Olfactory sensing device capable of judging smell source and mobile robot

Abstract

The invention relates to an olfactory sensing device capable of judging smell sources and a mobile robot, and the device comprises two olfactory sensors which are distributed at different positions, each sensor is provided with a plurality of detection points, and different detection points generate different responses to the same smell; an actuator that drives the sensor to displace; the encoder is used for pulse encoding of the output of each olfactory sensor, and the encoding rule is that when the signal change degree exceeds a first threshold value, an output peak is triggered to serve as an event; the SNN neural model is used for receiving the time sequence formed by encoding as model input and generating an output pulse sequence representing odor classification and concentration; the processor is used for controlling the actuator to drive the two sensors to perform at least one variable angle displacement in the connecting line direction, and acquiring an output pulse sequence generated by the gas detection executed by the two sensors before and after the variable angle displacement from the SNN neural model; and the smell source direction is judged according to the concentration data of the target gas output by the two sensors before and after the variable-angle displacement.

Description

Technical Field

[0001] The present invention relates to the field of mobile robots, and in particular to an olfactory sensor device and a mobile robot capable of determining the source of an odor. Background Art

[0002] Mobile robots such as quadrupedal mechanical dogs and bipedal robots are equipped with olfactory perception functions. Based on smell, they can quickly complete tasks that assist humans, such as tracking and observing target species in protected areas, and quickly inspecting the nighttime excretion of the elderly in nursing homes.

[0003] In order to complete the assistance task, the mobile robot needs to have the ability to determine the direction of the odor source.

[0004] JP3568618B2 discloses a method for a mobile robot to track the direction of an odor source. A gas sensor and a rotating mechanism for a rotating sensor are arranged on a movable trolley. The sensor is driven to rotate 90° at each measuring point to obtain the responses of the sensors in four directions. The direction of the odor source generates a greater sensor response than other directions, and the robot is continuously driven to move in the direction of high response and repeat until the odor source is reached. The problem with the scheme is that the device has only one electronic nose (i.e., a gas sensor). At present, the multi-point array on the gas sensor has different responses to the gas at different detection points in the array. The output combination of each detection point enables the sensor to identify and distinguish more than 200 types of gas. However, the sensor detection and analysis takes time, which is basically in the millisecond level. This time plus the four rotations of the device only detect the output of one sensor each time. The sampling time at a single measurement point is too long, and the gas is diffusible in the air, so the processing result will have a large lag. Summary of the invention

[0005] The purpose of the present invention is to improve the deficiencies in the prior art and provide an olfactory sensing device that can quickly determine the source of an odor during odor diffusion, so as to meet the needs of mobile robots such as robot dogs to quickly distinguish the direction of the odor source during dynamic movement.

[0006] The present invention provides an olfactory sensing device capable of determining the source of an odor, comprising: a first olfactory sensor and a second olfactory sensor, which are distributed at different positions, each olfactory sensor being arranged with a plurality of detection points to form a detection array, and different detection points generating different responses to the same odor; an actuator, used to drive the first olfactory sensor and the second olfactory sensor to perform displacement; an encoder, which performs discrete pulse encoding on the detection output signal of each olfactory sensor, wherein the encoding rule is configured to trigger an output spike as an event when the degree of signal change exceeds a first threshold; an SNN neural model, which receives a time series formed by encoding the detection output signal as a model input, and generates an output pulse sequence characterizing the classification and concentration of the odor; a processor, which controls the actuator to drive the connection direction of the first olfactory sensor and the second olfactory sensor to perform at least one variable angle displacement, and obtains from the SNN neural model the output pulse sequence generated by the first olfactory sensor and the second olfactory sensor performing gas detection before and after the variable angle displacement, and determines the direction of the odor source according to the concentration data of the target gas output by the first olfactory sensor and the second olfactory sensor before and after the variable angle displacement.

[0007] The present invention also provides a mobile robot, comprising the above-mentioned olfactory sensing device.

[0008] The technical solution of the present invention meets the needs of a robot (robot dog) to quickly distinguish the direction of an odor source so as to quickly locate the odor source or track a dynamic odor source, while improving the temporal resolution of the sensor, establishing a good foundation for the robot to quickly move to a position to distinguish odors, and achieving a good balance in terms of computing power, response speed, accuracy, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 An exemplary structure of a mobile robot is given;

[0010] Figure 2 The internal structure of the breathing gas detection instrument is shown;

[0011] Figure 3 The detection structure diagram of the olfactory sensor is given;

[0012] Figure 4 The process of judging the source of odor is given;

[0013] Figure 5 The schematic diagram of the olfactory sensor detecting gas by changing the angle of displacement is given;

[0014] Figure 6a The principle diagram of an optional method for determining the source of odor is given;

[0015] Figure 6b The principle of another optional method for determining the source of odor is shown;

[0016] Figure 6c The principle diagram of the third alternative method for determining the source of odor is given;

[0017] Figure 7 The principle of verifying the odor source assessment results is given;

[0018] Figure 8 A schematic diagram of the principle of further evaluating the location of the odor source based on the direction of the odor source is given; and

[0019] Fig. 9 A schematic diagram of the construction of the air flow channel using a hose is given. DETAILED DESCRIPTION

[0020] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0021] Mobile robots are autonomous mobile systems, such as four-legged robot dogs, that move freely within a set area according to instructions and are used to replace humans to complete tasks. Figure 1 An exemplary structure of a mobile robot is given, including a moving mechanism 100, a breathing gas detection instrument 200, and an internal main control circuit.

[0022] The mobile mechanism 100, such as the mobile foot of a robot dog, including but not limited to wheel foot type, track type, leg foot type or track type, acts as an actuator to drive the breathing gas detection instrument 200 to move. In this solution, the robot dog drives the gas detection instrument 200 to move as a whole, and the first olfactory sensor and the second olfactory sensor move synchronously. In another optional solution, the mobile mechanism 100 can be a device that drives each sensor to move independently, thereby allowing relative displacement between the first olfactory sensor and the second olfactory sensor.

[0023] Figure 2 The internal structure of the breathing gas detection instrument is shown in FIG. The gas detection instrument 200 includes a first olfactory sensor 201 and a second olfactory sensor 202. The first olfactory sensor 201 and the second olfactory sensor 202 are distributed at different positions. Figure 3 The detection structure of the olfactory sensor is shown in the figure. For example, a 3X3 array sensor, each olfactory sensor is arranged with multiple detection points N1, N2, N3...N9 to form a detection array. Different detection points produce different responses to the same odor. Through the sampling combination of each point, more than 200 types of gases can be distinguished.

[0024] The internal main control circuit includes an encoder and a processor. The encoder is used to discretize pulses of the detection output signal of each olfactory sensor. The encoding rule is to trigger an output spike as an event when the signal change exceeds the first threshold, forming event-driven sensing. The detection output signal can be configured as an analog signal to meet higher precision requirements and short-distance transmission occasions. At this time, the encoder uses a hardware circuit to implement differentials. After the analog signal is converted into a voltage, it is compared by hardware for time differential to form a time series. It has the advantage of high conversion speed and is conducive to the realization of higher time resolution of the sensor; or, in situations where volume is considered and computing power is sufficient, such as obtaining a large computing power based on SNN, this electrical signal can be converted or directly configured as a digital signal. The encoder is implemented by software, and the processor forms a time series based on software pulse code modulation.

[0025] The processor is a device for logic processing, such as a CPU or MCU, and the processor has a built-in SNN neural model. The SNN neural model receives the time series formed by encoding the detection output signal as the model input, and generates an output pulse sequence that represents the odor classification and concentration. Each sensor chip is arranged with N detection points, each detection point is used to detect different odors, that is, the same odor has different effects on different detection points; by training the detection of a large number of different odors, the training model of the SNN network can be obtained. After the weights of the training model are solidified, they are stored in the sensor chip to obtain a sensor chip with an SNN neural pulse network for detecting different odors.

[0026] The processor obtains a set of data (L1, L2, L3....LN) in each detection cycle, which represents the detection value of each detection point; the encoder converts it into a pulse signal and enters the SNN network for analysis. In each analysis cycle, the SNN network analyzes multiple groups of pulse signals to obtain the type and relative concentration data of the odor and send it to the processor.

[0027] Based on the analysis of odor classification and concentration recognition using the pulse neural network (SNN), further analysis of the odor source is achieved using logical rule judgment. Figure 4 A flow chart for determining the source of an odor is provided. The method for determining the source of an odor includes: S1, controlling an actuator to drive the connecting direction of the first olfactory sensor and the second olfactory sensor to perform at least one variable angle displacement, such as Figure 5 As shown; S2, obtaining from the SNN neural model the output pulse sequence generated by the first olfactory sensor and the second olfactory sensor performing gas detection before and after the variable angle displacement; S3, judging the direction of the odor source based on the concentration data of the target gas output by the first olfactory sensor and the second olfactory sensor before and after the variable angle displacement.

[0028] The technical advantages of the above implementation scheme are: (1) for the changing characteristics of 3X3 or more array sensors, due to the complex characteristic changes, artificial learning analysis using a pulse neural network (SNN) can greatly improve the accuracy. At the same time, the pulse code is converted by an encoder and then input into the SNN to perform event-responsive gas classification and concentration identification, further improving the sensor's time resolution and establishing a good foundation for the robot to quickly move and distinguish odors; (2) the analysis of the odor source adopts a logical rule judgment method. Since the data changes are more intuitive, the use of a logical rule judgment method greatly reduces the computing power requirements. In addition, the judgment rule uses the difference in concentration values ​​before and after the two olfactory sensors change angles. When the odor is in the process of diffusion, the direction of the odor source can also be quickly distinguished to quickly locate the odor source or track the dynamic odor source; (3) The combination of the two analysis methods achieves a good balance in computing power, response speed, accuracy, and other aspects.

[0029] To facilitate understanding of the above-mentioned method for determining the source of odor, four optional specific implementations are given below to assist understanding.

[0030] As an optional specific implementation solution, refer to Figure 6a , the method for determining the source of odor further comprises:

[0031] The first olfactory sensor 201 and the second olfactory sensor 202 are started to continuously perform gas detection. In each round of detection, the concentration of the two olfactory sensors 201 and 202 for the target gas is obtained from the SNN neural model, and the actuator is controlled to drive the two olfactory sensors 201 and 202 to synchronously perform variable angle displacement after each round of gas detection:

[0032] When the difference between the concentration values ​​of the first olfactory sensor 201 and the second olfactory sensor 202 is less than the second threshold, the odor source is marked as the vertical midline direction of the line connecting the first olfactory sensor 201 and the second olfactory sensor 202 in this round.

[0033] Compared with one sensor, using two sensors can effectively detect different spatial positions at the same time, avoiding the time difference caused by the movement of different spatial positions by one sensor, and the error caused by the change of the diffusion shape of the detected object; compared with more than two sensors, using two can effectively save costs.

[0034] As another optional implementation, refer to Figure 6b , the method for determining the source of odor further comprises:

[0035] The distance between the first olfactory sensor and the second olfactory sensor is denoted as l, and the measured gas concentrations are denoted as S1 and S2 respectively:

[0036] At the initial measurement point P0, select any direction as the x-axis direction, the direction of the connecting line is parallel to the x-axis direction, and use S1 and S2 to calculate the difference ΔS=s2-s1 / l as the directional derivative of the concentration function f(x0,y0) at point P0 along the x-axis direction.

[0037] Drive the two olfactory sensors to perform variable angle displacement to adjust the direction, measure the directional derivatives in multiple directions, and select the directional derivative with the largest modulus as the estimate of the concentration gradient. The corresponding direction is the direction of the odor source.

[0038] This measurement method has a simple principle and is easy to operate. It has no special requirements for the measurement direction, does not require additional calculations, and requires little computing power resources.

[0039] As another optional specific implementation scheme, the method for determining the source of odor further includes:

[0040] The distance between the first olfactory sensor and the second olfactory sensor is denoted as l, and the measured gas concentrations are denoted as S1 and S2 respectively;

[0041] At the initial measurement point P0, at least three directions are selected, the direction of the connecting line is driven to be parallel to each direction, and the difference ΔS=s2-s1 / l is calculated with S1 and S2 as the concentration function f(x0, y0) at point P0 along these directions Directional derivative of

[0042] Estimate the concentration gradient using the least squares method:

[0043]

[0044] This solution is calculated based on multiple measurement results, which reduces the direction estimation error and makes the result more stable and accurate.

[0045] As another optional implementation, refer to Figure 6c , the method for determining the source of odor further comprises:

[0046] The distance between the first olfactory sensor and the second olfactory sensor is denoted as l, and the measured gas concentrations are denoted as S1 and S2 respectively:

[0047] At the initial measurement point P0, select any direction as the x-axis direction, the direction of the connecting line is parallel to the x-axis direction, and use S1 and S2 to calculate the difference ΔS=s2-s1 / l as the directional derivative of the concentration function f(x0, y0) at point P0 along the x-axis direction.

[0048] Drive the two olfactory sensors to set the angle to perform variable angle displacement, preferably 90° to save computing power. After the variable angle displacement, measure S1 and S2 again to calculate the difference, which is used as the directional derivative of the concentration function at point P0 along the second direction.

[0049] Estimate the gradient of the concentration function This direction is taken as the direction of the odor source.

[0050] This solution only requires measurement results in any two perpendicular directions, with a small number of measurements, thus achieving a balance between computing power and accuracy.

[0051] As an optional verification implementation method, it includes: controlling the actuator to rotate the perpendicular midline of the line connecting the two olfactory sensors to the direction of the odor source, and starting gas detection; based on the difference in concentration values ​​of the target gas between the first olfactory sensor and the second olfactory sensor, judging based on the threshold that the evaluation of the odor source direction is correct when the difference is low. By verifying the evaluated odor source direction, the accuracy of the direction judgment is further improved. Or, as another optional verification implementation method, refer to Figure 7 , at the measuring point P 01 Estimating the direction of the odor source Control the actuator in the direction of the odor source Move the two olfactory sensors a certain distance to reach the new measurement point P 02 , again estimate the direction of the odor source Based on the direction of the smell Odor source direction The difference between the two is based on the threshold judgment. When the difference is low, the direction of the odor source is determined to be correct. Compared with the above verification scheme, the concentration change caused by the position change can more accurately verify the evaluation result.

[0052] Figure 8 The principle of further evaluating the location of the odor source based on the odor source direction is given, including: at the initial measurement point P, estimating the direction of the odor source Control the actuator to move the two olfactory sensors a distance L to the new measurement point P1, and estimate the direction of the odor source again. Based on the two estimated odor source directions, the odor source position is calculated through geometric relationships. Further, for the moving direction selection of the distance L, the actuator is controlled to move toward the current measurement point P0 in a direction perpendicular to the current concentration gradient. At this time, the odor source position can be estimated as (Ltanαcosθ, Ltanαsinθ) in a coordinate system with P0 as the origin and the initial measurement direction as the x-axis and y-axis directions.

[0053] Reference Figure 2The gas detection instrument 200 is provided with a neutralizing air chamber 203, and each olfactory sensor is provided with a corresponding breathing component. The first airflow channel of each group of breathing components converges to the neutralizing air chamber, and the air compression device is provided on the converging main channel for unified driving, simulating the double nose structure of the human body. Each group of breathing components is provided with an air inlet and outlet, a first airflow channel 205, and a second airflow channel 206.

[0054] In each group of breathing components, the air inlet and outlet serve as windows for gas input and gas output of the detection instrument, and are connected to the neutralization air chamber 203 to form a first airflow channel 205, which serves as a path for the detection gas to enter the neutralization air chamber 203. The first airflow channel 203 has a detection section, and the corresponding olfactory sensor is built into the detection section. The first airflow channel 205 is built with an air compression device 2051 and an air intake valve 2052. The air compression device 2051 drives the air chamber to inlet, and can be selected as a device that generates gas pressure energy through mechanical work, such as an air pump. The air intake valve 2052 is arranged at the inlet and / or outlet of the air pump to form a one-way air intake to the neutralization air chamber 203. The air intake valve 2052 can be an electrically controlled air intake valve, which is opened when needed and closed when not needed; or a one-way valve formed by a mechanical structure such as a duckbill valve can be used to save control logic participation. The neutralization air chamber 203 has an air outlet, which is connected back to the detection section to form a second airflow channel 206, which serves as a path for the neutralization gas to flush the sensor. The second air flow channel 206 has a built-in electrically controlled air outlet valve 2061 to control the air outlet of the neutralization air chamber 203 .

[0055] The processor is coupled to each olfactory sensor, air compression device, and electric control valve. The neutralization chamber 203 stores gas for neutralization, which is pre-input and stored using a clean gas source, and the clean gas source is configured so that the concentration of the target gas or the concentration and the percentage of the reference do not exceed the set threshold at most. The neutralization chamber 203 has a first state of air intake through the first air channel 205 driven by the air compression device 2051, and a second state of deflation and cleaning of the olfactory sensor through the second air channel 206 when the electric control outlet valve 2061 is opened.

[0056] During the process of the gas to be tested being inhaled from the air inlet and outlet, the sensor detection is implemented in the front section of the first air flow channel 205, and then it is sent to the air chamber to be neutralized with the clean gas stored in advance, and the gas in the air chamber is compressed at the same time; after the second air flow channel 206 is opened, the neutralized gas flushes the sensor and is exhaled through the air inlet and outlet. Among them, the gas detection instrument achieves the following effects with the help of the anthropomorphic bionics of the human nose and lungs and their connections: the air chamber, as a bionic lung, stores the gas for neutralization, and neutralizes with the gas to be tested to quickly reduce the gas concentration; the unique air flow channel topology structure composed of the air inlet and outlet, the sensor, the air chamber, the first and second flow channels forms a coordinated cooperation between the breathing bionic electronic nose and the bionic breathing lung, simulating the breathing and ventilation process of the human nose and lungs, achieving both gas detection and sensor flushing during breathing, and achieving excellent cleaning effect while the detection instrument works efficiently, which is conducive to the dynamic response of the mobile robot to quickly move the position to distinguish the odor.

[0057] In the scheme where the mobile mechanism 100 rotates the first olfactory sensor 201 and the second olfactory sensor 202 as a whole, such as using a mechanical foot to achieve translation and / or rotation of the gas detection instrument 200, the air flow channel uses a rigid component to form an air flow channel to improve the overall stability; in the scheme where the mobile mechanism 100 drives each sensor to move independently, refer to Fig. 9 The air flow channel is formed by a hose to support the relative displacement of the first olfactory sensor 201 and the second olfactory sensor 202 .

[0058] Furthermore, the neutralization air chamber 203 uses an elastic airbag, which swells when inflated and forms greater flushing kinetic energy when deflated in combination with the elastic recovery of the airbag. Preferably, the neutralization air chamber 203 is equipped with a molecular filtration membrane or an air filter such as activated carbon to purify the gas, thereby extending the usable time of the gas stored in the air chamber.

[0059] The detection cycle of the gas detection instrument 200 is composed of two sub-cycles T1 and T2, and T1 and T2 are alternating in time. In T1, the electrically controlled air inlet valve or the one-way valve is opened, the electrically controlled air outlet valve is closed, and the air compression device drives the air intake to form the first state, and the corresponding olfactory sensor is started for detection; in T2, the electrically controlled air inlet valve or the one-way valve is closed, the air compression device is closed, and the electrically controlled air outlet valve is opened to switch from the first state to the second state, and the detection of the corresponding olfactory sensor is stopped. In this scheme, the air pressure difference is generated by compressing the interior of the air chamber with the help of the air compression device, forming an inhalation and exhalation alternating detection and cleaning, which is conducive to the application of occasions with high requirements for detection speed.

[0060] Or, as another alternative embodiment, refer to Figure 2, a front-end electric control valve 207 of a coupling processor is arranged in the flow channel between the air inlet and outlet and the corresponding detection section, wherein the air compression device is selected as a device capable of adjusting the air intake volume, and / or the front-end electric control valve is a flow valve. When the gas concentration is detected to be too high, the air intake volume of the air inlet and outlet is reduced or closed by the air compression device and / or the front-end electric control valve to simulate the human body's nose-covering action. On this basis, further, the detection cycle of the gas detection instrument is alternately constituted by two sub-cycles T1 and T2. In T1, the front-end electric control valve is opened, the electric control air inlet valve or the one-way valve is opened, the electric control air outlet valve is closed, and the air compression device drives the air intake to form the first state, and starts the corresponding olfactory sensor for detection; in T2, the front-end electric control valve is closed, the electric control air inlet valve or the one-way valve is opened, the electric control air outlet valve is opened, and the air compression device drives to form the first state and the second state at the same time, and stops the detection of the corresponding olfactory sensor. In this scheme, with the help of the driving of the air compression device in the flushing stage, a one-inhalation-multiple-exhalation cycle cleaning is formed, which is conducive to the application of occasions with high requirements for sensor detection accuracy. More preferably, the gas detection instrument is further provided with a measuring device for sensing the air intake amount of the neutralization chamber, and the measuring device can reflect the air intake amount through the change of the internal air pressure by using a gas pressure sensor, or directly detect the air intake amount by using a flow meter, or control the air intake rate per unit time, and reflect the air intake amount based on time by using a timer. The processor is coupled to the measuring device, and the internal pressure of the neutralization chamber is evaluated according to the air intake amount. Based on the comparison between the internal pressure of the neutralization chamber and the third threshold value, the processor switches from the sub-period T1 to T2 to perform flushing when the internal pressure of the neutralization chamber is greater than the third threshold value.

[0061] As another optional implementation, after executing the gas detection process or when the concentration of the target gas output from the neutralization chamber for cleaning the sensor or the percentage of the concentration to the benchmark is higher than the fourth threshold, a first request to replace the gas used for neutralization in the neutralization chamber is generated to avoid the internal gas source of the chamber becoming a pollution source after multiple neutralizations, which affects the detection. In the case where the gas detection instrument is configured on a mobile robot, further, the control robot is triggered to move to the clean gas area to perform the ventilation process based on the reception of the first request. By combining the mobility of the robot, the air in the air chamber is kept clean in real time, ensuring the continuous operation capability of the mobile robot's olfactory perception. Furthermore, the clean gas area is configured as a space where the concentration of the target gas or the percentage of the concentration to the benchmark is lower than the fifth threshold, which can be addressed by detection or by pre-dividing the specified area by an electronic fence.

[0062] The ventilation process is configured to continuously switch between the first state and the second state until the concentration of the target gas output by the neutralization chamber or the percentage of the concentration to the benchmark is lower than the sixth threshold. As an optional implementation, the concentration of the target gas output by the neutralization chamber or the percentage of the concentration to the benchmark can be detected by reusing the olfactory sensor and characterized by the olfactory sensor data collected in the second state. In the second state, the gas in the air chamber is fully neutralized, which can obtain higher accuracy and is conducive to cost simplification; or, the neutralization chamber is built with an olfactory sensor coupled to a processor to independently detect the concentration of the target gas output by the neutralization chamber or the percentage of the concentration to the benchmark.

[0063] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention, rather than to limit the scope of protection of the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the essence and scope of the technical solution of the present invention.

Claims

1. An olfactory sensing device capable of determining the source of an odor, characterized in that: include: The first olfactory sensor and the second olfactory sensor are distributed at different positions, each olfactory sensor is arranged with a plurality of detection points to form a detection array, and different detection points produce different responses to the same odor; an actuator, used for driving the first olfactory sensor and the second olfactory sensor to move; An encoder, which discretizes pulses to encode the detection output signal of each olfactory sensor, wherein the encoding rule is configured to trigger an output spike as an event when the degree of signal change exceeds a first threshold; The SNN neural model receives the time series formed by encoding the detection output signal as the model input, and generates an output pulse sequence representing the odor classification and concentration; The processor controls the actuator to drive the connection direction of the first olfactory sensor and the second olfactory sensor to perform at least one variable angle displacement, and obtains from the SNN neural model the output pulse sequence generated by the first olfactory sensor and the second olfactory sensor performing gas detection before and after the variable angle displacement, and determines the direction of the odor source according to the concentration data of the target gas output by the first olfactory sensor and the second olfactory sensor before and after the variable angle displacement.

2. The olfactory sensor device according to claim 1, characterized in that The processor performs at least the following steps: The first olfactory sensor and the second olfactory sensor are started to continuously perform gas detection, the concentration of the two olfactory sensors for the target gas is obtained from the SNN neural model in each round of detection, and the actuator is controlled to drive the two olfactory sensors to synchronously perform variable angle displacement after each round of gas detection; When the difference between the concentration values ​​of the first olfactory sensor and the second olfactory sensor is less than the second threshold, the odor source is marked as the vertical midline direction of the line connecting the first olfactory sensor and the second olfactory sensor in this round.

3. The olfactory sensor device according to claim 1, characterized in that The processor performs at least the following steps: The distance between the first olfactory sensor and the second olfactory sensor is denoted as l, and the measured gas concentrations are denoted as S1 and S2 respectively; At the initial measurement point P0, any direction is selected as the x-axis direction, the direction of the connecting line is parallel to the x-axis direction, and the difference ΔS=s2-s1 / t is calculated with S1 and S2 as the directional derivative of the concentration function f(x0, y0) at point P0 along the x-axis direction. Drive the two olfactory sensors to perform variable angle displacement to adjust the direction, measure the directional derivatives in multiple directions, and select the directional derivative with the largest modulus as the estimate of the concentration gradient. The corresponding direction is the direction of the odor source.

4. The olfactory sensor device according to claim 1, characterized in that The processor performs at least the following steps: The distance between the first olfactory sensor and the second olfactory sensor is denoted as l, and the measured gas concentrations are denoted as S1 and S2 respectively; At the initial measurement point P0, at least three directions are selected, the direction of the connecting line is driven to be parallel to each direction, and the difference ΔS=s2-s1 / l is calculated with S1 and S2 as the concentration function f(x0, y0) at point P0 along these directions Directional derivative of Estimate the concentration gradient using the least squares method: in 5. The olfactory sensor device according to claim 1, characterized in that: The processor performs at least the following steps: The distance between the first olfactory sensor and the second olfactory sensor is denoted as l, and the measured gas concentrations are denoted as S1 and S2 respectively; At the initial measurement point P0, select any direction as the x-axis direction, the direction of the connecting line is parallel to the x-axis direction, and use S1 and S2 to calculate the difference ΔS=s2-s1 / l as the directional derivative of the concentration function f(x0, y0) at point P0 along the x-axis direction. Drive the two olfactory sensors to set the angle to perform variable angle displacement, measure S′1 and S′2 again to calculate the difference, and use it as the directional derivative of the concentration function at point P0 along the second direction Estimate the gradient of the concentration function This direction is taken as the direction of the odor source.

6. The olfactory sensor device according to claim 5, characterized in that: The set angle is 90°.

7. The olfactory sensing device according to any one of claims 3 to 6, characterized in that: At the measuring point P 01 Estimating the direction of the odor source Control the actuator in the direction of the odor source Move the two olfactory sensors a certain distance to reach the new measurement point P 02 , again estimate the direction of the odor source Based on the direction of the smell Odor source direction The difference between them is used to verify the estimation result of the odor source direction.

8. The olfactory sensing device according to any one of claims 3 to 6, characterized in that: Controlling the actuator to rotate the perpendicular midline of the line connecting the two olfactory sensors to a direction toward the source of the odor, thereby starting gas detection; Based on the difference in concentration values ​​of the target gas detected by the first olfactory sensor and the second olfactory sensor, the estimation result of the odor source direction is verified.

9. The olfactory sensing device according to any one of claims 3 to 7, characterized in that: The processor performs at least the following steps: Estimate the direction of the odor source at the initial measurement point P0 Control the actuator to move the two olfactory sensors a distance L to the new measurement point P1, and estimate the direction of the odor source again. Based on the two estimated odor source directions, the odor source position is calculated through geometric relationship.

10. The olfactory sensor device according to claim 9, characterized in that: Control the actuator to move a distance L along the direction perpendicular to the current concentration gradient at the current measurement point P0; In a coordinate system with P0 as the origin and the initial measurement direction as the x-axis and y-axis directions, the estimated position of the odor source is (Ltanαcosθ, Ltanαsinθ).

11. The olfactory sensor device according to claim 1, characterized in that: The detection output signal of the olfactory sensor is configured as an analog signal, and the analog signal is converted into a voltage and then compared with the time difference through a hardware encoder to form a time series; Alternatively, the detection output signal of the olfactory sensor is converted or configured into a digital signal, and the encoder forms a time sequence based on software pulse code modulation.

12. The olfactory sensor device according to claim 1, characterized in that: Includes a neutralizing air chamber; Each olfactory sensor is correspondingly provided with a breathing assembly, and each group of breathing assemblies is provided with an air inlet and outlet, a first air flow channel, and a second air flow channel; In each group of breathing components, the air inlet and outlet are connected to the neutralization air chamber to form a first air flow channel, the first air flow channel has a detection section, the corresponding olfactory sensor is built in the detection section, the neutralization air chamber is connected back to the detection section via a second air flow channel, the first air flow channel has at least one air compression device built in to drive the air intake of the neutralization air chamber, and an electrically controlled air intake valve or a one-way valve forms a one-way air intake to the neutralization air chamber, the second air flow channel has an electrically controlled air outlet valve built in to control the air outlet of the neutralization air chamber, the neutralization air chamber stores gas for neutralization, and has at least a first state of air intake through the first air flow channel driven by the air compression device, and a second state of deflation and cleaning of the olfactory sensor through the second air flow channel when the electrically controlled air outlet valve is opened; The processor is respectively coupled to each olfactory sensor, the air compression device, and the electric control valve.

13. The olfactory sensor device according to claim 12, characterized in that: The detection cycle of each group of respiratory components consists of two sub-cycles T1 and T2. The internal control logic of the respiratory components includes: In T1, the front-end electric control valve is opened, the electric control outlet valve is closed, the air compression device drives the air intake to form the first state, and the corresponding olfactory sensor is started for detection; In T2, the front-end electric control valve is maintained, the air compression device is closed, and the electric control air outlet valve is opened to switch from the first state to the second state, and the detection of the corresponding olfactory sensor is stopped.

14. The olfactory sensor device according to claim 12, characterized in that: A front-end electric control valve coupled to a processor is arranged in the flow channel between the air inlet and outlet and the corresponding detection section.

15. The olfactory sensor device according to claim 14, characterized in that: The detection cycle of each group of respiratory components consists of two sub-cycles T1 and T2. The internal control logic of the respiratory components includes: In T1, the front-end electric control valve is opened, the electric control outlet valve is closed, the air compression device drives the air intake to form the first state, and the corresponding olfactory sensor is started for detection; In T2, the front-end electric control valve is closed, the electric control air outlet valve is opened, and the air compression device is driven to form the first state and the second state at the same time, and the detection of the corresponding olfactory sensor is stopped.

16. The olfactory sensor device according to claim 13 or 14, characterized in that: including a measuring device for sensing the amount of air intake into the neutralizing chamber; The processor is coupled to the measuring device and is used to evaluate the internal pressure of the neutralization chamber according to the intake air volume, and to use the comparison between the internal pressure of the neutralization chamber and the third threshold as the judgment condition for switching between sub-periods T1 and T2.

17. The olfactory sensor device according to claim 12, characterized in that: The first air flow channels of each group of breathing components converge to the neutralizing air chamber, and the air compression device is arranged on the converging main channel for unified driving.

18. The olfactory sensor device according to claim 12, characterized in that: After the gas detection process is performed or when the gas cleanliness in the neutralization chamber is lower than a fourth threshold, a first request for replacing the gas used for neutralization in the neutralization chamber is generated.

19. The olfactory sensor device according to claim 18, characterized in that: In the second state, olfactory sensor data is collected to characterize the cleanliness of the gas inside the neutralization chamber.

20. A mobile robot, characterized in that: Comprising the olfactory sensing device according to any one of claims 1-19.

Images