A control method and a path control method for an on-site sales robot

Abstract

The present application provides a control method for an on-site sales robot. The method includes: a collection module collects signals in an area of interest, where the area of interest is an area of interest related to motor imagery, and sends the collected signals to a signal processing module; the signal processing module converts the collected signals into digital signals through an analog-to-digital converter; uses an AI large model to preprocess the converted digital signals, extracts the high-order features and patterns of the signals; performs feature extraction on the signals based on discrete wavelet transform; performs pattern recognition on the signals according to the extracted features to obtain the motion instructions that the user wants to give to the on-site sales robot; wirelessly transmits the motion instructions that the user wants to give to the on-site sales robot to the single-chip microcomputer on the on-site sales robot through a wireless interface; the single-chip microcomputer on the sales robot sends control information to the drive circuit board according to the motion instructions, so that the single-chip microcomputer on the sales robot controls the movement of the sales robot.

Description

Technical Field

[0001] The present invention relates to the field of automation technology, and particularly to a control method and a path control method for an on-site sales robot. Background Art

[0002] With the continuous development of technology, robots are increasingly widely used in the commercial field. In the real estate sales industry, the current position is sensed by a sensor device installed inside the sales robot, and after obtaining various data of obstacles through image recognition technology and combining the target point position data to obtain obstacle avoidance parameters to avoid obstacles, the optimal path from the current position to the target position is planned. Unmanned driving technology and the movement of controllable sliding objects by mind have extremely broad prospects, and the integration of the two is bound to become a new trend and will usher in a booming period in the future. The movement of controllable sliding objects by mind is widely used in medical treatment, big health, game entertainment, and smart home. It can be predicted that in the near future, it will enter our lives.

[0003] In the prior art, when a remote on-site sales robot is running, the user needs to hold a remote controller to control the operation of the on-site sales robot. There are too many accessories and it is not convenient to carry. If the user can control the operation of the on-site sales robot through his own mind, the user can control the on-site sales robot at will. During the running process of the on-site sales robot, the user does not need to hold a remote controller, which improves the operation convenience of the on-site sales robot, and can also avoid the problem that the on-site sales robot cannot be controlled due to the loss of the remote controller. Although the existing on-site sales robots have solved the problem of the planned route of the running path to a certain extent, there are some problems in path control, such as unreasonable path planning and poor obstacle avoidance ability. There is a lack of path planning to avoid obstacles from the current position to the target position, and there is a lack of control of the speed of the on-site sales robot when encountering obstacles through obstacle avoidance parameters obtained from various data values of obstacles, which affects the working efficiency of the robot and the customer experience. Summary of the Invention

[0004] The present invention provides a control method and a path control method for an on-site sales robot, realizing the free control of the on-site sales robot by the user's mind.

[0005] On the one hand, the present application provides a control method for an on-site sales robot. The method includes: Signal acquisition: The acquisition module acquires signals in the region of interest, which is the region of interest related to motor imagery, and sends the acquired signals to the signal processing module; Signal conversion: The signal processing module converts the acquired signals into digital signals through an analog-to-digital converter; AI large model preprocessing: The signal processing module uses an AI large model to preprocess the converted digital signals and extract the high-order features and patterns of the signals; Feature extraction: The signal processing module performs feature extraction on the signals based on discrete wavelet transform; Pattern recognition: Perform pattern recognition on the signals according to the extracted features to obtain the motion instructions for the user to control the on-site sales robot; Wireless transmission: The signal processing module wirelessly transmits the motion instructions for the user to control the on-site sales robot to the single-chip microcomputer on the on-site sales robot through a wireless interface; Motion control: The single-chip microcomputer on the on-site sales robot sends control information to the drive circuit board according to the motion instructions, so that the single-chip microcomputer on the on-site sales robot controls the movement of the on-site sales robot.

[0006] Optionally, the method further includes: The signal processing module processes the digital signals; Each input sample of the digital signal x[n] sequentially enters a digital filter, and the output y[n] of the digital filter at any moment is a simple weighted sum of the current and past sample inputs:

[0007]

[0008] x[n-L] is the Lth past input sample, y[n-L] is the Lth past output sample, b L 、a L are the scale weights of each input sample and output sample respectively, and M and N are the numbers of input sample weights and output sample weights used by the filter; By adjusting the length of the filter and the weights assigned to consecutive samples, the signal frequencies in a specific frequency range will be amplified, attenuated, retained, or disappeared.

[0009] Optionally, the method further includes: Discrete wavelet transform discretizes c as the scale factor and d as the translation factor. The discrete wavelet transform of the signal f(t) is:

[0010]

[0011] ψ j,k (t) is the discretized wavelet function and is the kth low-frequency component of the jth layer;

[0012] Discrete wavelet transform is a uniformly discretized time series. The discrete wavelet transform DWTx(m,n) of the signal x(t) is defined as:

[0013]

[0014] where m is the scale factor, n is the dilation factor, and ψ m,n (t) is the wavelet function after discretization.

[0015] Optionally, the feature extraction of the signal by the signal processing module based on discrete wavelet transform includes: Wavelet transform can realize the frequency-domain conversion of the signal. Wavelet transform convolves a basic wavelet after scale transformation and time shift with the signal to be analyzed. The continuous wavelet transform CWTx(c, d) of the signal x(t) is defined as:

[0016]

[0017] where CWTx(c, d) is the time-frequency feature map, x(t) is the electrical signal, c is the scale factor, d is the translation factor, is the wavelet basis function under the translation factor d and the scale factor c, φ(t) is the basic wavelet (or mother wavelet), which can be a real signal or a complex signal; the non-zero real number c is the scale factor, which is the parameter for scaling the basic wavelet φ(t) on the time axis; the larger c is, the smaller the scale is, and the corresponding wavelet φ(t / c) is more stretched in the time domain and more compressed in the frequency domain. The time translation parameter d, and different d values represent that the wavelet is shifted to different positions along the time axis.

[0018] Optionally, the method further includes: processing the motor imagery electrical signal in the discrete domain with continuous wavelet transform and discretizing the transform, and the processing is as follows: cj = 2 j , d j,k = k * 2 j , where c is the scale factor, d is the translation factor, continuous wavelet transform maps the signal from one-dimensional space to two-dimensional space, discretizes the scale factor c and the time shift parameter d, and uses the wavelet transform values at some discrete points on the time-scale plane to characterize the signal, realizes the feature extraction of the frequency band of interest, and discretizes the scale according to the integer power of the constant 2, that is, takes c = 2^j, j ∈ Z.

[0019] Optionally, the method further includes:

[0020] When the user imagines the movement of the left hand, the energy in the right hemisphere decays, showing the related desynchronization ERD phenomenon, and the energy in the left hemisphere increases, showing the related synchronization ERS phenomenon. The quantization method of ERD / ERS is: E represents the energy value of the EEG signal in a certain frequency band after motor imagery; R represents the energy value of the EEG signal in a certain frequency band before motor imagery; when ERD / ERS is positive, the energy after motor imagery increases, and the ERS phenomenon appears; when ERD / ERS is negative, the energy after motor imagery decays, and the ERD phenomenon appears.

[0021] Optionally, the method further includes: by detecting the ERD / ERS value of the electrode, performing pattern recognition on the signal to obtain the motion instruction of the user for the on-site sales robot; when the detected ERD / ERS value of the left hemisphere electrode is positive, the energy of the left hemisphere increases and the energy of the right hemisphere decreases, it is concluded that the user performs left hand motor imagery, so as to recognize that the motion instruction of the user for the on-site sales robot is to move left; when the detected ERD / ERS value of the right hemisphere electrode is positive, the energy of the right hemisphere increases and the energy of the left hemisphere decreases, it is concluded that the user performs right hand motor imagery, so as to recognize that the motion instruction of the user for the on-site sales robot is to move right.

[0022] Optionally, the single-chip microcomputer on the on-site sales robot sends control information to the drive circuit board according to the motion instruction, so that the single-chip microcomputer on the on-site sales robot controls the motion of the on-site sales robot, including: when the single-chip microcomputer sends the control information to the drive circuit board with a pulse width modulation PWM signal frequency of 14 - 30Hz, the motors driving the left and right wheels of the on-site sales robot rotate in the same direction and at the same speed at the same time, and the on-site sales robot moves forward; when the single-chip microcomputer sends the control information to the drive circuit board with a pulse width modulation signal frequency of 30 - 40Hz, the left wheel motor of the on-site sales robot rotates forward and the right wheel motor rotates backward, and the two wheels form a differential speed, and the on-site sales robot moves right; when the single-chip microcomputer sends the control information to the drive circuit board with a pulse width modulation signal frequency of 8 - 13Hz, the left wheel motor of the on-site sales robot rotates backward and the right wheel motor rotates forward, and the two wheels also form a differential speed, and the on-site sales robot moves left; when the single-chip microcomputer sends the control information to the drive circuit board with a pulse width modulation signal frequency of 1 - 3Hz, the left and right wheel motors of the on-site sales robot stop rotating at the same time, and due to the effect of the speed reducer, the on-site sales robot stops moving quickly.

[0023] Optionally, after the single-chip microcomputer on the on-site sales robot sends control information to the drive circuit board according to the motion instruction, and the single-chip microcomputer on the on-site sales robot controls the movement of the on-site sales robot, the method further includes: the feedback module is used to test the physiological signal of the user. If a happy emotion is feedback, the movement mode of the on-site sales robot controlled by the single-chip microcomputer is correct; if a frustrated emotion is feedback, the movement mode of the on-site sales robot controlled by the single-chip microcomputer is incorrect; the feedback module sends an error report to the single-chip microcomputer, and the single-chip microcomputer sends an instruction for braking movement to the drive circuit board, so that the single-chip microcomputer controls the on-site sales robot to brake; when the movement mode of the on-site sales robot controlled by the single-chip microcomputer is correct, the user shows a happy emotion to feedback the recognition result; when the movement mode of the on-site sales robot controlled by the single-chip microcomputer is incorrect, the user shows a frustrated emotion to feedback the recognition result; by calculating the multi-dimensional information between EEG leads to obtain the asymmetry feature ASI, it can effectively determine whether an emotion occurs:

[0024]

[0025] Among them, refers to the information that the left brain signal X flows to the right brain signal Y. Similarly, represents the information flow from Y to X; S f is the total bidirectional information; S n refers to the information flow when no emotion occurs.

[0026] The second aspect of the present invention provides a path control method for an on-site sales robot, including the following steps:

[0027] S1: A global map is pre-stored inside the sales robot, the map is divided into multiple regions, and the central position points of each region are recorded as target points A, B, C, D, E, and F. The system receives the request information containing the node ID sent externally;

[0028] S2: Sense the current position through the sensor device installed inside the sales robot, mark the current position coordinates of the sales robot as the starting coordinates, parse the node ID in the request information, determine the marked coordinate position information corresponding to the node ID according to the preset coordinate system, and record the path from the starting coordinates to the target coordinates as the optimal path;

[0029] S3: The image acquisition device installed on the sales robot captures the images of the surrounding environment in real time, detects the position information of the obstacles from the images through the Yolo target detection algorithm, converts the obstacle information in the images into coordinates in the two-dimensional coordinate system, and marks the coordinate positions, lengths, and widths of the obstacles in the map. Obtain the obstacle avoidance parameters according to the obstacle coordinates, the distance from the initial coordinates, the minimum safety distance, and the length and width of the obstacles;

[0030] S4: Obtain the deceleration value of the sales robot based on the difference between the current speed and the target speed of the sales robot, add it to the steering-induced deceleration value of the sales robot obtained by multiplying the adjustment coefficient by the difference between the current speed and the linear speed to obtain the final deceleration value, and decelerate while avoiding obstacles;

[0031] S5: Compare the distance value between the current position of the sales robot and the target node with a preset distance value, and determine whether the target node has been reached according to the comparison result.

[0032] As a further solution of the present invention: The step S1 includes the following steps:

[0033] Create a global map through simultaneous localization and mapping technology, divide the map into multiple regions, mark the central positions inside each region as multiple target points, including target points A, B, C, D, E, and F, record them inside the sales robot, and the system receives request information containing node IDs sent externally; among them, the node IDs include reception desk point A, model display area point B, show flat entrance point C, negotiation area point D, signing area point E, and rest area point F.

[0034] As a further solution of the present invention: The step S2 includes the following steps:

[0035] Sense the current position through the sensor device installed inside the sales robot, mark the current position coordinates of the sales robot as the starting coordinates, the robot parses the received request information and extracts the node ID therein, constructs a preset two-dimensional plane coordinate system through the map structure, searches for the corresponding marked coordinate position information in the preset coordinate system according to the parsed node ID, makes the node ID correspond to the coordinate position one by one, and uses the path planning algorithm to obtain the optimal path from the starting coordinates to the target coordinates according to the starting point coordinates and the target point coordinates; among them, the node IDs are points A, B, C, D, E, F, and O, and the corresponding coordinates are target point A (X A , Y A ), B (X B , Y B ), C (X C , Y C ), D (X D , Y D ), E (X E , Y E ) and F (X F , Y F ) and starting point O (X O , Y O ).

[0036] As a further solution of the present invention: According to the starting point coordinates and the target point coordinates, use a path planning algorithm to calculate the optimal path from the starting point coordinates to the target point coordinates, including the following steps:

[0037] Search for the path from the current position to the target point through the path planning algorithm. During the search process, the path planning algorithm will evaluate according to the map data and the distance values between different nodes to find the optimal path. When an obstacle is detected during the detection process, update the map and re-plan the optimal path.

[0038] As a further solution of the present invention: The step S3 includes the following steps:

[0039] Use the image acquisition device installed on the sales robot to capture the images of the surrounding environment in real time, preprocess the captured images, extract the size and shape features of the obstacles from the preprocessed images, use deep learning algorithms to classify and identify the extracted features. After identifying the obstacles, detect the position information of the obstacles from the images through the Yolo target detection algorithm, use spatial geometric transformation to convert the obstacle information in the images into coordinates in the two-dimensional coordinate system, use the grid method to update the internal environment map of the robot, and mark the coordinate positions, lengths and widths of the obstacles in the map. Obtain the obstacle avoidance parameters through the obstacle coordinates, the distance between the initial coordinates, the minimum safety distance, and the length and width of the obstacles, adjust the speed of the robot, and assist the sales robot to perform positioning and avoid obstacles during movement;

[0040] Obtain the obstacle avoidance parameter calculation formula according to the obstacle coordinates, the distance between the initial coordinates, the minimum safety distance, the length and width of the obstacles:

[0041]

[0042] Among them, Distance2 is the distance value between the obstacle coordinates and the initial coordinates, Distance1 is the minimum safety distance value, (X O , Y O ) are the coordinates of the starting point O, (X Z , Y Z ) are the obstacle coordinates, denoted as point G; avoidance parameter is the calculated value of the obstacle avoidance parameter, L Z is the length value of the obstacle, W Z is the width value of the obstacle;

[0043] If the calculated value of the obstacle avoidance parameter is positive, the sales robot can pass safely without decelerating; if the calculated value of the obstacle avoidance parameter is negative, the sales robot needs to take deceleration and rotation angle measures to avoid obstacles.

[0044] As a further solution of the present invention: Step S4 includes the following steps:

[0045] Formula for calculating the safe braking distance at the current speed:

[0046]

[0047] Wherein, Distance3 is the calculated value of the safe braking distance, V is the current speed of the sales robot, a is the maximum deceleration of the sales robot, T is the reaction time of the robot, g is the acceleration due to gravity, and μ is the friction coefficient between the sales robot and the ground;

[0048] Formula for calculating the target speed value based on the maximum deceleration value and the minimum safe distance deceleration value of the sales robot:

[0049]

[0050] Wherein, V target is the target speed, a is the maximum deceleration of the sales robot, T is the reaction time of the robot, and Distance1 is the minimum safe distance value;

[0051] Formula for calculating the deceleration value of the sales robot based on the difference between the current speed and the target speed of the sales robot:

[0052] V s = V - V target

[0053] Wherein, V target is the target speed, V is the current speed of the sales robot, and V s is the deceleration value of the sales robot;

[0054] Multiply the adjustment coefficient by the difference between the current speed and the linear speed to obtain the steering deceleration value of the sales robot caused by steering. Formula for the steering deceleration value:

[0055]

[0056] Wherein, R is the radius of the wheel of the sales robot, V is the current speed of the sales robot, t is the change in time, α is the deceleration value adjustment coefficient, π is the pi, and V z is the calculated value of the steering deceleration value;

[0057] Formula for the final deceleration value:

[0058] V final = V s + V z

[0059] Wherein, V s is the deceleration value, and V zis the calculated value of the steering deceleration value, V final is the final deceleration value.

[0060] As a further solution of the present invention: the step S5 includes the following steps:

[0061] Compare the distance value between the current position of the sales robot and the target node with the preset distance value, and judge whether the target node has been reached according to the comparison result; if the distance value between the sales robot and the target node is greater than 0.3 meters, the sales robot has not reached the target node and needs to continue moving forward to the target point; if the distance value between the sales robot and the target node is less than or equal to 0.3 meters, the sales robot has reached the target node, and corresponding tasks are executed according to the information of different node IDs; after the sales robot completes the task, it starts to receive new task instructions, re-plan the path and execute the task.

[0062] The present invention uses a path planning algorithm to perform corresponding evaluations based on map data and distance values between different nodes to find the optimal path to the target point. When an obstacle is detected during the detection process, the map is updated and the optimal path to avoid the obstacle is re-planned, realizing the normal operation of the robot and avoiding potential risks; the present invention uses an image acquisition device to capture images of the surrounding environment in real time, extracts the size and shape features of the obstacle after preprocessing, and classifies and identifies the extracted features through a deep learning algorithm, realizing the accurate identification of the type of obstacle; the present invention uses the Yolo target detection algorithm to detect the position information of the obstacle from the image, converts the obstacle information in the image into coordinates in a two-dimensional coordinate system through spatial geometric transformation, and obtains obstacle avoidance parameters based on the distance between the obstacle coordinates and the initial coordinates, the minimum safety distance, and the length and width of the obstacle, and adjusts the speed of the robot, realizing the positioning and timely avoidance of obstacles by the sales robot during movement. Description of the Drawings

[0063] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings of the embodiments will be briefly introduced below. Obviously, the drawings described below only relate to some embodiments of the present invention and do not limit the present invention.

[0064] Figure 1 is a flowchart of a control method for a field sales robot provided by an embodiment of the present application;

[0065] Figure 2 is a flowchart of a control method for a field sales robot provided by another embodiment of the present application;

[0066] Figure 3 is a system framework diagram for controlling a field sales robot provided by an embodiment of the present application;

[0067] Figure 4 It is an interaction flowchart between system frameworks for controlling an on-site sales robot provided by an embodiment of the present application;

[0068] Figure 5 It is a physical interaction schematic diagram for controlling an on-site sales robot provided by an embodiment of the present application;

[0069] Figure 6 It is a flowchart of a path control method for an on-site sales robot provided by an embodiment of the present application. Detailed implementation manners

[0070] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the described embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0071] The present application proposes a control method and system for an on-site sales robot. The control system includes a signal processing device and a single-chip microcomputer installed on the on-site sales robot. This document aims at the information interaction between the signal processing device and the single-chip microcomputer on the on-site sales robot, and then the single-chip microcomputer controls the motion mode and motion trajectory of the on-site sales robot. The control system includes a collection module to collect signals; the signal processing module uses an AI large model to preprocess the converted digital signals and extract the high-order features and patterns of the signals; the signal processing module performs feature extraction on the signals based on discrete wavelet transform; performs pattern recognition on the signals according to the extracted features to obtain the motion instructions that the user wants to send to the on-site sales robot; wirelessly transmits the motion instructions that the user wants to send to the on-site sales robot to the single-chip microcomputer on the on-site sales robot through a wireless interface; the single-chip microcomputer on the on-site sales robot sends control information to the drive circuit board according to the motion instructions, thereby controlling the motion of the on-site sales robot.

[0072] Motor imagery refers to the process of only relying on the brain itself to imagine the movement of a certain part of the body without actually generating obvious movements. The basis for the generation of motor imagery electrophysiological signals is the event-related synchronization / desynchronization phenomenon. The process of motor imagery can activate certain regions of the brain, causing changes in the energy of some specific electrophysiological signals. For example, when people imagine the movement of their left arm without actually moving the left arm, the brain will generate an electrophysiological signal for executing the movement of the left arm.

[0073] The on-site sales robot in this application can be a smart robot, a mobile payment device, a smart navigation dog, or other movable terminal devices. Optionally, in the application scenario of this application, when the on-site sales robot is an on-site sales robot, when a marketer is conducting real estate sales, a salesperson can control the movement mode and movement direction of the on-site sales robot through mental control, and control the on-site sales robot to actively approach customers to explain real estate information to customers.

[0074] Embodiment 1

[0075] As Figure 1 Shown in the flowchart of the control method of an on-site sales robot provided by the embodiment of this application, the embodiment of this application proposes a method for controlling an on-site sales robot based on motor imagery, and the method includes the following steps:

[0076] Step 101, signal acquisition: The acquisition module acquires signals in the region of interest, which is a region of interest related to motor imagery, and sends the acquired signals to the signal processing module.

[0077] Acquire and detect signals under different motor imagery actions of the user, and divide the signals into several frequency bands including alpha wave, beta wave, gamma wave, and theta wave according to frequency.

[0078] Alpha wave: The frequency range is 8 - 13 Hz, and the amplitude is 20 - 100 μV. This frequency band is closely related to human motor imagery and actual movement. Beta wave: The frequency range is 14 - 30 Hz, and the amplitude is 5 - 20 μV. It often appears when humans are thinking or suddenly stimulated by the outside world. The appearance of beta wave is often accompanied by relatively strong emotional states such as fear, excitement, and excitement in humans. Theta wave: The frequency range is 4 - 7 Hz, and the amplitude is 20 - 150 μV. It often appears when adults feel slightly sleepy or extremely relaxed. Delta wave: The frequency range is 0.5 - 3 Hz, and the amplitude is 20 - 200 μV. It often appears when humans are in a deep sleep state or in a state of dizziness or coma.

[0079] The sampling frequency of the acquisition module can be 128 Hz. The acquisition module includes a signal generator, a photomultiplier tube, an amplifier, and an analog-to-digital converter. The signal generator includes detection electrodes and reference electrodes, and each detection electrode is connected to the input of a differential amplifier. The signal reflects the superposition of the postsynaptic potentials generated by the collected neurons. Functional magnetic resonance imaging obtains electrical signals through the change of blood flow during neuron activity. Similarly, functional infrared imaging measures the change of blood oxygen level. Magnetoencephalography measures the change of the magnetic field caused by electrical signals using instruments such as superconducting quantum interference devices.

[0080] Optionally, the acquired signal consists of N mixed signals, represented as a two-dimensional matrix x(t). The rows of the matrix represent the signals of each channel, and the columns represent the sampling points. x(t) = [x1(t), x2(t), x3(t),..., x N (t)] T , the unknown source signal a(t) = [a1(t), a2(t), a3(t),..., a N (t)] T . The mixing mechanism of the unknown signals is represented by matrix B. Then the acquired original signal is represented by the unknown signal source a(t), so x(t) = B·a(t). The unknown signals are mixed by matrix B to obtain the acquired signal x(t). Each column in matrix B represents the spatial distribution of a certain independent component on the electrodes.

[0081] A system for controlling an on-site sales robot proposed in this application further includes a signal recording module, which records the action potentials emitted by a single neuron collected by the electrodes in the acquisition module. The action potentials generated by neurons closer to the electrodes have a greater amplitude deviation in the recorded signals.

[0082] The signal processing module processes the digital signals; each input sample of the digital signal x[n] enters the filter in sequence. The output y[n] of the digital filter at any moment is a simple weighted sum of the current and past sample inputs:

[0083]

[0084] x[n-L] is the Lth past input sample, y[n-L] is the Lth past output sample, b L , a L are the scale weights of each input sample and output sample respectively. M and N are the numbers of input sample weights and output sample weights used by the filter respectively. By adjusting the length of the filter and the weights assigned to consecutive samples, the signal frequencies in a certain specific frequency range will be amplified, attenuated, retained, or disappeared.

[0085] After the amplified signal passes through an anti-aliasing filter, it is then digitized by an A / D (analog-to-digital) converter. By appropriately adjusting the length of the filter and the weights assigned to consecutive samples, the signal frequencies in a certain specific frequency range will be amplified, attenuated, retained, or disappeared.

[0086] Filters are divided into high-pass filters, low-pass filters, band-pass filters, and notch filters. A high-pass filter is used to remove low-frequency noise. Signals above a certain frequency can pass through, while signals below that frequency are attenuated and filtered out. Therefore, a high-pass filter retains high-frequency EEG activities, such as beta waves and gamma waves, and can improve the clarity of EEG signals. A low-pass filter is used to remove high-frequency noise. Signals below a certain frequency pass through, while those above the frequency do the opposite. A band-pass filter means that signals within a certain frequency range can pass through, while frequencies below or above this range are removed. Signals outside this frequency range are attenuated and filtered out, which is equivalent to performing high-pass filtering and low-pass filtering simultaneously. Notch filtering means that signals within a certain frequency range are attenuated and filtered out, while signals outside this frequency range are retained. For example, it is used to remove interference caused by the power supply frequency (such as 50 Hz or 60 Hz) and other frequency noises, which helps to reduce the impact of power interference on EEG signals.

[0087] Step 102, signal conversion: The signal processing module converts the collected signal into a digital signal through an analog-to-digital converter.

[0088] Discrete wavelets consist of two functions: a scaling function and a wavelet function. The unit step function is used as the scaling function, and the wavelet function is composed of the offset from the unit step function. Discrete wavelets are scaled only using specific time scales; the scaling is usually a power of 2. The Haar wavelet is a type of discrete wavelet, and the formula is as follows:

[0089]

[0090] As a type of digital filter, discrete wavelet transform can extract the characteristics of different frequency bands of electrical signals. Using wavelets, the original electrical signal is divided into five frequency bands, d1 - d5, and the corresponding frequency band ranges are: 64 - 128 Hz (d1), 32 - 64 Hz (d2), 16 - 32 Hz (d3), 8 - 16 Hz (d4), and 4 - 8 Hz (d5). The researchers calculated the signal energy and fluctuation index of the d3 - d5 frequency bands, and used a support vector machine to construct a discriminant model, and finally obtained a sensitivity of 94.46% and a specificity of 95.26%.

[0091] Optionally, feature extraction includes: reducing signal noise and enhancing relevant characteristics through signal conditioning, extracting features from the conditioned signal, and using feature conditioning to appropriately prepare the feature vector for the feature transformation stage.

[0092] Optionally, data augmentation: After step 102, the method for controlling the on-site sales robot based on motor imagery further includes: adding Gaussian noise to the digital signal in order to enhance the data, and the distribution function is:

[0093] The Gaussian distribution is also called the normal distribution N(μ, δ 2 ), where μ and δ 2 are the expected value and variance value of the Gaussian distribution, respectively.

[0094] Step 103, AI large model preprocessing: The signal processing module uses the AI large model to preprocess the converted digital signal and extract the high-order features and patterns of the signal.

[0095] The AI large model performs multi-level feature extraction and analysis on the signal through deep learning algorithms (such as convolutional neural networks or recurrent neural networks) to identify the feature signals related to motor imagery.

[0096] Specifically, the steps for extracting the high-order features of the signal are as follows. Let the probability density function of the random variable x(t) be f(x), and its first characteristic function be

[0097]

[0098] The first characteristic function Φ(ω) is the Fourier transform of the probability density function f(x). The second characteristic function Ψ(ω) is ψ(ω) = In[Φ(ω)];

[0099] The definition of the high-order cumulant is that the k-th cumulant of the random variable x is the k-th derivative of its second characteristic function at the origin, that is, C k = ψ (k) (ω) ω＝0 ;

[0100] The bispectrum is defined as the two-dimensional Fourier transform of the third-order cumulant,

[0101] is the frequency, is the time delay constant, When, the diagonal slice of the signal in the bispectrum is used to obtain the frequency of the signal after removing Gaussian white noise, that is, the frequency corresponding to the maximum value of the bispectrum.

[0102] Optionally, the formula for extracting the high-order features of the signal is as follows:

[0103] , where is the weight of the i-th neuron in layer L1 and the j-th neuron in layer L2, is the corresponding bias, k is the number of neurons, is the eigenvalue of layer L1 of the output layer for extracting the high-order features of the signal, is the one-dimensional feature.

[0104] Step 104, Feature Extraction: The signal processing module performs feature extraction on the signal based on discrete wavelet transform.

[0105] The discrete wavelet transform discretizes c as the scale factor and d as the translation factor. The discrete wavelet transform of the signal f(t) is:

[0106]

[0107] ψ j,k (t) is the discretized wavelet function and is the k-th low-frequency component of the j-th layer;

[0108] The discrete wavelet transform is a uniformly discretized time series. The discrete wavelet transform DWTx(m,n) of the signal x(t) is defined as:

[0109]

[0110] where m is the scale factor, n is the dilation factor, and ψ m,n (t) is the discretized wavelet function.

[0111] The feature extraction of the signal by the signal processing module based on discrete wavelet transform includes: Wavelet transform can achieve frequency-domain conversion of the signal. Wavelet transform convolves a basic wavelet as a scale transformation and time shift with the signal to be analyzed. The continuous wavelet transform CWTx(c,d) of the signal x(t) is defined as:

[0112]

[0113] where φ(t) is the basic wavelet (or mother wavelet), which can be a real signal or a complex signal; common mother wavelets include Haar wavelet, Morlet wavelet. The non-zero real number c is the scale factor, which is a parameter for scaling the basic wavelet φ(t) on the time axis; the larger c is, the smaller the scale is, and the corresponding wavelet φ(t / c) is more stretched in the time domain and more compressed in the frequency domain. The real number d is the time translation parameter, and different d values represent that the wavelet is shifted to different positions along the time axis.

[0114] Optionally, the corresponding common spatial pattern feature vector f(x) is obtained through covariance logarithm, and the formula is as follows:

[0115]

[0116] where, Z j is the j-th row signal of the signal Z after the spatial filter, m is the characteristic parameter of the common spatial pattern spatial filter, var(.) is the variance operator, and In(.) is the logarithm operator.

[0117] Optionally, the continuous wavelet transform extracts the input signal according to different frequency bands, so as to realize multi-scale decomposition to obtain the corresponding frequency band signals, and finally realize the time-frequency domain decomposition of the original signal. Compared with the Fourier transform, the wavelet transform can pay more attention to the local characteristics of the signal and can better analyze the signal in the time-frequency domain.

[0118] Optionally, the method further includes: processing the electroencephalogram signal of motor imagery in the discrete domain with the continuous wavelet transform, and discretizing the transform, and the processing is as follows: cj = 2 j ,d j,k = k * 2 j , where c is the scale factor and d is the translation factor. The continuous wavelet transform maps the signal from one-dimensional space to two-dimensional space, discretizes the scale factor c and the time shift parameter d, and uses the wavelet transform values at some discrete points on the time-scale plane to characterize the signal, so as to realize the feature extraction of the frequency band of interest. The scale is discretized according to the integer power of the constant 2, that is, c = 2^j, j ∈ Z.

[0119] The continuous wavelet transform maps the signal from one-dimensional space to two-dimensional space, so a lot of the information contained is redundant. This allows the scale factor c and the time shift parameter d to be discretized to realize the feature extraction of the frequency band of interest.

[0120] Optionally, this step further includes extracting time domain, frequency domain, time-frequency domain and spatial domain features of the signal.

[0121] The time domain function becomes a frequency domain function through the Fourier transform.

[0122]

[0123] Time domain feature extraction is a relatively primitive technology. It analyzes the signal according to time, enabling us to quantify how the signal changes over time. This is particularly important in EEG signals because they are usually recorded over a time range of several hours. Usually, windowing and segmentation of the signal are required for time domain feature extraction. Each window will extract a local feature, and researchers will be able to view how these features change over each window.

[0124] Frequency domain analysis techniques focus on extracting features from the sine wave signals that make up the data. Usually, it is first transformed from the time domain to the frequency domain and then further analyzed.

[0125] Optionally, optimize the features of the signal. Calculate the SU value SU i between each feature F ic in each frequency band and the category C, and sort them in descending order according to the value of SU ic . Then, according to each feature and the category label SU icSelect different thresholds according to λ% of the value change range, and observe the remaining feature dimensions and related recognition accuracies after feature screening. The threshold selection formula is as follows:

[0126]

[0127] where λ = 10 - 90 and δ is the threshold.

[0128] Step 105, pattern recognition: Perform pattern recognition on the signal based on the extracted features to obtain the motion instructions for the user to control the on-site sales robot.

[0129] The motion modes of the on-site sales robot include: acceleration, braking, left turn, and right turn. The motion instructions of the on-site sales robot include: acceleration motion, braking motion, left turn motion, and right turn motion.

[0130] Use the SVM support vector machine with a radial basis kernel function as the motion imagination EEG signal identification classifier. The classification decision function is as follows:

[0131]

[0132] where sgn(·) represents the sign function, a i * represents the optimal solution of a i b * represents the classification threshold, K(x, x i ) represents the kernel function, f(x) represents the classification function, x i represents the i-th sample, y i represents the result label of the i-th sample. The category to which the classification sample x belongs is judged according to the positive or negative of the classification function. Select the optimal channels for each sub-band according to the characteristics of each individual and the frequency band signals, effectively avoiding the interference of non-dominant channel noise. Through the feature characterization method, extract and optimize the features of the motion imagination EEG signal, and then realize the effective characterization of the motion imagination EEG signal. Finally, use the support vector machine SVM classifier to realize the pattern recognition of the motion imagination EEG signal.

[0133] Control the left and right turning and walking of the on-site sales robot by imagining the movement of the left and right arms, control the braking of the on-site sales robot by imagining the movement of the tongue, and control the forward and backward movement of the on-site sales robot by imagining the movement of the left and right feet. After determining that the on-site sales robot needs to turn right and walk, the user needs to imagine raising the right arm. After the recognition module in the on-site sales robot recognizes the user's electrical signal, it obtains the control instruction and sends the control instruction to the control module to control the steering of the universal wheel of the on-site sales robot.

[0134] When humans are stimulated by visual, auditory, or tactile stimuli, specific functional areas of the cerebral cortex are activated to analyze the stimuli and transmit relevant information, thereby accelerating the metabolic rate and blood flow velocity. These physiological activities consume a large amount of energy, resulting in a decrease in the amplitude of the corresponding electroencephalogram (EEG) signals, that is, a decrease in the energy of a certain rhythm. This phenomenon is called event-related desynchronization (ERD). Conversely, when the brain is in a relaxed and peaceful state, the amplitude of the EEG signals of the relevant rhythm will increase significantly, and the energy will also increase accordingly. This phenomenon is called event-related synchronization (ERS). The electrode channels C3 and C4 for collecting EEG signals are located in the parietal lobe area of the brain responsible for limb movement sensation. Therefore, when the subject imagines moving the left or right hand, event-related desynchronization / synchronization phenomena will occur in the C3 and C4 areas of the cerebral cortex. Also, since the left and right hemispheres of the brain control opposite body movements, taking the imagination of left hand movement as an example, the energy of the EEG rhythm related to the right hemisphere cortex of the brain decays, showing the ERD phenomenon, while the energy of the EEG rhythm related to the left hemisphere cortex increases, showing the ERS phenomenon. The corresponding relationship between the ERD / ERS phenomenon and the imagination of left and right hand movements on the C3 and C4 channels can be obtained as shown in the following table:

[0135] Imagery task Channel ERD / ERS phenomenon Left hand C3 ERS C4 ERD Right hand C3 ERS C4 ERD

[0136] When the user imagines moving the left hand, the energy of the right hemisphere decays, showing the related desynchronization ERD phenomenon; the energy of the left hemisphere increases, showing the related synchronization ERS phenomenon. The quantization method of ERD / ERS is as follows:

[0137] E represents the energy value of the EEG signal of a certain frequency band after motor imagery at a certain electrode; R represents the energy value of the EEG signal of a certain frequency band before motor imagery at a certain electrode. From the above formula, it can be concluded that when ERD / ERS is positive, it indicates an increase in energy after motor imagery, so the ERS phenomenon appears; when ERD / ERS is negative, it indicates a decrease in energy after motor imagery, so the ERD phenomenon appears.

[0138] Optionally, by detecting the ERD / ERS value of the electrode, pattern recognition is performed on the signal to obtain the motion instruction for the user to control the on-site sales robot.

[0139] For example, when the detected ERD / ERS value of the C3 electrode in the left hemisphere is positive, the energy of the left hemisphere increases and the energy of the right hemisphere decreases. The user imagines moving the left hand, and it is recognized that the motion instruction for the user to control the on-site sales robot is to move to the left; when the detected ERD / ERS value of the electrode in the right hemisphere is positive, the energy of the right hemisphere increases and the energy of the left hemisphere decreases. It is concluded that the user imagines moving the right hand, and thus it is recognized that the motion instruction for the user to control the on-site sales robot is to move to the right.

[0140] Step 106, Wireless Transmission: The signal processing module wirelessly transmits the motion instruction that the user wants to use to control the on-site sales robot to the single-chip microcomputer on the on-site sales robot through a wireless interface.

[0141] Wirelessly transmit the motion instructions of the user for the on-site sales robot: accelerating motion, braking motion, left-turn motion, or right-turn motion, to the single-chip microcomputer on the on-site sales robot.

[0142] The on-site sales robot can obtain information from the control system through wired or wireless connection methods. The above wireless connection methods include but are not limited to 3G / 4G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra-wideband) connection, and other currently known or future-developed wireless connection methods.

[0143] The single-chip microcomputer is an STM32F microcontroller or an 89C51 microcontroller. There is a high-gain inverting amplifier that constitutes an internal oscillator in the STM32F single-chip microcomputer. The pins XTALl and XTAL2 are the input and output terminals of the amplifier respectively, and are used to externally connect a crystal oscillator; the STM32F single-chip microcomputer enables reset with a high level. In the design of the minimum single-chip microcomputer system, it is necessary to ensure that the single-chip microcomputer can be reset when powered on. At the same time, when the system runs away during operation, the single-chip microcomputer can be reset through the corresponding button. Therefore, the single-chip microcomputer reset requires two reset methods: power-on reset and button reset.

[0144] Step 107, Motion Control: The single-chip microcomputer on the on-site sales robot sends control information to the drive circuit board according to the motion instruction, so that the single-chip microcomputer on the on-site sales robot controls the motion of the on-site sales robot.

[0145] Specifically, the single-chip microcomputer controls the steering and rotation speed of the motors of the on-site sales robot. When the frequency of the pulse-width modulation (PWM) signal sent by the single-chip microcomputer to the drive circuit board is 14 - 30 Hz, the motors driving the left and right wheels of the on-site sales robot rotate simultaneously in the same direction and at the same speed, and the on-site sales robot moves forward. After algorithm processing, the speed of the on-site sales robot is proportional to the strength of the signal. The stronger the signal intensity, the faster the speed of the on-site sales robot. To play a role in safety protection, a maximum value is set for the rotation speed control of the motor, and when the signal jumps at different frequency bands, the system immediately stops working and the on-site sales robot stops moving. When the single-chip microcomputer sends a pulse-width modulation (PWM) signal with a frequency of 40 Hz, the left-wheel motor of the on-site sales robot rotates forward, and the right-wheel motor rotates backward. The two wheels form a differential speed, and the on-site sales robot moves to the right. At this time, the signal strength is not proportional to the rotation speed of the motors in the on-site sales robot, that is, in this frequency band, the turning speed of the on-site sales robot is a uniform turn. When the single-chip microcomputer sends a pulse-width modulation (PWM) signal with a frequency of 8 - 13 Hz, the left-wheel motor of the on-site sales robot rotates backward, and the right-wheel motor rotates forward. The two wheels also form a differential speed, and the on-site sales robot moves to the left. At this time, the signal strength is not proportional to the rotation speed of the motors in the on-site sales robot, that is, in this frequency band, the turning speed of the on-site sales robot is also a uniform turn. When the single-chip microcomputer sends a pulse-width modulation (PWM) signal with a frequency of 1 - 3 Hz, the left and right wheel motors of the on-site sales robot stop rotating simultaneously. Due to the effect of the reduction gearbox, the on-site sales robot quickly stops moving.

[0146] In some embodiments, sensors are arranged around the on-site sales robot. The sensors detect the distance between the on-site sales robot and surrounding obstacles and send the distance information to the single-chip microcomputer, and the single-chip microcomputer controls the movement of the on-site sales robot according to the distance information.

[0147] After step 107, the single-chip microcomputer on the on-site sales robot sends control information to the drive circuit board according to the movement instruction. After the single-chip microcomputer on the on-site sales robot controls the movement of the on-site sales robot, the above method further includes:

[0148] The feedback module is used to test the physiological signals of the user. If it feedbacks a happy emotion, the movement mode of the on-site sales robot controlled by the single-chip microcomputer is correct; if it feedbacks a frustrated emotion, the movement mode of the on-site sales robot controlled by the single-chip microcomputer is incorrect; the feedback module sends an error report to the single-chip microcomputer, and the single-chip microcomputer sends an instruction for braking movement to the drive circuit board, so that the single-chip microcomputer controls the on-site sales robot to perform braking movement.

[0149] When the microcontroller controls the movement of the on-site sales robot in the correct mode, the user shows a happy emotion and gives feedback on the recognition result; when the microcontroller controls the movement of the on-site sales robot in the wrong mode, the user shows a frustrated emotion and gives feedback on the recognition result.

[0150] Optionally, by calculating the multi-dimensional information between EEG leads to obtain the asymmetry feature ASI, it can effectively determine whether an emotion occurs:

[0151]

[0152] Among them, refers to the information flowing from the left brain signal X to the right brain signal Y. Similarly, represents the information flow from Y to X; S f is the total amount of bidirectional information; S n refers to the information flow when no emotion occurs.

[0153] Emotion recognition is to analyze the characteristics of EEG signals shown by people in different emotional states and then perform emotion recognition. So the first step is to extract relevant characteristics from the EEG signal data collected in our laboratory, providing the original feature set for subsequent feature selection and classification recognition. Features are the basis and the key to classification recognition, and the quality of features directly affects the quality of classification recognition results. The experiment first collects EEG signals of people in different emotional states through a physiological signal recorder; we directly calculate these data, and the values calculated are called original features. By calculating various statistical values of EEG signals, these original statistical values are used as the original features. Also, according to the particularity and difference of EEG signals, the average energy of the β-wave rhythm of EEG signals in the time domain and frequency domain is extracted.

[0154] Feature extraction of electrical signals: the mean A of the β-wave of the electrical signal, the standard deviation B of the β-wave of the electrical signal, the minimum value Min of the β-wave of the electrical signal, the maximum value Max of the β-wave of the electrical signal, the ratio min Ratio of the number of minimum values of the β-wave of the electrical signal to the signal length, the ratio max Ratio of the number of maximum values of the β-wave of the electrical signal to the signal length, the average energy EnergyA of the β-wave of the electrical signal.

[0155]

[0156] Among them, in the formula, E represents the electrical signal data, and N represents the length of the electrical signal data.

[0157] The present application provides a control method for a field sales robot, which uses the electrical signals generated when the user imagines different movements as the control information source to intelligently control the movement mode of the field sales robot, thereby realizing the user's free control of the field sales robot with his mind. Through the above method, the deep learning ability of the AI large model can be used to more accurately identify and process signals related to movement imagination, thereby realizing precise control of the field sales robot.

[0158] Embodiment 2

[0159] like Figure 2 The flowchart of a control method of a field sales robot provided by an embodiment of the present application is shown. Embodiment 2 of the present application proposes a method for controlling a field sales robot based on motor imagery, and the method comprises the following steps:

[0160] Step 201: The acquisition module acquires signals in an area of interest, which is an area of interest related to motor imagery, and sends the acquired signals to a signal processing module.

[0161] Step 202: The signal processing module converts the collected signal into a digital signal through an analog-to-digital converter.

[0162] Step 203: The signal processing module uses the AI big model to pre-process the converted digital signal to extract the high-order features and patterns of the signal.

[0163] Step 204: The signal processing module performs feature extraction on the signal based on discrete wavelet transform.

[0164] Step 205: Perform pattern recognition on the signal based on the extracted features to obtain the motion instructions that the user wants to use to control the on-site sales robot.

[0165] Step 206: The signal processing module wirelessly transmits the motion instructions that the user wants to give to the on-site sales robot to the single-chip microcomputer on the on-site sales robot through the wireless interface.

[0166] Step 207: The single-chip microcomputer on the on-site sales robot sends control information to the driving circuit board according to the motion instruction, so that the single-chip microcomputer on the on-site sales robot controls the movement of the on-site sales robot.

[0167] Step 208, the feedback module is used to test the user's physiological signals. If the feedback is joyful, the single-chip microcomputer controls the movement mode of the on-site sales robot correctly.

[0168] Step 209 , if the feedback is of frustrated emotion, the single chip microcomputer controls the mode of the on-site sales robot's movement in error, the feedback module sends an error report to the single chip microcomputer, and jumps to step 207 .

[0169] Step 210, the single-chip microcomputer sends an instruction for braking movement to the drive circuit board, so that the single-chip microcomputer controls the braking movement of the on-site sales robot.

[0170] Embodiment III

[0171] As Figure 3 shown, a system framework diagram for controlling an on-site sales robot provided by an embodiment of the present application, as Figure 4 shown, an interaction flowchart between system frameworks for controlling an on-site sales robot provided by an embodiment of the present application, as Figure 5 shown, a physical interaction schematic diagram for controlling an on-site sales robot provided by an embodiment of the present application. Embodiment III of the present application proposes a system for controlling an on-site sales robot. The system 300 includes: a collection module 301, configured to collect signals in an interest area, where the interest area is an interest area related to motor imagery, and send the collected signals to a signal processing module; the signal processing module 302, configured to convert the collected signals into digital signals through an analog-to-digital converter, preprocess the converted digital signals using an AI large model, extract high-order features and patterns of the signals, and the signal processing module performs feature extraction on the signals based on discrete wavelet transform; an identification module 303, configured to perform pattern recognition on the signals according to the extracted features to obtain a motion instruction for the user to control the on-site sales robot 307; a transmission module 304, configured to wirelessly transmit the motion instruction for the user to control the on-site sales robot 307 by the signal processing module through a wireless interface to a single-chip microcomputer on the on-site sales robot; a control module 305, configured to send control information to a drive circuit board by the single-chip microcomputer on the on-site sales robot according to the motion instruction, so that the single-chip microcomputer on the on-site sales robot controls the motion of the on-site sales robot; a feedback module 306, configured to test the physiological signals of the user. If a happy emotion is feedback, the mode of the single-chip microcomputer controlling the motion of the on-site sales robot is correct; if a frustrated emotion is feedback, the mode of the single-chip microcomputer controlling the motion of the on-site sales robot is incorrect; the feedback module sends an error report to the single-chip microcomputer, and the single-chip microcomputer sends an instruction for braking movement to the drive circuit board, so that the single-chip microcomputer controls the braking movement of the on-site sales robot.

[0172] The collection module 301 can be a module independent of the control system 300 and send the collected signals to the control system 300, or can be a module embedded in the control system. The collection module 301 and the control system 300 are an integral whole.

[0173] Information transmission between the on-site sales robot and the control system can be carried out through a wired connection or a wireless connection method.

[0174] A system for controlling a on-site sales robot provided in the above embodiment corresponds one-to-one with a method for controlling a on-site sales robot provided in the above embodiment. The system for controlling a on-site sales robot executes all steps of the method for controlling a on-site sales robot, which will not be described in detail herein.

[0175] Embodiment 4

[0176] A method for controlling a on-site sales robot provided in an embodiment of the present application includes the following steps: establishing a coordinate system, determining the coordinates of multiple nodes in the coordinate system, obtaining the current position coordinates of the sales robot in real time, planning a path trajectory of the sales robot from the current position to the multiple nodes, and the sales robot moves according to the planned path trajectory.

[0177] Establish a coordinate system and determine the coordinates of multiple nodes in the coordinate system. These nodes may include important positions in the sales area, such as product display areas, cash registers, etc. Obtain the current position coordinates of the sales robot in real time, and the position of the sales robot in the coordinate system can be accurately determined through sensor technologies such as GPS and lidar. According to the coordinates of the multiple nodes and the current position coordinates of the sales robot, plan a path trajectory of the sales robot from the current position to the multiple nodes. The sales robot moves according to the planned path trajectory and updates its own position coordinates in real time so as to be able to adjust the path in time when encountering obstacles or other abnormal situations.

[0178] Optionally, the sales robot receives a request message sent by a node, and the request message includes a node ID; determines the coordinates of the node in the coordinate system according to the node ID in the request message, plans a path trajectory of the sales robot from the current position to the node, and the sales robot moves to the position of the node according to the planned path trajectory.

[0179] Embodiment 5

[0180] Please refer to Figure 6 The flowchart of a path control method for a on-site sales robot shown. The present invention is a path control method for a on-site sales robot, including the following steps:

[0181] S601: Pre-store a global map inside the sales robot in advance, divide the map into multiple areas, record the central position points of each area as target points A, B, C, D, E, and F, and the system receives a request message containing a node ID sent externally.

[0182] S602: Sense the current position through the sensor device installed inside the sales robot, mark the current position coordinates of the sales robot as the starting coordinates, parse the node ID in the request information, determine the marked coordinate position information corresponding to the node ID according to the preset coordinate system, and record the path from the starting coordinates to the target coordinates as the optimal path.

[0183] S603: Use the image acquisition device installed on the sales robot to capture the images of the surrounding environment in real time, detect the position information of the obstacles from the images through the Yolo target detection algorithm, convert the obstacle information in the images into coordinates in the two-dimensional coordinate system, mark the coordinate positions, lengths, and widths of the obstacles in the map, and obtain the obstacle avoidance parameters based on the distance between the obstacle coordinates and the initial coordinates, the minimum safety distance, and the lengths and widths of the obstacles.

[0184] S604: Obtain the final deceleration value by adding the deceleration value of the sales robot obtained from the difference between the current speed and the target speed of the sales robot and the steering-induced deceleration value of the sales robot obtained by multiplying the difference between the current speed and the linear speed by the adjustment coefficient, and then decelerate and avoid obstacles.

[0185] S605: Compare the distance value between the current position of the sales robot and the target node with the preset distance value, and determine whether the target node has been reached according to the comparison result.

[0186] Specifically, a global map is pre-stored inside the sales robot. The global map is divided into multiple areas, and each area represents a specific task execution area. The central position points of each area are marked and recorded as target points A, B, C, D, E, and F. These target points serve as reference points for the robot's navigation and task execution. The sensor device is activated to continuously monitor and record the current position information of the sales robot. When a request message containing a node ID is received from the outside, the current position coordinates of the sales robot obtained by the sensor at this time are immediately marked as the starting coordinates. The node ID in the request message is parsed, converted into a target point identifier recognizable by the system, and the coordinate position of the parsed node ID is searched. The starting coordinates, target coordinates, and map data are input into the selected path planning algorithm to calculate the optimal path from the starting coordinates to the target coordinates. The image acquisition device is activated to capture and record the images of the robot's surrounding environment. The coordinate information of the obstacles is extracted from the output results of the pre-trained Yolo object detection algorithm, and two-dimensional coordinate conversion is performed to mark the coordinate information of the obstacles in the map data, and the obstacle avoidance parameters are calculated. Deceleration is performed by adding the deceleration value of the sales robot obtained from the difference between the current speed and the target speed of the sales robot and the deceleration value caused by turning of the sales robot obtained by multiplying the adjustment coefficient by the difference between the current speed and the linear speed to obtain the final deceleration value. Whether the target node has been reached is judged according to the comparison result of the distance value between the current position of the sales robot and the target node and the preset distance value.

[0187] In one embodiment of the present invention, step S601 includes the following steps: creating a global map through simultaneous localization and mapping technology, dividing the map into multiple areas, marking the central positions inside each area as multiple target points, including target points A, B, C, D, E, and F, and storing and recording them inside the sales robot. The system receives a request message containing a node ID sent from the outside; wherein, the node ID includes point A of the reception desk, point B of the model display area, point C of the sample room entrance, point D of the negotiation area, point E of the signing area, and point F of the rest area.

[0188] Specifically, a global map is pre-stored inside the sales robot. The global map is divided into multiple areas, and each area represents a specific task execution area. The central position points of each area are marked and recorded as target points A, B, C, D, E, and F. These target points serve as reference points for the robot's navigation and task execution. The node ID includes point A of the reception desk, point B of the model display area, point C of the sample room entrance, point D of the negotiation area, point E of the signing area, and point F of the rest area.

[0189] In one embodiment of the present invention, the step S602 includes the following steps: sensing the current position through a sensor device installed inside the sales robot, marking the current position coordinates of the sales robot as the starting coordinates, parsing the received request information by the robot and extracting the node ID therein, constructing a preset two-dimensional plane coordinate system through the map structure, searching for the corresponding marked coordinate position information in the preset coordinate system according to the parsed node ID, making the node ID correspond to the coordinate position one by one, and obtaining the optimal path from the starting coordinate to the target coordinate using a path planning algorithm; where the node IDs are points A, B, C, D, E, F, and O, and the corresponding coordinates are the target point A (X A , Y A ), point B (X B , Y B ), point C (X C , Y C ), point D (X D , Y D ), point E (X E , Y E ), and point F (X F , Y F ) and the starting point O (X O , Y O ).

[0190] Specifically, start the sensor device installed inside the sales robot to continuously monitor and record the current position information of the sales robot. When receiving a request information containing a node ID sent from the outside, immediately mark the current position coordinates of the sales robot obtained by the sensor at this time as the starting coordinates, parse the node ID in the request information, convert it into a target point identifier recognizable by the system, and search for the coordinate position of the parsed node ID. Input the starting coordinates, target coordinates, and map data into the selected path planning algorithm to calculate the optimal path from the starting coordinates to the target coordinates; where the node IDs are points A, B, C, D, E, F, and O, and the corresponding coordinates are the target point A (X A , Y A ), point B (X B , Y B ), point C (X C , Y C ), point D (X D , Y D ), point E (X E , Y E ), and point F (X F , Y F ) and the starting point O (X O , Y O ).

[0191] In one embodiment of the present invention, according to the starting point coordinates and the target point coordinates, an optimal path from the starting point coordinates to the target point coordinates is calculated using a path planning algorithm, including the following steps: Search for a path from the current position to the target point through the path planning algorithm. During the search process, the path planning algorithm will evaluate based on the map data and the distance values between different nodes to find the optimal path. When an obstacle is detected during the detection process, update the map and re-plan the optimal path.

[0192] Specifically, search for a path from the current position to the target point through the path planning algorithm. In the pre-stored global map data structure, mark the coordinate positions, lengths, and widths of the obstacles. When a new obstacle is detected or an existing obstacle moves, update the map data in a timely manner, and re-plan the optimal path for the sales robot from the current position to the target position while avoiding the obstacles.

[0193] In one embodiment of the present invention, step S603 includes the following steps: Use the image acquisition device installed on the sales robot to capture images of the surrounding environment in real time, preprocess the captured images, extract the size and shape features of the obstacles from the preprocessed images, use deep learning algorithms to classify and identify the extracted features. After identifying the obstacles, detect the position information of the obstacles from the images through the Yolo target detection algorithm, convert the obstacle information in the image into coordinates in a two-dimensional coordinate system using spatial geometric transformation, update the internal environment map of the robot using the grid method, and mark the coordinate positions, lengths, and widths of the obstacles in the map. Obtain obstacle avoidance parameters based on the obstacle coordinates and the distances from the initial coordinates, the minimum safety distance, and the lengths and widths of the obstacles, and adjust the speed of the robot to assist the sales robot in positioning and avoiding obstacles during movement; Obtain the calculation formula for the obstacle avoidance parameters based on the obstacle coordinates and the distances from the initial coordinates, the minimum safety distance, the lengths and widths of the obstacles:

[0194]

[0195] Among them, Distance2 is the distance value between the obstacle coordinates and the initial coordinates, Distance1 is the minimum safety distance value, (X O , Y O ) are the coordinates of the starting point O, (X Z , Y Z ) are the obstacle coordinates, denoted as point G; avoidance parameter is the calculated value of the obstacle avoidance parameter, L Z is the length value of the obstacle, W Z is the width value of the obstacle;

[0196] If the calculated value of the obstacle avoidance parameter is positive, the sales robot can pass safely without decelerating; if the calculated value of the obstacle avoidance parameter is negative, the sales robot needs to take measures of decelerating and rotating the angle to avoid the obstacle.

[0197] Specifically, the image acquisition device installed on the sales robot captures the images of the surrounding environment in real time, preprocesses the captured images, extracts the size and shape features of the obstacles from the preprocessed images, classifies and identifies the extracted features using deep learning algorithms. After identifying the obstacles, the position information of the obstacles is detected from the images through the Yolo object detection algorithm, and the center point coordinates in the images are obtained. The upper left and upper right coordinates of the bounding box are used to convert the obstacle information in the images into coordinates in the two-dimensional coordinate system using spatial geometric transformation. The grid method is used to update the internal environment map of the robot, and the coordinate positions, lengths, and widths of the obstacles are marked in the map. The map data structure can be queried to find the map area closest to the current robot position. The obstacle avoidance parameter is obtained through the distance between the obstacle coordinates and the initial coordinates, the minimum safety distance, and the length and width of the obstacle. If the calculated value of the obstacle avoidance parameter is positive, it means that the sales robot can pass safely without decelerating; if the calculated value of the obstacle avoidance parameter is negative, the sales robot needs to take measures of decelerating and rotating the angle to avoid the obstacle; among them, the minimum safety distance Distance1 is 12 meters.

[0198] In one embodiment of the present invention, the step S604 includes the following steps:

[0199] The calculation formula for the safe braking distance at the current speed:

[0200]

[0201] Among them, Distance3 is the calculated value of the safe braking distance, V is the current speed of the sales robot, a is the maximum deceleration of the sales robot, T is the reaction time of the robot, g is the acceleration due to gravity, and μ is the friction coefficient between the sales robot and the ground;

[0202] The calculation formula for the target speed value obtained from the maximum deceleration value and the minimum safety distance deceleration value of the sales robot:

[0203]

[0204] Among them, V target is the target speed, a is the maximum deceleration of the sales robot, T is the reaction time of the robot, and Distance1 is the minimum safety distance value;

[0205] The calculation formula for the deceleration value of the sales robot obtained from the difference between the current speed of the sales robot and the target speed of the sales robot:

[0206] V s = V - V target

[0207] where V target is the target speed, V is the current speed of the sales robot, and V s is the deceleration value of the sales robot;

[0208] Multiply the adjustment coefficient by the difference between the current speed and the linear speed to obtain the steering deceleration value of the sales robot caused by steering. The formula for calculating the steering deceleration value is:

[0209]

[0210] where R is the radius of the wheels of the sales robot, V is the current speed of the sales robot, t is the change in time, α is the deceleration value adjustment coefficient, π is the pi, and V z is the calculated value of the steering deceleration value;

[0211] The formula for calculating the final deceleration value is:

[0212] V final = V s + V z

[0213] where V s is the deceleration value, V z is the calculated value of the steering deceleration value, and V final is the final deceleration value.

[0214] Specifically, first calculate the distance between the current position of the robot and the nearest obstacle; determine whether obstacle avoidance measures need to be taken according to the preset minimum safety distance, and obtain the safe braking distance at the current speed through the current robot speed, maximum deceleration value, reaction time, gravitational acceleration, and friction coefficient between the sales robot and the ground. Among them, the gravitational acceleration g is 9.8, the adjustment coefficient α is 0.8, the pi π is 3.14, and the friction coefficient μ is 0.5. Obtain the target speed value according to the maximum deceleration value of the sales robot and the minimum safety distance deceleration value. The target speed value is the speed to ensure that the robot can stop within the safe distance. Obtain the deceleration value of the sales robot through the difference between the current speed of the sales robot and the target speed of the sales robot. Multiply the adjustment coefficient by the difference between the current speed and the linear speed to obtain the steering deceleration value caused by steering. The final deceleration value is obtained by adding the deceleration value and the steering deceleration value.

[0215] In one embodiment of the present invention, step S605 includes the following steps: comparing the distance value between the current position of the sales robot and the target node with a preset distance value, and judging whether the target node has been reached according to the comparison result; if the distance value between the sales robot and the target node is greater than 0.3 meters, the sales robot has not reached the target node and needs to continue moving forward to the target point; if the distance value between the sales robot and the target node is less than or equal to 0.3 meters, the sales robot has reached the target node, and corresponding tasks are executed according to the information of different node IDs; after the sales robot completes the task, it starts to receive new task instructions, re-plan the path and execute the task.

[0216] Specifically, the distance value between two points is obtained through the current position coordinate point of the sales robot and the coordinate point of the target node. A preset distance value is set to 0.3 meters, and the distance value between the two points is compared with the preset distance value. If the distance value between the current position of the sales robot and the target node is greater than 0.3 meters, it continues to move towards the target node; if the distance value between the current position of the sales robot and the target node is less than or equal to 0.3 meters, it starts to execute subsequent tasks; after the task ends, it receives new task instructions, re-plans the path and executes the task.

[0217] A control method and a path control method for a field sales robot provided by an embodiment of the present application solve the problem that the accessories for remotely controlling the field sales robot are too many and inconvenient to carry, and realize the free control of the field sales robot by the user's mind. In the present application, when a marketer conducts real estate sales, a salesperson can control the movement mode and movement direction of the field sales robot through the mind, control the field sales robot to actively walk towards the customer and explain the real estate information to the customer, improve the efficiency of the marketer's real estate sales, and reduce a large amount of labor costs.

[0218] The above is only an exemplary embodiment of the present invention, and is not used to limit the protection scope of the present invention. The protection scope of the present invention is determined by the appended claims.

Claims

1. A control method for an on-site sales robot, characterized in that The method includes: Signal acquisition: The acquisition module acquires signals in the region of interest, which is the region of interest related to motor imagery, and sends the acquired signals to the signal processing module; Signal conversion: The signal processing module converts the acquired signals into digital signals through an analog-to-digital converter; AI large model preprocessing: The signal processing module uses the AI large model to preprocess the converted digital signals and extract the high-order features and patterns of the signals; Feature extraction: The signal processing module performs feature extraction on the signals based on discrete wavelet transform; Pattern recognition: Pattern recognition is performed on the signals according to the extracted features to obtain the motion instructions for the user to control the on-site sales robot; Wireless transmission: The signal processing module wirelessly transmits the motion instructions for the user to control the on-site sales robot to the single-chip microcomputer on the on-site sales robot through a wireless interface; Motion control: The single-chip microcomputer on the on-site sales robot sends control information to the drive circuit board according to the motion instructions, so that the single-chip microcomputer on the on-site sales robot controls the motion of the on-site sales robot; The feature extraction of the signals by the signal processing module based on discrete wavelet transform includes: Wavelet transform can achieve frequency-domain conversion of signals. Wavelet transform convolves a basic wavelet after scale transformation and time shift with the signal to be analyzed. The continuous wavelet transform CWTx(c,d) of the signal x(t) is defined as: Among them, CWTx(c, d) is the time-frequency feature map, x(t) is the electrical signal, c is the scale factor, and d is the translation factor. is the wavelet basis function under the translation factor d and the scale factor c. φ(t) is the basic wavelet, which can be a real signal or a complex signal; the non-zero real number c is the scale factor, which is a parameter for scaling the basic wavelet φ(t) on the time axis; the larger c is, the smaller the scale is, and the corresponding wavelet φ(t / c) is more stretched in the time domain and more compressed in the frequency domain. The time translation parameter d, and different d values indicate that the wavelet is shifted to different positions along the time axis.

2. The method according to claim 1, wherein The method further includes: The signal processing module processes the digital signals; Each input sample of the digital signal x[n] enters the digital filter in sequence. The output y[n] of the digital filter at any moment is a simple weighted sum of the current and past sample inputs: x[n-L] is the L-th past input sample, y[n-L] is the L-th past output sample, and b L , a L are the scaling weights for each input sample and output sample respectively, and M and N are the numbers of input sample weights and output sample weights used by the filter; By adjusting the length of the filter and the weights assigned to consecutive samples, the signal frequencies in a specific frequency range will be amplified, attenuated, retained, or disappeared; The method further includes: Discrete wavelet transform discretizes c as the scale factor and d as the translation factor. The discrete wavelet transform of the signal f(t) is: ψ j,k (t) is the wavelet function after discretization and is the k-th low-frequency component of the j-th layer; Discrete wavelet transform is a uniformly discretized time series. The discrete wavelet transform DWTx(m,n) of the signal x(t) is defined as: where m is the scale factor, n is the dilation factor, and ψ m,n (t) is the wavelet function after discretization.

3. The method according to claim 1, characterized in that, The method further includes: processing the motor imagery electroencephalogram signals in the discrete domain with continuous wavelet transform, discretizing the transform, and the processing is as follows: cj = 2 j , d j,k = k * 2 j , where c is the scale factor and d is the translation factor. The continuous wavelet transform maps the signal from one-dimensional space to two-dimensional space, discretizes the scale factor c and the time shift parameter d, and uses the wavelet transform values at some discrete points on the time-scale plane to characterize the signal, so as to realize the feature extraction of the frequency band of interest. The scale is discretized according to the integer power of the constant 2, that is, c = 2^j, j ∈ Z; When the user imagines moving the left hand, the energy in the right hemisphere decays, manifested as the related desynchronization ERD phenomenon, and the energy in the left hemisphere increases, manifested as the related synchronization ERS phenomenon. The quantification method of ERD / ERS is as follows: E represents the energy value of the EEG signal in a certain frequency band after motor imagery for a certain electrode; R represents the energy value of the EEG signal in a certain frequency band before motor imagery for a certain electrode; when ERD / ERS is positive, the energy after motor imagery increases, and the ERS phenomenon appears; when ERD / ERS is negative, the energy after motor imagery decays, and the ERD phenomenon appears.

4. The method according to claim 3, characterized in that, The method further includes: By detecting the ERD / ERS values of the electrodes, pattern recognition is performed on the signals to obtain the motion instructions for the user to control the on-site sales robot; When the detected ERD / ERS value of the left hemisphere electrode is positive, the energy of the left hemisphere increases and the energy of the right hemisphere decreases, indicating that the user is performing left hand motor imagery. Thus, it is recognized that the motion instruction for the user to control the on-site sales robot is to move left; When the detected ERD / ERS value of the right hemisphere electrode is positive, the energy of the right hemisphere increases and the energy of the left hemisphere decreases, indicating that the user is performing right hand motor imagery. Thus, it is recognized that the motion instruction for the user to control the on-site sales robot is to move right; The single-chip microcomputer on the on-site sales robot sends control information to the drive circuit board according to the motion instructions, so that the single-chip microcomputer on the on-site sales robot controls the motion of the on-site sales robot includes: When the frequency of the pulse width modulation (PWM) signal sent by the single-chip microcomputer to the drive circuit board for the control information is 14 - 30 Hz, the motors driving the left and right wheels of the on-site sales robot rotate in the same direction and at the same speed simultaneously, and the on-site sales robot moves forward; when the frequency of the pulse width modulation signal sent by the single-chip microcomputer to the drive circuit board for the control information is 30 - 40 Hz, the left-wheel motor of the on-site sales robot rotates forward, and the right-wheel motor rotates backward, and the two wheels form a differential speed, and the on-site sales robot moves to the right; when the frequency of the pulse width modulation signal sent by the single-chip microcomputer to the drive circuit board for the control information is 8 - 13 Hz, the left-wheel motor of the on-site sales robot rotates backward, and the right-wheel motor rotates forward, and the two wheels also form a differential speed, and the on-site sales robot moves to the left; when the frequency of the pulse width modulation signal sent by the single-chip microcomputer to the drive circuit board for the control information is 1 - 3 Hz, the left and right wheel motors of the on-site sales robot stop rotating simultaneously, and due to the effect of the speed reducer, the on-site sales robot stops moving quickly.

5. The method according to claim 1, wherein After the single-chip microcomputer on the on-site sales robot sends control information to the drive circuit board according to the motion instruction, and thus the single-chip microcomputer on the on-site sales robot controls the movement of the on-site sales robot, the method further includes: The feedback module is used to test the physiological signals of the user. If it feedbacks a happy emotion, the single-chip microcomputer controls the movement mode of the on-site sales robot correctly; if it feedbacks a frustrated emotion, the single-chip microcomputer controls the movement mode of the on-site sales robot incorrectly; the feedback module sends an error report to the single-chip microcomputer, and the single-chip microcomputer sends an instruction for braking movement to the drive circuit board, so that the single-chip microcomputer controls the on-site sales robot to perform a braking movement; When the single-chip microcomputer controls the movement mode of the on-site sales robot correctly, the user shows a happy emotion to feedback the recognition result; when the single-chip microcomputer controls the movement mode of the on-site sales robot incorrectly, the user shows a frustrated emotion to feedback the recognition result; By calculating the multi-dimensional information between electroencephalogram leads to obtain the asymmetry feature ASI, it can effectively determine whether an emotion occurs: Among them, refers to the information flow from the left brain signal X to the right brain signal Y. Similarly, represents the information flow from Y to X; S f is the total amount of two-way information; S n refers to the information flow when no emotion occurs.

6. A path control method for an on-site sales robot, characterized in that, The method includes the following steps: S1: Pre-store a global map inside the sales robot in advance, divide the map into multiple regions, record the central position points of each region as target points A, B, C, D, E, and F, and the system receives the request information containing the node ID sent externally; S2: Sense the current position through the sensor device installed inside the sales robot, mark the current position coordinates of the sales robot as the starting coordinates, analyze the node ID in the request information, determine the marked coordinate position information corresponding to the node ID according to the preset coordinate system, and record the path from the starting coordinates to the target coordinates as the optimal path; S3: The image acquisition device installed on the sales robot captures the images of the surrounding environment in real time. The position information of the obstacles is detected from the images through the Yolo target detection algorithm. The obstacle information in the images is converted into coordinates in a two-dimensional coordinate system, and the coordinate positions, lengths, and widths of the obstacles are marked on the map. The obstacle avoidance parameters are obtained based on the obstacle coordinates, the distance between the initial coordinates, the minimum safety distance, and the lengths and widths of the obstacles. S4: The final deceleration value is obtained by adding the deceleration value of the sales robot obtained from the difference between the current speed and the target speed of the sales robot and the steering-induced deceleration value of the sales robot multiplied by the difference between the current speed and the linear speed and the adjustment coefficient, and then decelerating to avoid the obstacles. S5: The distance value between the current position of the sales robot and the target node is compared with the preset distance value, and it is judged whether the target node has been reached according to the comparison result. The step S2 includes the following steps: The current position is sensed through the sensor device installed inside the sales robot, and the current position coordinates of the sales robot are marked as the starting coordinates. The robot parses the received request information and extracts the node ID therein. A preset two-dimensional plane coordinate system is constructed through the map structure. According to the parsed node ID, the corresponding marked coordinate position information is searched in the preset coordinate system, and the node ID is put into one-to-one correspondence with the coordinate position. According to the starting point coordinates and the target point coordinates, the optimal path from the starting coordinates to the target coordinates is obtained using the path planning algorithm. Among them, the node IDs are points A, B, C, D, E, F, and O, and the corresponding coordinates are the target point A (X A , Y A ), point B (X B , Y B ), point C (X C , Y C ), point D (X D , Y D ), point E (X E , Y E ), and point F (X F , Y F ), and the starting point O (X O , Y O ); Calculating the optimal path from the starting point coordinates to the target point coordinates using the path planning algorithm includes the following steps: The path from the current position to the target point is searched through the path planning algorithm. During the search process, the path planning algorithm will evaluate according to the map data and the distance values between different nodes to find the optimal path. When an obstacle is detected during the detection process, the map is updated and the optimal path is re-planned.

7. The path control method of an on-site sales robot according to claim 6, characterized in that, The step S1 includes the following steps: A global map is created through the simultaneous localization and mapping technology. The map is divided into multiple regions, and the central positions inside each region are marked as multiple target points, including target points A, B, C, D, E, and F, which are stored and recorded inside the sales robot. The system receives the request information containing the node ID sent externally. Among them, the node ID includes point A of the reception desk, point B of the model display area, point C of the sample room entrance, point D of the negotiation area, point E of the signing area, and point F of the rest area.

8. The path control method of an on-site sales robot according to claim 6, characterized in that, The step S3 includes the following steps: The images of the surrounding environment are captured in real time by the image acquisition device installed on the sales robot, and the captured images are preprocessed. The size and shape features of the obstacles are extracted from the preprocessed images. The extracted features are classified and recognized using deep learning algorithms. After the obstacles are recognized, the position information of the obstacles is detected from the images through the Yolo object detection algorithm. Spatial geometric transformation converts the obstacle information in the images into coordinates in a two-dimensional coordinate system. The grid method is used to update the internal environment map of the robot, and the coordinate positions, lengths, and widths of the obstacles are marked on the map. The obstacle avoidance parameters are obtained based on the distance between the obstacle coordinates and the initial coordinates, the minimum safety distance, and the length and width of the obstacles, and the speed of the robot is adjusted to assist the sales robot in positioning and avoiding obstacles during movement; Obstacle avoidance parameter calculation formula is obtained based on the distance between the obstacle coordinates and the initial coordinates, the minimum safety distance, the length and width of the obstacles: Among them, Distance2 is the distance value between the obstacle coordinates and the initial coordinates, Distance1 is the minimum safety distance value, (X O , Y O ) are the coordinates of the starting point O, (X Z , Y Z ) are the obstacle coordinates, denoted as point G; avoidance parameter is the calculated value of the obstacle avoidance parameter, L Z is the length value of the obstacle, W Z is the width value of the obstacle; If the calculated value of the obstacle avoidance parameter is positive, the sales robot can pass safely without decelerating; If the calculated value of the obstacle avoidance parameter is negative, the sales robot needs to take measures of decelerating and turning the angle to avoid the obstacle.

9. The path control method of an on-site sales robot according to claim 6, characterized in that, The step S4 includes the following steps: Safety braking distance calculation formula at the current speed: Among them, Distance3 is the calculated value of the safety braking distance, V is the current speed of the sales robot, a is the maximum deceleration of the sales robot, T is the reaction time of the robot, g is the acceleration due to gravity, and μ is the friction coefficient between the sales robot and the ground; Target speed value calculation formula is obtained based on the maximum deceleration value and the minimum safety distance of the sales robot: Among them, V target is the target speed, a is the maximum deceleration of the sales robot, T is the robot's reaction time, and Distance1 is the minimum safety distance value; Sales robot deceleration value calculation formula is obtained based on the difference between the current speed and the target speed of the sales robot: V s = V - V target Among them, V target is the target speed, V is the current speed of the sales robot, V s is the deceleration value of the sales robot; The steering deceleration value of the sales robot caused by steering is obtained by multiplying the adjustment coefficient by the difference between the current speed and the linear speed, and the steering deceleration value calculation formula is obtained: Wherein, R is the radius of the wheel of the sales robot, V is the current speed of the sales robot, t is the change in time, α is the deceleration value adjustment coefficient, π is the pi, and V z is the calculated value of the steering deceleration value; Final deceleration value calculation formula: V final = V s + V z Among them, V s is the deceleration value, V z is the calculated value of the steering deceleration, and V final is the final deceleration value; The step S5 includes the following steps: The distance value between the current position of the sales robot and the target node is compared with the preset distance value, and it is judged whether the target node has been reached according to the comparison result; If the distance value between the sales robot and the target node is greater than 0.3 meters, the sales robot has not reached the target node and needs to continue moving forward to the target point; If the distance value between the sales robot and the target node is less than or equal to 0.3 meters, the sales robot reaches the target node and performs corresponding tasks according to the information of different node IDs; After the sales robot completes the task, it starts to receive new task instructions, re-plans the path and executes the task.

Images