TW202221625A - Method for predicting power consumption of uav and uav using the same - Google Patents

Abstract

A method for predicting power consumption of an unmanned aerial vehicle (UAV) and a UAV using the same. The method includes: obtaining historical climate data and current climate data according to at least one sensor and a transceiver; monitoring power consumption of the UAV to obtain historical power consumption corresponding to the historical climate data; training a machine learning (ML) model according to the historical climate data and the historical power consumption, and inputting the current climate data to the ML model to generate predicted power consumption; and outputting the predicted power consumption.

Description

Method for predicting power consumption of unmanned aerial vehicle and unmanned aerial vehicle using said method

The present invention relates to a method for predicting the power consumption of an unmanned aerial vehicle and an unmanned aerial vehicle (UAV) using the method.

Generally speaking, the power of the drone will decrease with the usage time. The power consumption of a drone is affected by a variety of factors, including the hardware capabilities of the drone, the mission or payload, etc. In addition to this, the power consumption of drones is also vulnerable to climate impacts. However, the current technology cannot accurately predict the power consumption of the drone by grasping the influence of the climate on the power consumption of the drone. Therefore, how to predict the power consumption of UAVs according to different climates is an important topic in this field.

The present invention provides a method for predicting the power consumption of an unmanned aerial vehicle and an unmanned aerial vehicle using the method, which can predict the power consumption of the unmanned aerial vehicle according to climate data.

An unmanned aerial vehicle of the present invention includes a processor, a storage medium, a transceiver and at least one sensor. The storage medium stores multiple modules. The processor is coupled to the storage medium, at least one sensor and the transceiver, and accesses and executes a plurality of modules, wherein the plurality of modules include a data collection module, a power monitoring module, a power consumption prediction module and an output module. The data collection module obtains historical climate data and current climate data according to at least one of the at least one sensor and the transceiver. The power monitoring module monitors the power consumption of the drone to obtain historical power consumption corresponding to historical climate data. The power consumption prediction module trains the machine learning model according to the historical climate data and the historical power consumption, and inputs the current climate data into the machine learning model to generate the predicted power consumption. The output module predicts the power consumption through the transceiver output.

In an embodiment of the present invention, the above-mentioned current climate data corresponds to a first period, wherein the power monitoring module monitors the power consumption of the drone to obtain the previous power consumption corresponding to the second period, wherein the second period is earlier In the first period, the power consumption prediction module inputs the current climate data and the previous power consumption into the machine learning model to generate the predicted power consumption.

In an embodiment of the present invention, the above-mentioned data collection module obtains the specification data of the UAV through the transceiver, wherein the power consumption prediction module trains the machine learning model according to the specification data, historical climate data and historical power consumption, wherein The power consumption prediction module inputs current climate data and specification data into the machine learning model to generate predicted power consumption.

In an embodiment of the present invention, the above-mentioned data collection module obtains the historical flight instruction set and the current flight instruction set through the transceiver, wherein the power consumption prediction module is based on the historical flight instruction set, historical climate data and historical power consumption The machine learning model is trained, wherein the power consumption prediction module inputs the current flight instruction set and the current climate data into the machine learning model to generate the predicted power consumption.

In an embodiment of the present invention, the above-mentioned current flight instruction set is associated with at least one of the following instructions: flight attitude, flight speed and altitude, wherein the flight speed includes horizontal speed and vertical speed.

In an embodiment of the present invention, the above-mentioned at least one sensor includes a thermometer and a hygrometer, wherein the current climate data includes temperature and humidity.

In an embodiment of the present invention, the above-mentioned at least one sensor includes a barometer, an anemometer and an anemometer, wherein the current climate data includes air pressure, wind speed and wind direction.

In an embodiment of the present invention, the above-mentioned modules further include a data reading and correction module. The data reading and correction module removes outliers in historical climate data, current climate data, and historical power consumption to update historical climate data, current climate data, and historical power consumption.

A method for predicting the power consumption of an unmanned aerial vehicle of the present invention includes: obtaining historical climate data and current climate data according to at least one of at least one sensor and a transceiver; monitoring the power consumption of the unmanned aerial vehicle to obtain corresponding historical power consumption of historical climate data; train a machine learning model according to historical climate data and historical power consumption, and input current climate data to the machine learning model to generate predicted power consumption; and output predicted power consumption.

Based on the above, the present invention can train a machine learning model according to historical climate data and historical power consumption, and generate the predicted power consumption of the UAV according to the machine learning model.

In order to make the content of the present invention more comprehensible, the following specific embodiments are given as examples according to which the present invention can indeed be implemented. Additionally, where possible, elements/components/steps using the same reference numerals in the drawings and embodiments represent the same or similar parts.

FIG. 1 is a schematic diagram of an unmanned aerial vehicle 100 according to an embodiment of the present invention. The drone 100 may include a processor 110 , a storage medium 120 , a transceiver 130 , and a sensor 140 .

The processor 110 is, for example, a central processing unit (CPU), or other programmable general-purpose or special-purpose micro control unit (micro control unit, MCU), microprocessor (microprocessor), digital signal processing digital signal processor (DSP), programmable controller, application specific integrated circuit (ASIC), graphics processor (graphics processing unit, GPU), image signal processor (image signal processor, ISP) ), image processing unit (IPU), arithmetic logic unit (ALU), complex programmable logic device (CPLD), field programmable gate array (field programmable gate array) , FPGA) or other similar elements or a combination of the above. The processor 110 may be coupled to the storage medium 120 and the transceiver 130 , and access and execute a plurality of modules and various application programs stored in the storage medium 120 .

The storage medium 120 is, for example, any type of fixed or removable random access memory (random access memory, RAM), read-only memory (ROM), and flash memory (flash memory). , a hard disk drive (HDD), a solid state drive (SSD), or similar components or a combination of the above components for storing a plurality of modules or various application programs executable by the processor 110 . In this embodiment, the storage medium 120 can store a plurality of modules including a data collection module 121 , a power monitoring module 122 , a data reading and correction module 123 , a power consumption prediction module 124 and an output module 125 . , its function will be explained later. The storage medium 120 may include, for example, a relational database (RDB), a non-relational database, a distributed database, a search engine database, a directory service (lightweight directory access protocol, LDAP), a file, or storage such as memory.

The transceiver 130 transmits and receives signals in a wireless or wired manner. Transceiver 130 may also perform operations such as low noise amplification, impedance matching, frequency mixing, up or down frequency conversion, filtering, amplification, and the like.

Sensors 140 may include one or more sensors associated with measurements of climate data. In one embodiment, the sensor 140 may include a thermometer 141 , a barometer 142 , an anemometer 143 , an anemometer 144 or a hygrometer 145 , but the invention is not limited thereto. The data collection module 121 can obtain historical climate data and current climate data through the transceiver 130 or the sensor 140 . Historical climate data or current climate data may be associated with temperature, air pressure, wind speed, wind direction or humidity, but the invention is not limited thereto. For example, the sensor 140 can measure through the sensor 140 to obtain information such as temperature, air pressure, wind speed, wind direction or humidity. For another example, the sensor 140 can obtain information such as temperature, air pressure, wind speed, wind direction or humidity from the website of the Weather Bureau through the transceiver 130 . For another example, the sensor 140 can receive information such as temperature, air pressure, wind speed, wind direction or humidity input by the user from the terminal device through the transceiver 130 . The time period corresponding to the historical climate data is earlier than the time period corresponding to the current climate data.

The power monitoring module 122 can monitor the power consumption of the drone 100 to obtain historical power consumption corresponding to historical climate data. After obtaining historical climate data and historical power consumption, the power consumption prediction module 124 can train a machine learning model according to the historical climate data and historical power consumption.

In one embodiment, the data collection module 121 can obtain the specification data of the UAV 100 through the transceiver 130 . The power consumption prediction module 124 can train the machine learning model according to the specification data, historical climate data and historical power consumption.

In one embodiment, the data collection module 121 can obtain the historical flight instruction set corresponding to the historical climate data through the transceiver 130 . The power consumption prediction module 124 can train the machine learning model according to the historical flight instruction set, historical climate data and historical power consumption. The historical flight instruction set may be associated with flight instructions such as flight attitude, flight speed or altitude, wherein the flight speed may include horizontal speed or vertical speed.

In one embodiment, before the power consumption prediction module 124 trains the machine learning model according to historical climate data, historical power consumption, specification data or historical flight instruction sets, the data reading and correction module 123 may first remove the historical data Climate data, historical power consumption, specification data or abnormal values in the historical flight instruction set, so that the machine learning model trained by the power consumption prediction module 124 is more accurate.

In order to remove a data point from a data set, in one embodiment, the data reading and correction module 123 may determine that the data point is an outlier based on whether the data point is larger than the upper limit or smaller than the lower limit, and remove the data from the data set point (or update the data point). In one embodiment, the data reading and correction module 123 can remove outliers (or update the data points) in the data set through the normal distribution method. Specifically, the data reading and correction module 123 can establish the normal distribution curve of the data set. Since about 99.7% of the values are distributed within 3 standard deviations from the mean of the data set, the data reading and correction module 123 can remove or correct data points in the data set that are more than 3 standard deviations away. Corrected datasets (ie: corrected historical climate data, historical power consumption, specification data, or historical flight instruction sets) can be used to train machine learning models.

The machine learning model is, for example, a neural network-like model. A neural network-like model can include an input layer, a hidden layer, and an output layer. The power consumption prediction module 124 can divide the revised data set into a training data set, a validation data set and a test data set. The input layer can receive training datasets, validation datasets, or test datasets. The output layer can output the training results or calculation results of the neural network-like model. The hidden layer is located between the input layer and the output layer. The input layer, the hidden layer and the output layer can respectively include one or more nodes, wherein the one or more nodes can be changed according to the user's needs. One or more nodes in the input layer can be respectively connected to one or more nodes in the hidden layer, and one or more nodes in the hidden layer can be respectively connected to one or more nodes in the output layer. When the power consumption prediction module 124 performs the training of the neural network-like model, the power consumption prediction module 124 may first use the training data set to train the neural network-like model. Next, the power consumption prediction module 124 can use the verification data set to evaluate the training status of the neural network model, determine whether the neural network model is overfitting, and adjust the parameters of the neural network model according to the judgment result to Generate the best neural network-like model. After the neural network-like model is fully verified, the power consumption prediction module 124 can use the test data set to evaluate the performance of the final neural network-like model.

如公式（1）所示，並且類神經網路模型的輸出值

如公式（2）所示，其中n為正整數。耗電量預測模組124可根據如公式（3）所示的平均絕對誤差法（mean absolute error，MAE）、如公式（4）所示的平均絕對百分比誤差法（mean absolute percentage error，MAPE）或如公式（5）所示的均方根誤差法（root mean square error，RMSE）來計算類神經網路模型的損失函數f。損失函數f的值越小，代表類神經網路模型的效能越好。若損失函數f的值過大，則耗電量預測模組124可調整類神經網路模型中各個層（即：輸入層、隱藏層和輸出層）之中的節點的數量，並重新訓練類神經網路模型。

…(1)

…(2)

…(3)

…(4)

…(5) Hypothesis test dataset

As shown in formula (1), and the output value of the neural network-like model

As shown in formula (2), where n is a positive integer. The power consumption prediction module 124 can be based on the mean absolute error (MAE) method shown in formula (3) and the mean absolute percentage error (MAPE) method shown in formula (4). Or the root mean square error (RMSE) as shown in formula (5) to calculate the loss function f of the neural network-like model. The smaller the value of the loss function f, the better the performance of the neural network-like model. If the value of the loss function f is too large, the power consumption prediction module 124 can adjust the number of nodes in each layer (ie, the input layer, the hidden layer, and the output layer) in the neural network-like model, and retrain the neural network-like model. network model.

…(1)

…(2)

…(3)

…(4)

…(5)

After the power consumption prediction module 124 generates the machine learning model, the power consumption prediction module 124 can input the current climate data into the machine learning model to generate the predicted power consumption. The output module 125 can output the predicted power consumption through the transceiver 130 . For example, the output module 125 can transmit the predicted power consumption to the terminal device for controlling the drone 100 through the transceiver 130 . The terminal device may output the predicted power consumption through an output device (eg, a display) for reference by the operator or manager of the drone 100 . In one embodiment, before inputting the current climate data into the machine learning model, the data reading and correction module 123 may remove or update outliers in the current climate data.

In one embodiment, the predicted power consumption of the drone 100 may be related to the specification data of the drone 100 . The power consumption prediction module 124 may input specification data and current climate data into the machine learning model to generate predicted power consumption.

In one embodiment, the predicted power consumption of the drone 100 may be related to the flight instruction set. The power consumption prediction module 124 can input the current flight instruction set and the current climate data into the machine learning model to generate the predicted power consumption. In one embodiment, before inputting the previous flight instruction set into the machine learning model, the data reading and correction module 123 may remove or update outliers in the current flight instruction set.

In one embodiment, the predicted power consumption of the drone 100 may be related to the previous power consumption. Specifically, assuming that the current climate data corresponds to the first period, the power monitoring module 122 can monitor the power consumption of the drone 100 to obtain the previous power consumption corresponding to the second period, wherein the second period is earlier than the first period time period. The power consumption prediction module 124 can input the previous power consumption and the current climate data into the machine learning model to generate the predicted power consumption. In one embodiment, before inputting the previous power consumption into the machine learning model, the data reading and correction module 123 may remove or update the abnormal value in the previous power consumption.

FIG. 2 is a flowchart of a method for predicting power consumption of an unmanned aerial vehicle according to an embodiment of the present invention, wherein the method can be implemented by the unmanned aerial vehicle 100 shown in FIG. 1 . In step S201, historical climate data and current climate data are obtained according to at least one of the at least one sensor and the transceiver. In step S202, the power consumption of the drone is monitored to obtain the historical power consumption corresponding to the historical climate data. In step S203, the machine learning model is trained according to the historical climate data and the historical power consumption, and the current climate data is input into the machine learning model to generate the predicted power consumption. In step S204, the predicted power consumption is output.

To sum up, the present invention not only utilizes various flight status data and historical power consumption data of the UAV, but also integrates current weather data. The present invention can predict the power consumption by importing the current climate data into the prediction model, so it can provide a better prediction reference for the controller of the UAV to evaluate the power consumption required to perform various complicated flight tasks, and can influence the UAV to perform flight tasks in advance. The worst weather factors are dealt with in advance, thereby effectively reducing the occurrence of flight accidents.

100: Drone 110: Processor 120: Storage Media 121: Data Collection Module 122: Power monitoring module 123: Data reading and correction module 124: Power consumption prediction module 125: Output module 130: Transceiver 140: Sensor 141: Thermometer 142: Barometer 143: Anemometer 144: Anemometer 145: Hygrometer S201, S202, S203, S204: steps

FIG. 1 is a schematic diagram of an unmanned aerial vehicle according to an embodiment of the present invention. FIG. 2 is a flowchart illustrating a method for predicting power consumption of an unmanned aerial vehicle according to an embodiment of the present invention.

S201, S202, S203, S204: steps

Claims (9)

A drone comprising: transceiver; at least one sensor; storage media, storing multiple modules; and a processor, coupled to the storage medium, the at least one sensor and the transceiver, and accesses and executes the multiple modules, wherein the multiple modules include: a data collection module for obtaining historical climate data and current climate data according to at least one of the at least one sensor and the transceiver; a power monitoring module, which monitors the power consumption of the drone to obtain historical power consumption corresponding to the historical climate data; a power consumption prediction module, which trains a machine learning model according to the historical climate data and the historical power consumption, and inputs the current climate data into the machine learning model to generate predicted power consumption; and The output module outputs the predicted power consumption through the transceiver. The drone of claim 1, wherein The current climate data corresponds to a first time period, where The power monitoring module monitors the power consumption of the drone to obtain previous power consumption corresponding to a second period, wherein the second period is earlier than the first period, wherein The power consumption prediction module inputs the current climate data and the previous power consumption into the machine learning model to generate the predicted power consumption. The drone of claim 1, wherein The data collection module obtains the specification data of the UAV through the transceiver, wherein The power consumption prediction module trains the machine learning model according to the specification data, the historical climate data and the historical power consumption, wherein The power consumption prediction module inputs the current climate data and the specification data into the machine learning model to generate the predicted power consumption. The drone of claim 1, wherein The data collection module obtains the historical flight instruction set and the current flight instruction set through the transceiver, wherein The power consumption prediction module trains the machine learning model according to the historical flight instruction set, the historical climate data and the historical power consumption, wherein The power consumption prediction module inputs the current flight instruction set and the current climate data into the machine learning model to generate the predicted power consumption. The drone of claim 4, wherein the current flight instruction set is associated with at least one of the following instructions: Flight attitude, flight speed and altitude, wherein the flight speed includes horizontal speed and vertical speed. The drone of claim 1, wherein the at least one sensor includes a thermometer and a hygrometer, and wherein the current climate data includes temperature and humidity. The drone of claim 1, wherein the at least one sensor includes a barometer, an anemometer, and an anemometer, and wherein the current climate data includes air pressure, wind speed, and wind direction. The drone of claim 1, wherein the plurality of modules further comprise: A data reading and correction module to remove abnormal values in the historical climate data, the current climate data and the historical power consumption to update the historical climate data, the current climate data and the historical power consumption power. A method of predicting the power consumption of a drone, comprising: Obtain historical climate data and current climate data according to at least one of the at least one sensor and the transceiver; monitoring the power consumption of the drone to obtain historical power consumption corresponding to the historical climate data; Train a machine learning model based on the historical climate data and the historical power consumption, and input the current climate data into the machine learning model to generate predicted power consumption; and The predicted power consumption is output.

Images