WO2017008454A1

WO2017008454A1 - Robot positioning method

Info

Publication number: WO2017008454A1
Application number: PCT/CN2015/099467
Authority: WO
Inventors: 欧勇盛; 肖晨; 肖月; 彭安思
Original assignee: 中国科学院深圳先进技术研究院
Priority date: 2015-07-10
Filing date: 2015-12-29
Publication date: 2017-01-19
Also published as: CN105004336A

Abstract

A robot positioning method, comprising: when a robot moves, collecting orientation data by a photoelectric image sensor, a position-sensitive sensor, a speedometer and a gyroscope (201); inputting the orientation data collected by the photoelectric image sensor into all the local filters and a main filter; inputting the orientation data collected by the position-sensitive sensor into a first local filter; the first local filter processing the input orientation data, and inputting a processing result into the main filter; inputting the orientation data collected by the speedometer into a second local filter; the second local filter processing the input orientation data, and inputting a processing result into the main filter; and inputting the orientation data collected by the gyroscope into a third local filter; the third local filter processing the input orientation data, and inputting a processing result into the main filter (202); and the main filter processing the data output by the photoelectric image sensor and the data output by all the local filters, so as to obtain positioning information about the robot (203).

Description

Positioning method of a robot

The present application claims priority to Chinese Patent Application No. 201510403964.5, filed on Jan. 10,,,,,,,,,,,,,,,,

Technical field

The present invention relates to the field of robot technology, and in particular, to a method for positioning a robot.

Background technique

The positioning and navigation technology of indoor cleaning robots is a key research hotspot in the field of service robot research. In related technology research, positioning and navigation technology is the key to realize intelligent and fully autonomous movement of mobile robots. For navigation, it is to let the robot identify the strange environment independently and identify its own relative position without external human intervention, so as to accurately move to the specified target position and complete the designated work task. For the sweeping robot, it is to accurately complete the cleaning task according to the established sweeping path and strategy. At present, the common positioning technology of sweeping robots is as follows:

Patent application (1) A sweeping robot intelligent navigation system, application number: 201310087158.2

It is an indoor positioning scheme designed by infrared technology. A base station is placed indoors. An infrared transmitting head and a noise wave transmitting head are installed on the base station, and then an infrared receiving head and a noise wave receiving head are installed on the cleaning robot. The robot calculates the distance by receiving the infrared and acoustic signals transmitted by the base station to achieve the purpose of positioning. The requirement is that the trajectory of the robot motion must always be perpendicular to the base station.

The infrared locating technology of the technical solution is susceptible to the occlusion of the infrared signal, and the distance is affected by the signal intensity distribution, so that the positioning is not reliable.

Patent application (2) Speed measuring, ranging system and method for sweeping robot based on universal wheel, application number: 201410266593.6

The technology mainly designs an encoder disk. The Hall sensor and the multi-pole magnetic ring are mounted on the wheel to form an electromagnetic induction system. When the wheel rotates, the Hall sensor can generate current, and the processor determines the wheel by the magnitude of the current. Speed and number of revolutions. Then, the cumulative speed is used to achieve the purpose of robot positioning.

Although the locating technology based on the odometer is relatively easy to implement, it has the problem of accumulated error, which will affect the positioning accuracy of the robot after a long time of work.

Patent Application (3) A vision-based robot indoor positioning navigation method. Application No.: 201010611473.7

It is a vision-based robot indoor positioning and navigation method. According to the idea of two-dimensional code, design a simple road sign that is easy to identify, easy to identify, contains absolute position coordinates and has certain error correction capability, and sets the road sign on the ceiling. The camera is mounted on the robot and the optical axis is perpendicular to the ceiling. Then, through the image threshold segmentation, connected domain extraction, contour curve matching, and landmark feature recognition, a series of steps are used to locate the road sign and analyze the coordinate information contained in the road sign. Finally, the robot's current position and heading angle are obtained by the robot's position estimation algorithm.

On the one hand, the visual positioning method of the technical solution may need to change the appearance of the original indoor environment. On the other hand, the algorithm has low reliability, poor anti-interference ability, high performance requirement on the processor, and high cost.

Patent application (4) Earth-moving robot obstacle avoidance, positioning system and method. Application number: 201410266597.4

A nine-segment collision detector is introduced. Since there are many collision detectors around the robot, it is possible to detect collisions at multiple angles, which is helpful for obstacle avoidance and positioning. Then, in terms of distance calculation, the Hall sensor is used to encode the wheel speed, and the accumulated speed constitutes an odometer.

The technical solution based on collision avoidance is very blunt and not intelligent, and the long-term use of the collision detector may cause mechanical damage and affect reliability.

In addition, the laser sensor-based positioning technology has attracted much attention in the navigation field due to its high precision and strong data reliability. However, the large size of the laser sensor is inconvenient to install on small indoor robots, and the data volume is inconvenient to handle. The point is that it is expensive and has not yet been promoted in home service robot applications. GPS technology is widely used in the navigation field, but it has no signal indoors and is not suitable for indoor robot positioning.

Summary of the invention

In order to solve the problems of the prior art, the present invention provides a positioning method of a robot, which can make the positioning accuracy of the robot moving while being low, and the cost is low.

To achieve the above object, the present invention provides a method for positioning a robot, the method comprising:

When the robot moves, the corresponding orientation data is collected by the photoelectric image sensor, the position sensitive sensor, the odometer and the gyroscope;

The orientation data collected by the photoelectric image sensor is input to the first local filter, the second local filter, the third local filter, and the main filter; the orientation data collected by the position sensitive sensor is input to the first local filter. The first local filter processes the orientation data collected by the photoelectric image sensor and the orientation data collected by the position sensitive sensor according to the latest information fed back by the main filter, and inputs the processing result to the main filter; the odometer collects The orientation data is input to a second local filter; the second local filter is based on the most feedback of the main filter The new positioning information processes the orientation data collected by the photoelectric image sensor and the orientation data collected by the odometer, and inputs the processing result to the main filter; the orientation data collected by the gyroscope is input to the third local filter; The local filter processes the orientation data collected by the photoelectric image sensor and the orientation data collected by the gyroscope according to the latest information fed back by the main filter, and inputs the processing result to the main filter;

The main filter processes data output by the photoelectric image sensor, data output by the first local filter, data output by the second local filter, and data output by the third local filter to obtain positioning information of the robot; Information is fed back to the first local filter, the second local filter, and the third local filter.

The above technical solution has the following beneficial effects: the technical solution adopts the photoelectric image sensor, has low cost, reliable data, and low requirements on processor performance. Further, the technical solution combines the information of the four sensors of the photoelectric image sensor, the position sensitive sensor, the odometer and the gyroscope, so that the positioning error does not accumulate and the positioning accuracy is high; and there is no need to arrange any road signs indoors, and the robot There is no limit to the path of motion.

In addition, the technical solution adopts a position sensitive sensor, and can simultaneously realize the functions of obstacle avoidance and returning to the charging station.

DRAWINGS

In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly described below. It is obvious that the drawings in the following description are only some embodiments of the present invention. Other drawings may also be obtained from those of ordinary skill in the art in light of the inventive work. In the drawing:

1 is a flow chart of a positioning method of a robot according to the present invention;

Figure 2 is a top view of the position of the photoelectric image sensor;

3 is a flow chart of acquiring data collected by each sensor by the host computer software of the robot;

4 is a schematic structural diagram of a joint Kalman filter algorithm according to the present invention;

FIG. 5 is a schematic diagram of the distance measurement of the position sensitive sensor of the present invention.

detailed description

The embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. The illustrative embodiments of the present invention and the description thereof are intended to explain the present invention, but are not intended to limit the invention.

The working principle of this technical solution is: Most robots are faced with various size problems, such as price and positioning error; in addition, indoor moving objects affect the original algorithm of the robot; The person needs to deploy a lot of sensors and accessories in the room, which changes the original appearance of the room, and it is not what we want. With these defects in mind, we have proposed a better approach. The technical solution adopts a lower cost sensor, combines the data of each sensor, overcomes the respective deficiencies, and makes the positioning more accurate. Another is its energy efficiency. Due to the high quality positioning navigation, he will not go back and forth in the same place, so that it does not need to be recharged for a long time.

The technical solution adopts the photoelectric image sensor as the main motion parameter measuring unit, and integrates it with the data of the odometer, the gyroscope and the position sensitive sensor to increase the reliability thereof. Since the photoelectric image sensor data is not affected by the drift of the robot, it is not easy to generate cumulative errors, and it is installed under the robot, and is not interfered by the moving object, and can provide accurate motion data to perform positioning calculation to the utmost. In order to realize the intelligent obstacle avoidance of the robot, the technical solution is based on the position sensitive sensor distance measuring system, which enables the robot to perceive obstacles in a large range and make an obstacle avoidance strategy.

Based on the above working principle, the present invention proposes a positioning method of a robot. As shown in Figure 1. The method includes:

Step 101): when the robot moves, the corresponding orientation data is collected by the photoelectric image sensor, the position sensitive sensor, the odometer and the gyroscope;

Step 102): the orientation data collected by the photoelectric image sensor is input to the first local filter, the second local filter, the third local filter, and the main filter; and the orientation data collected by the position sensitive sensor is input to the first a local filter; the first local filter processes the orientation data collected by the photoelectric image sensor and the orientation data collected by the position sensitive sensor according to the latest information fed back by the main filter, and inputs the processing result to the main filter; The orientation data acquired by the odometer is input to the second local filter; the second local filter processes the orientation data collected by the photoelectric image sensor and the orientation data collected by the odometer according to the latest positioning information fed back by the main filter, and the processing is processed The result is input to the main filter; the orientation data acquired by the gyroscope is input to the third local filter; and the third local filter collects the orientation data and the gyroscope collected by the photoelectric image sensor according to the latest information fed back by the main filter The orientation data is processed, and the processing result is input to the main filter;

Step 103): the main filter processes the data output by the photoelectric image sensor, the data output by the first local filter, the data output by the second local filter, and the data output by the third local filter to obtain the positioning of the robot. Information; at the same time, information is fed back to the first local filter, the second local filter, and the third local filter.

As shown in FIG. 2, it is a top view of the position of the photoelectric image sensor. The photoelectric image sensor is used as the main robot motion parameter sensing unit. The photoelectric image sensor 1 is mounted at the bottom of the robot body 2 and between the robot wheels 3. Bring it close to the ground. Transfer the information received by the sensor to the host computer of the robot through the serial port. Software, the PC software extracts the displacement and direction data of the robot motion, and then applies to the indoor positioning of the robot. The photoelectric image sensor obtains the distance and direction of the robot movement by comparing the difference between the two images.

As shown in FIG. 3, a flow chart of data collected by each sensor is acquired for the host computer software of the robot. The displacement information detected by the sensor is placed in the corresponding X, Y register, and an X, Y register is read for each distance to prevent register overflow. The internal motion status register marks whether the sensor generates displacement. If there is no displacement, the value of the motion status register is circulated continuously; if a displacement is generated, the values of the X and Y registers are read, and the values in the X and Y directions are accumulated and saved. . Then clear the X, Y registers, continue to read the motion status register, and so on.

The technical solution uses various sensors such as photoelectric image sensors, position sensitive sensors, odometers, gyroscopes, etc. The output data of various sensors is different in form, and these data must be converted into a unified standard, and the data of the multi-sensor is used for the target. State estimation, looking for a state vector that best fits the observed data. The technical solution adopts the structure of the joint Kalman filter to fuse various sensor data.

The local filter is filtered according to the state equation and the measurement equation, and the filtering result of each step is transmitted to the main filter, and the main filter completes the optimal synthesis of the information to form a comprehensive information of the global system.

P _g . The global estimate of the synthesis by the main filter after each filtering phase is completed

And P _g . At the same time, the information distribution amount formed by the main filter according to the information distribution principle is fed back to each local filter.

Record the state estimation vector of the local filter i as

The system covariance matrix is Q _i , and the state vector covariance matrix is P _i , where i=1...N. The state estimation vector, system covariance matrix and state vector covariance matrix of the main filter are:

P _g , Q _g . The calculation process of the joint Kalman filter is as follows (1) to (11):

a. Given initial value:

Assume that the initial value of the global state at the start time is X ₀ , the system covariance matrix is Q ₀ , and the state vector covariance matrix is P _{0 .} This information is regularly distributed to the local filters and global filters by the information distribution factor.

Where β _i satisfies the principle of information conservation β ₁ + β ₂ + ... + β _N + β _m =1, 0 ≤ β _i ≤ 1.

b. The time update of the information is:

X _i,(k|k-1) =Φ _k|k-1 X _i,(k-1) (6)

c. When the measurement of the information is updated, the measurement of the i-th local filter is updated to:

d. The optimal information fusion conforms to the following formula:

The joint Kalman filter is an ideal method. The basic idea of the joint Kalman filter design is: first decentralized processing, then global fusion, that is, selecting a subsystem with comprehensive information, high output rate and reliable reliability as a common reference system among many non-similar subsystems, and other The subsystems are combined in pairs to form several sub-filters.

For the present technical solution, the common reference system is a photoelectric image sensor. In order to balance the accuracy and fault tolerance of the system, the technical solution adopts a fusion reset structure. In this structure, each subsystem is sent to the main filter after Kalman filtering, and the main filter only completes the synthesis of the local filtering information without filtering. At this time, the main filter state equation has no information distribution, so The estimated value of the main filter is taken as a global estimate. which is:

FIG. 4 is a schematic structural diagram of a joint Kalman filter algorithm according to the present invention. Let the state vector of the joint Kalman filter be X _g , the variance matrix P _g , the state vector of the local filter be X _i , the variance matrix be P _i , the state vector of the main filter be X _m , and the variance matrix be P _m The measurement information is measured by the inverse R ^{-1 of the} noise variance matrix. The system information is represented by the inverse Q ^{-1 of the} system noise variance matrix, and the filtered estimation error information is represented by the inverse P ^{-1 of the} estimated error variance matrix. The fusion algorithm of the present technical solution includes four filters, namely a main filter, a local filter 1, a local filter 2, and a local filter 3. The local filter 1 is responsible for the information collected by the position sensitive sensor and the information collected by the photoelectric image sensor; the local filter 2 is responsible for the information collected by the photoelectric image sensor and the information collected by the odometer; the local filter 3 is responsible for the photoelectric image sensor acquisition. The information is combined with the information collected by the gyroscope. On the one hand, the main filter performs information synthesis and distribution on each local filter, and on the other hand, the estimated value of the system state error is fed back to each local filter to correct its accumulated error.

In Figure 4, the fusion reset structure is used, so there are:

β ₁ =β ₂ =β ₃ =1/3

Then, in the joint filtering structure, the overall system information is distributed among the filters according to the following rules:

The process noise variance of each filter model is also assigned according to the same rules:

X _i =X _m =X _g

Where β _i represents the information distribution coefficient of the i-th filter;

When the solutions of the local filters and the main filters are statistically independent, they can be optimally synthesized according to the following algorithm:

For the present implementation, the positioning information of the robot is the position information e of the robot in the longitude direction, the speed information v _e of the robot in the longitude direction, the acceleration information a _e of the robot in the longitude direction, and the robot in the latitude direction. The position information n, the speed information v _n of the robot in the latitude direction, the acceleration information a _{n of the} robot in the latitude direction, the attitude information θ of the robot, and/or the rotational speed information ω of the robot.

Preferably, the method further comprises:

Applying the triangle similarity principle to the position information collected by the position sensitive sensor to obtain the distance between the front of the robot and the obstacle, and simultaneously acquiring the moving speed of the robot;

The obstacle avoidance strategy is obtained by using the distance between the front of the robot and the obstacle and the moving speed of the robot.

FIG. 5 is a schematic diagram of distance measurement of the position sensitive sensor of the present invention. Use the triangle principle to measure the distance between the front of the robot and the obstacle. Two equal focal length converging lenses 13 are mounted in front of the robot, and two position sensitive sensors 14 are respectively located at the focal points of the two converging lenses 13, which are in line with the light source 12. X1, x2 are the positions of the photosensitive region of the position sensitive sensor 14 after the obstacle 11 reflects the light passing through the condenser lens; l is the distance between the two position sensitive sensors 14; y is the distance from the obstacle to the front of the robot.

According to the similar triangle relationship in the figure:

Formula (1) + (2)

Because l ₁ +l ₂ =l, so

The relative speed is also available:

In the above formula, y _n is the distance between the vehicle and the obstacle at time n, y _n-1 is the distance between the vehicle and the obstacle at time n-1, and t _n -t _n-1 is the time difference between the two distances.

For the present embodiment, the robot moves to the target position using the positioning information of the robot and the obstacle avoidance strategy. The technical solution realizes accurate autonomous navigation of the robot indoors, and efficiently completes the cleaning task. In general, home cleaning robots now have no ability to move to all corners of the house without repeating them. It is very likely that they will move back and forth to the same area over and over again. The technical solution enables the robot to be accurately positioned, so that the robot can avoid repeated paths, thereby saving energy and saving cleaning time.

The specific embodiments of the present invention have been described in detail with reference to the preferred embodiments of the present invention. All modifications, equivalent substitutions, improvements, etc., made within the spirit and scope of the invention are intended to be included within the scope of the invention.

Claims

A method for positioning a robot, characterized in that the method comprises:

When the robot moves, the corresponding orientation data is collected by the photoelectric image sensor, the position sensitive sensor, the odometer and the gyroscope;

The orientation data collected by the photoelectric image sensor is input to the first local filter, the second local filter, the third local filter, and the main filter; the orientation data collected by the position sensitive sensor is input to the first local filter. The first local filter processes the orientation data collected by the photoelectric image sensor and the orientation data collected by the position sensitive sensor according to the latest information fed back by the main filter, and inputs the processing result to the main filter; the odometer collects The orientation data is input to the second local filter; the second local filter processes the orientation data collected by the photoelectric image sensor and the orientation data collected by the odometer according to the latest positioning information fed back by the main filter, and inputs the processing result to the main a filter; the orientation data acquired by the gyroscope is input to a third local filter; and the third local filter performs the orientation data collected by the photoelectric image sensor and the orientation data collected by the gyroscope according to the latest information fed back by the main filter. Processing, inputting the processing result to the main filter;

The main filter fuses data output by the photoelectric image sensor, data output by the first local filter, data output by the second local filter, and data output by the third local filter to obtain positioning information of the robot; Information is fed back to the first local filter, the second local filter, and the third local filter.
The method according to claim 1, wherein the positioning information of the robot is position information e of the robot in the longitude direction, speed information v e of the robot in the longitude direction, and acceleration information of the robot in the longitude direction. e , position information n of the robot in the latitude direction, speed information v n of the robot in the latitude direction, acceleration information a n of the robot in the latitude direction, attitude information θ of the robot, and/or rotational speed information ω of the robot.
The method of claim 1 or 2, further comprising:

Applying the triangle similarity principle to the position information collected by the position sensitive sensor to obtain the distance between the front of the robot and the obstacle, and simultaneously acquiring the moving speed of the robot;

The obstacle avoidance strategy is obtained by using the distance between the front of the robot and the obstacle and the moving speed of the robot.
The method of claim 3, further comprising:

The robot moves to the target position using the positioning information of the robot and the obstacle avoidance strategy.
The method according to claim 1 or 2, wherein the photoelectric image sensor is disposed under the robot body.
A method according to claim 1 or 2, wherein two position sensitive sensors are provided on the body of the robot.