The entire control mechanism forms a closed-loop system covering the entire lifecycle from operation and charging to storage. Example 3

[0060] This embodiment focuses on describing the specific implementation details of the battery management method at the hardware level and the deployment optimization of the algorithm on a low-power processor to ensure the efficient execution of each step.

[0061] In step S1, the hardware link for acquiring battery operating data employs a redundant design. Each battery module is equipped with an independent acquisition unit, and data is aggregated to the main control processor via an isolated CAN bus or I2C bus. The ambient temperature information extracted by the sensors includes the air intake temperature, the battery pack center temperature, and the temperature of the power devices on the circuit board. The task load is extracted through a dedicated current monitoring chip, which has an integration function and can directly provide the processor with the ampere-hours of power consumption.

[0062] In step S2, to reduce the computational overhead of the support vector machine algorithm, the system employs model compression and quantization techniques. The pre-trained RBF kernel function model is converted into a fixed-point arithmetic format and accelerated using the processor's DSP instruction set. The classification result generation cycle is set to once every 10 seconds, maximizing processor efficiency while maintaining real-time performance.

[0063] In step S3, the calculation logic for the loss acceleration index is hard-coded in the firmware's protective layer. The weighted product operation used in the calculation process is converted into a combination of displacement operation and lookup table method. For example, for the drag coefficient of different ground materials, the system pre-stores a two-dimensional lookup table, directly obtaining the corresponding coefficient based on the motor torque feedback value, thereby eliminating complex real-time multiplication calculations.

[0064] In step S4, the random forest algorithm leverages the structural characteristics of decision trees and employs an efficient pruning strategy. By removing weak feature branches that contribute less than 5% to the prediction results, the model size is reduced to one-third of its original size, making it suitable for embedded flash memory with only a few hundred KB of storage space. The generation of the prediction decay curve uses an incremental update mechanism, recalculating the full curve only when significant changes in user behavior are detected.

[0065] In step S5, the adjustment of the dynamic adaptation parameters is achieved through the parameter mirroring mechanism of the electronic control system. The optimized battery management rules are first written into a shadow register, and then updated to the main control register through atomic operations during system idle time, avoiding motor operation jitter that may occur during parameter update.

[0066] In step S6, the iterative computation of the gradient boosting algorithm is distributed across multiple task cycles. Only one decision tree residual is calculated in each cycle, and a complete solution update is completed after dozens of cycles. This distributed computing strategy ensures that the algorithm's operation does not preempt the robot's highly real-time tasks such as navigation and obstacle avoidance.

[0067] In step S7, the real-time monitoring module utilizes the processor's hardware comparator function. By setting the thresholds for ambient temperature and workload as inputs to the hardware comparator, the comparator immediately triggers a system interrupt and activates the strategy adjustment module once the signal exceeds the red line defined by the predicted attenuation curve. This hardware-level response mechanism ensures that the system can complete shutdown or power reduction operations within one millisecond under extreme abnormal conditions.

[0068] In step S8, the calculation of the device efficiency improvement path incorporates the robot's kinematic model. By dynamically adjusting the PID control parameters of the walking wheel motors, the robot avoids the high-rate discharge zone of the battery as much as possible during acceleration and deceleration. In the control mechanism for stabilizing lifespan, the logic of intermittent hibernation is integrated into the task scheduler. The system automatically inserts necessary "cooling tasks" into the cleaning sequence based on the battery's real-time thermal history, such as reducing the moving speed on the path back to the charging dock.

[0069] Furthermore, this embodiment also discloses a self-calibration method for the aforementioned accelerated wear index. After each charging cycle, the system measures the battery's voltage recovery characteristics under standard load. If the measured recovery voltage deviates from the theoretical value, it indicates a systematic error in the current accelerated wear index calculation model. In this case, the system automatically adjusts the weighted product coefficients in step S3. Through this online self-calibration mechanism, the accuracy of the management method throughout the battery's entire lifespan is ensured.

[0070] For storing the initial dataset in step S1, this embodiment employs a circular buffer storage structure. The buffer size is set according to the user's preferred period (e.g., historical data from the last 30 days). When new data is stored, the oldest data is overwritten, ensuring constant memory usage. The data undergoes Huffman compression before storage, effectively improving data storage density and providing space for long-term algorithm analysis.

[0071] In step S4, the determination of the personalized degradation model also incorporates an equivalent estimation of the battery impedance spectrum. By analyzing the battery's voltage response under different frequency loads, the ohmic internal resistance, charge transfer resistance, and diffusion impedance are estimated. These electrochemical parameters are used as high-level input features for the random forest algorithm, enabling the predicted degradation curve to capture the battery's performance degradation trend at a microscopic level.

[0072] The optimized battery management rules in step S5 also include an "emergency protection" subset. When the system detects that the battery temperature is approaching a critical point, the rules will disregard all cleaning efficiency requirements, forcibly shut down the suction motor, and only maintain the walking motor running at low speed to guide the robot to a well-ventilated area. This tiered response mechanism slows down the aging process of the battery to the greatest extent possible while ensuring safety.

[0073] In step S6, the gradient boosting algorithm employs a "smooth transition" technique for the periodic updates of the strategy. The switch between the old and new strategies is not abrupt, but rather completed gradually over several minutes through a linear interpolation function, ensuring the continuity of the robot's cleaning performance and making the adjustment of the management strategy imperceptible to the user.

[0074] The real-time monitoring in step S7 also includes monitoring the battery pack pressure. Under high temperature and high load conditions, if even the slightest battery venting occurs, the pressure sensor will detect abnormal pressure fluctuations. Upon receiving this signal, the strategy adjustment module will update the operating parameters to "extreme limitation mode," retaining only the most basic communication functions and issuing a safety warning to the user.

[0075] The stable lifespan control mechanism in step S8 guides the battery to "shallow charge and discharge" through intelligent adjustment of the charging logic. If the predicted degradation curve shows that the battery has entered a sensitive aging period, the system will perform a small-current replenishment operation during non-cleaning tasks, taking advantage of peak solar power generation or other low electricity cost periods, to maintain the battery charge in the range of 40% to 60% where chemical stability is highest.

[0076] This invention constructs a multi-layered, cross-dimensional intelligent cleaning robot battery management system through the methods described in Examples 1, 2, and 3. From single-machine sensor perception to cloud-based swarm intelligence, from underlying algorithm optimization to upper-level strategy evolution, it comprehensively covers the operational needs of intelligent cleaning robots in high-temperature, high-load, and complex ground environments. This method not only achieves accurate identification and prediction of battery wear but also provides refined protection for the battery throughout its entire lifecycle through closed-loop parameter adjustment and efficiency path planning, significantly extending the stable service life of the equipment.

[0077] The above description is only a preferred embodiment of the present invention. Therefore, any equivalent changes or modifications made to the structure, features and principles described in the claims of this patent application are included within the protection scope of this patent application.

Claims

1. A battery management method for an intelligent cleaning robot, characterized in that: Includes the following steps: S1. Obtain battery operation data, extract ambient temperature, task load and ground material from the sensors of the intelligent cleaning robot, and combine them with user habits to obtain an initial dataset; S2. Based on the initial dataset, the support vector machine algorithm is used to classify the degree of influence of the ambient temperature on the task load, determine the battery wear acceleration characteristics under high load scenarios, and obtain the classification results. S3. Based on the classification results, obtain the battery charge and discharge cycle log. If the ambient temperature exceeds a preset threshold, calculate the interaction effect between the task load and the ground material to obtain the wear acceleration index. S4. For the aforementioned degradation acceleration index, the random forest algorithm is used to predict the long-term impact of user habits on battery performance, determine the personalized degradation model, and obtain the predicted degradation curve. S5. Extract the cause of attenuation based on the predicted attenuation curve. If the task load continues to be higher than the average level, adjust the dynamic adaptation parameters to obtain the optimized battery management rules. S6. Based on the optimized battery management rules, the gradient boosting algorithm is used to integrate the user habits and the ground material to generate a personalized management solution for complex environments. S7. Monitor the battery status in real time through the personalized management scheme. If the cause of degradation matches the high load scenario, activate the strategy adjustment module to obtain the updated operating parameters. S8. From the updated operating parameters, combined with ambient temperature, determine the path to improve equipment efficiency and obtain a control mechanism for stable service life.

2. The battery management method for an intelligent cleaning robot according to claim 1, characterized in that: The process of acquiring battery operation data involves extracting ambient temperature, workload, and ground material from the sensors of the intelligent cleaning robot, and combining this data with user habits to obtain an initial dataset, including: The ambient temperature is collected in real time by multiple temperature sensors arranged in the battery pack. The suction motor's drive current is collected by a current sensor to characterize the dust collection load in the task load; the output power of the walking motor is calculated by combining the suction motor's drive current with the voltage data collected by a voltage sensor to characterize the movement load in the task load. The ground material can be identified by capturing the attenuation characteristics of infrared reflected signals through photoelectric sensors or by identifying torque changes fed back by suction motor torque sensors. The user habits include the set of path coordinates of the intelligent cleaning robot in historical cleaning tasks, the distribution of time points when the cleaning task is started, and the power output mode selected in different cleaning areas. The initial dataset is constructed into a multi-dimensional time series matrix by associating and mapping the ambient temperature, task load, and ground material with the user habits along a unified time axis.

3. The battery management method for an intelligent cleaning robot according to claim 1, characterized in that: Based on the initial dataset, the support vector machine algorithm is used to classify the impact of ambient temperature on the task load, determine the battery wear acceleration characteristics under high load scenarios, and obtain the following classification results: During the training phase of the support vector machine algorithm, a battery aging dataset from historical high-load scenarios stored in the database is used for supervised learning. The kernel function of the support vector machine algorithm is selected as the radial basis function. The nonlinear feature data in the initial dataset is mapped to a high-dimensional feature space using the support vector machine algorithm, and an optimal classification hyperplane is constructed in the high-dimensional feature space. The optimal classification hyperplane is used to identify the degree of nonlinear influence of the task load on the battery voltage drop rate and internal resistance change when the ambient temperature is in different temperature ranges. The classification results include an identifier that characterizes whether the current operating condition is in a risk zone of accelerated battery wear and the corresponding confidence score; the characteristics of accelerated battery wear include the nonlinear rate of drop in battery terminal voltage, the temperature rise of the battery surface per unit time, and the increase in the voltage difference between individual cells in the battery pack.

4. The battery management method for an intelligent cleaning robot according to claim 1, characterized in that: Based on the classification results, a battery charge / discharge cycle log is obtained. If the ambient temperature exceeds a preset threshold, the interaction effect between the task load and the ground material is calculated to obtain the wear acceleration index, including: The battery charge / discharge cycle log is read from the local memory. The battery charge / discharge cycle log records the number of cycles, depth of discharge, and the amount of capacity decay corresponding to a single cleaning task. The depth of discharge is determined by the ratio of the cumulative amount of electricity consumed in a single cleaning task to the battery's rated total capacity. Obtain the current intensity corresponding to the task load and the friction resistance coefficient corresponding to the ground material. Multiply the current intensity and the friction resistance coefficient to obtain the load resistance product. Obtain the thermal index correction factor corresponding to the current ambient temperature. Then, perform a weighted summation of the load resistance product and the thermal index correction factor to obtain the loss acceleration index. The loss acceleration index is used to quantitatively characterize the combined severity of polarization and heat accumulation inside the battery under the coupled effects of high load and high temperature.

5. The battery management method for an intelligent cleaning robot according to claim 1, characterized in that: Regarding the aforementioned accelerated degradation index, the random forest algorithm is used to predict the long-term impact of user habits on battery performance, determine a personalized degradation model, and obtain the predicted degradation curve, including: Configure multiple independent decision trees, and train each decision tree in parallel based on randomly selected user habits and loss acceleration indicators; The loss acceleration index, the user habits, and the historical capacity decay sequence in the initial dataset are used as input feature vectors and input into the random forest algorithm. The personalized degradation model outputs a predicted degradation trend of the battery capacity relative to the cumulative number of cleaning tasks, thereby generating the predicted degradation curve. In the predicted degradation curve, the inflection point where the slope changes abruptly is identified to determine the range of rapid capacity decline dominated by user habits; based on the distribution of the inflection point on the time axis, the contribution rate of different influencing factors to battery life degradation is calculated, and then the weight ranking of the causes of degradation is determined.

6. The battery management method for an intelligent cleaning robot according to claim 1, characterized in that: The step of extracting the cause of degradation based on the predicted degradation curve, and adjusting the dynamic adaptation parameters to obtain optimized battery management rules if the task load remains above the average level, includes: Based on the weighted sorting, temperature interaction dominant factor, load duration dominant factor or ground material dominant factor are identified from the causes of attenuation. If the weight of the load dominance factor is found to exceed a preset ratio, the upper limit of power output is lowered by modifying the pulse width modulation drive limit of the electronic control system. If the weight of the temperature interaction dominant factor is found to exceed a preset ratio, the charging current limit value in the charging process is reduced. The dynamic adaptation parameters also include an intermittent hibernation trigger threshold set based on the difference between the battery surface temperature and the internal estimated temperature. The optimized battery management rules differentiate and combine the power output upper limit, the intermittent sleep trigger threshold, and the charging current limit based on the weight of the cause of degradation, forming a logic control sequence with preventive protection function.

7. The battery management method for an intelligent cleaning robot according to claim 1, characterized in that: The step of integrating user habits and ground material using a gradient boosting algorithm based on the optimized battery management rules to generate a personalized management solution for complex environments includes: In each iteration of the gradient boosting algorithm, the residual between the actual monitored battery state parameters and the predicted values ​​of the personalized degradation model is calculated. A new weak learner is trained using the residuals and then weighted and accumulated into the current battery management rules to achieve dynamic adjustment of the weight coefficients. The personalized management solution presets differentiated power allocation strategies for different types of floor materials. When the floor material is identified as carpet, the current allocation of the suction motor is increased, and a temperature compensation strategy is activated simultaneously. The temperature compensation strategy linearly adjusts the proportional coefficient of the suction motor drive current based on the deviation between the real-time ambient temperature and the historical average ambient temperature, ensuring that the battery temperature rise curve is maintained within the preset safety envelope range. The personalized management scheme also incorporates the cleaning frequency in the user's habits to adaptively update the execution cycle of the power allocation strategy.

8. The battery management method for an intelligent cleaning robot according to claim 1, characterized in that: The personalized management scheme monitors the battery status in real time. If the cause of battery degradation matches the high-load scenario, the strategy adjustment module is activated to obtain updated operating parameters, including: The ambient temperature and task load of the intelligent cleaning robot are collected in real time during operation, and the current temperature rise slope and current integral value are extracted. The temperature rise slope and the current integral value are matched with the risk characteristic interval defined in the predicted decay curve; When the matching degree exceeds the preset threshold and it is determined that the cause of the attenuation belongs to the high load scenario, the strategy adjustment module is triggered to perform an action, reduce the duty cycle of the suction motor to reduce the output power, and increase the pause time interval in the cleaning path to promote heat dissipation. The updated operating parameters include a charging cut-off voltage and a discharge depth limit value corrected based on the capacity inflection point position in the predicted degradation curve; the charging cut-off voltage is gradually reduced as the capacity degradation increases, and the discharge depth limit value is reduced by increasing the battery low voltage protection threshold to prevent the battery from entering a deep discharge state.

9. A battery management method for an intelligent cleaning robot according to claim 1, characterized in that: The control mechanism that determines the equipment efficiency improvement path and obtains a stable service life by integrating the updated operating parameters with ambient temperature includes: Based on the coupling relationship between the updated operating parameters and the current ambient temperature, an evaluation model for cleaning efficiency and temperature rise ratio is established; the equipment efficiency improvement path optimizes the operating trajectory with the lowest total energy consumption by adjusting the moving speed and turning radius of the walking motor while meeting the cleaning coverage requirements. The stable service life control mechanism monitors the duration of continuous high-load operation per unit time and forcibly triggers an intermittent sleep process when the duration reaches a preset limit. During the intermittent hibernation process, the intelligent cleaning robot adopts a progressive power reduction mode, first reducing the speed of the suction motor, then reducing the walking speed, until it stops running for active cooling. If the battery charge / discharge cycle log shows that high-load tasks are executed multiple times in a row, the control mechanism will automatically reduce the starting current of subsequent charging cycles to suppress battery polarization and maintain the stability of the internal chemical reaction of the battery.

Images