AGV robot: Theoretical basis of visual obstacle avoidance

Introduction to AGV (Automated Guided Vehicle):

The American Supply Chain Management Association defines AGV as an automatic guided vehicle, which is equipped with an electromagnetic or other automatic guidance device and can travel on a specified navigation path. It is a transport vehicle with safety protection and various transfer functions.

With the improvement of automation, the production mode of traditional manufacturing industry has undergone profound changes. The automation of warehousing and logistics has become an inevitable trend. Production has transitioned from a single fixed mode to a flexible production line with strong adaptability. In order to save costs and improve economic benefits, advanced production methods such as flexible manufacturing systems and automated warehousing systems have been widely used. They are based on high-tech systems as decision-making centers, and automated stereoscopic warehouses and unmanned transport vehicles as main equipment. Among them, unmanned transport equipment is marked by automatic guided vehicles, which integrate advanced technologies such as computer science, image signal processing, and automatic control, and is a key equipment for modern logistics systems and flexible production organization systems.

At present, the working environment of AGV can be divided into indoor environment and outdoor environment. More systems work indoors or use indoors as the main workplace. Because indoors are universal, they can provide theoretical and technical support for the development of mobile robots for various occasions. The indoor environment is regarded as a structured environment, where the light is relatively stable and the complexity of the environment is limited. The signs on the structured roads often have obvious characteristics, such as color, width, boundaries, etc., so that a relatively simple method can be used for feature extraction during road recognition, and then the road scene can be restored.

AGV application areas include: manufacturing, warehousing, post offices, ports, airports, tobacco, medicine, food, chemicals, nuclear materials, photosensitive materials and other special industries.

AGV navigation method:

The navigation problem of mobile robots mainly involves three questions: "Where are you now?", "Where are you going?", and "How are you going?". The first question is the positioning problem in the navigation system, which determines the position of the mobile robot in the working environment relative to the global coordinates and its own posture; the second and third questions are the path planning and tracking of the navigation system.

The purpose of studying navigation is to enable machines to move purposefully and complete specific tasks without human intervention. Therefore, the flexibility of the logistics system depends on the guidance and navigation methods. The guidance and navigation methods used in systems in different application scenarios are also diverse.

There is a difference between guidance and navigation. Guidance is to calculate the operating parameters of the next cycle based on the current state data. It only requires relative position and has nothing to do with the global coordinates. Navigation refers to determining its own position and heading. The main guidance and navigation methods of AGV are:

(1) Coordinate guided AGV

The principle of direct coordinate guidance is: first, the driving area is divided into several standard unified coordinate small areas by positioning blocks, and then the number of small areas passed through is counted during driving to achieve guidance. There are two commonly used forms: magnetic type and electromagnetic type. The former is to divide the coordinate small areas by different colors, and then use color-sensitive to count; the latter is to divide the coordinates by magnetic blocks or metal blocks, and then use metal-sensitive devices to count.

The advantages of this guidance method are: simple path modification, good guidance reliability, and no influence from the environmental background. The disadvantages are: complex installation of positioning blocks, the guidance positioning is completely determined by the size and number of positioning blocks, large workload, and low precision.

(2) Electromagnetic guided AGV

Electromagnetic guidance is one of the most commonly used guidance methods. The wire buried underground carries an electromagnetic frequency. The frequency in the wire is turned on or off by a device called "ground". Electromagnetic guidance relies on the electromagnetic frequency generated by induction to guide the vehicle along the buried route. This guidance method is mature in technology, economical and reliable, with concealed leads that are not easily contaminated or damaged. The guidance principle is simple and easy to communicate, and it is not affected by sound or light. However, it has poor flexibility, high requirements for the flatness of the ground, and it is difficult to change or expand the path.

(3) Optically guided AGV

Optical guidance is based on the principle that the propagation process of a single light source will not change. A color ribbon with stable reflectivity is laid on the driving path. At the same time, a photoelectric device that can transmit and receive light sources is installed on the vehicle. The direction of the vehicle is adjusted by comparing the transmitted and received light in real time. Its advantages are low cost for laying the guide line and good flexibility, but it is very sensitive to the pollution and wear of the color ribbon, has high requirements for the environment, and has poor guidance reliability.

(4) Laser infrared navigation AGV

Laser infrared navigation is to equip the vehicle with a scanner that can emit and receive laser infrared rays, install a sufficient number of laser reflectors around the guide area, emit laser beams through the laser scanner, collect the laser beams reflected by the reflectors, and determine its current position and heading through trigonometric calculations to achieve guidance. The advantages of this navigation method are accurate positioning, flexible and changeable driving paths, and adaptability to a variety of on-site environments; the disadvantages are high manufacturing costs, complex position calculations, and limited error correction capabilities against light interference.

(5) Vision-guided AGV

Visual navigation, also known as image recognition guidance, uses a camera to dynamically capture path information, identifies the path to be tracked through image processing technology, and guides the operation. Visual navigation can not only obtain a large amount of information, but also has the characteristics of simple path setting and change, and good system flexibility. In addition, inertial navigation and GPS navigation are mostly used in the military.

The various guidance/navigation methods of AGV can be divided into external guidance and internal guidance according to the source of guidance information. According to whether the guidance has a predetermined path, it can be divided into two categories: predetermined path guidance and free path guidance.

Research on Navigation Marking Line Detection:

The visual detection of lane markings as the edge of the road is a basic function that AGV path recognition needs to achieve. The visual navigation AGV uses CCD cameras to collect strip markings laid on the ground, and uses image processing and analysis to obtain information about the surrounding environment of the guided vehicle. This method is the core technology in the AGV recognition system.

Contents involved: Coordinate system establishment, lane model analysis, image preprocessing, etc. The following focuses on image processing:

There is a basic consensus on automotive research: images will be contaminated during acquisition, conversion and transmission, which will inevitably reduce image quality. Therefore, the image should be preprocessed first, then threshold segmentation should be performed, and then the path should be identified and tracked.

In addition to the available information, the original image obtained by the onboard camera of the visual navigation AGV has a lot of redundant information and poor recognizability due to environmental restrictions and random interference. First, the original image must be preprocessed. The process is as follows:

Image smoothing is a low-pass filtering technique that can be performed in both the frequency domain and the spatial domain.

(1) Template operation.

The template operation implements a neighborhood operation, that is, the result of a certain pixel is not only related to the grayscale of the pixel, but also to the values ​​of its neighboring points. The mathematical description of the template operation is called convolution.

(2) Median filtering.

Median filtering is to sort the image pixels in the neighborhood by grayscale and take the middle value as the output pixel. It is a nonlinear spatial filtering technology and a filtering method that can remove noise while protecting the target boundary from becoming blurred. Its principle is to select a moving window with an odd number of pixels and replace the grayscale value of the central pixel of the window with the median of the grayscale in the window, thereby eliminating isolated noise points. Its mathematical expression is as follows:

Among them, f(x,y), g(x,y) are the original image and the processed image respectively, and φ is a two-dimensional template, usually a 2*2, 3*3 area.

(3) Morphological correction.

After the above processing, there may still be a small number of scattered points in the binary image, and the edge of the black part is not very clear, with burrs and holes. Perform mathematical morphological filtering on the binary image to achieve local background smoothing. The main operations of mathematical morphology are morphological dilation, morphological difference erosion, opening operation, and closing operation.