CN116088512A

CN116088512A - Unmanned construction vehicle technology capable of being accurately positioned

Info

Publication number: CN116088512A
Application number: CN202211718733.XA
Authority: CN
Inventors: 田野; 刘培培; 胡超玄
Original assignee: Chengdu Univeristy of Technology
Current assignee: Chengdu Univeristy of Technology
Priority date: 2022-12-30
Filing date: 2022-12-30
Publication date: 2023-05-09

Abstract

The invention discloses an unmanned construction vehicle technology, and particularly relates to an unmanned construction vehicle technology capable of being accurately positioned. After fully researching the existing unmanned construction vehicle and related positioning technology, an unmanned construction vehicle platform capable of accurately positioning is designed, a laser radar automatic lifting device and a laser radar balancing device are mounted, and a main controller, an IMU, a laser radar and a monocular camera are selected on a hardware platform. The laser radar automatic lifting device, the laser radar balancing device and the enhanced repositioning-based Cartograper algorithm are designed and verified: the relative error rate is reduced by 1.04%, the omission factor is reduced to 0%, and the obstacle avoidance rate is improved by 93.33%.

Description

Unmanned construction vehicle technology capable of being accurately positioned

Technical Field

The invention relates to unmanned construction vehicle technology, in particular to unmanned construction vehicle technology capable of being positioned accurately.

Background

At present, as the scale of the construction industry is continuously enlarged, the requirements of construction vehicles are increasingly larger, the construction sites are also increasingly complex, and a plurality of security problems derived from the requirements are increasingly obvious, wherein the positioning technology of unmanned construction vehicles is particularly important. When a construction vehicle works in a complex scene, a driver is easy to feel tired, the surrounding environment needs to be noticed at any time, and safety problems easily occur when the driver is slightly not noticed. Therefore, the method has important significance in solving the problem of personnel driving safety, reducing labor consumption and researching unmanned vehicles.

Disclosure of Invention

In order to solve the problem of unmanned positioning of a construction vehicle, the invention provides a positioning technology of an unmanned construction vehicle based on laser SLAM, which can realize the accurate positioning of the unmanned vehicle by constructing a high-precision map and improve the track precision and safety of the movement of the unmanned vehicle.

The technical scheme adopted by the invention for solving the technical problems is as follows: the system comprises a lifting device, a 2D laser radar, a balancing device and the like, wherein an upper computer is a Jetson Nano development board, a sensor module is a 2D laser radar, a monocular camera and an IMU (inertial measurement unit), a personal computer remote real-time monitoring module with a Linux system is mounted, and a lower computer is an STM32 control module and an Arduino control module, wherein the STM32 control module controls a chassis motor to move, and the Arduino control module controls the laser radar to automatically lift.

In order to realize the positioning and obstacle avoidance requirements of the unmanned vehicle, the software system is built on the ROS platform, and the hardware platform is designed in a layered manner aiming at each functional module of the positioning and obstacle avoidance system of the unmanned vehicle platform, wherein the functional modules comprise a command layer, an algorithm layer, a control layer and a hardware layer.

The instruction layer is to install Ubuntu system between PC and Jetson Nano host, connect 5G signal, and then configure relevant IP parameters in basherc file, so that PC and Jetson Nano host are under the same local area network, then Jetson Nano host can be controlled remotely by ssh network protocol. The Rviz visualization tool can be opened in the ROS to remotely monitor the construction state and the construction environment of the construction vehicle in real time, and monitor the position of the construction vehicle in the construction environment in real time.

The algorithm layer is mainly composed of a Jetson Nano host and a 2D laser radar, and is mainly used for receiving data of the laser radar and inertial navigation data acquired by an STM32, and local path planning of a construction vehicle platform is achieved through SLAM and obstacle avoidance algorithm. The SLAM positioning algorithm of the construction vehicle uses a Cartographer algorithm, adopts a graph optimization method, and accurately estimates the moving track and map of the robot according to the stored sensor data and the constraint relation between the sensor data and the robot.

The control layer is meant by STM32 control panel, arduino control panel, inertial navigation sensor, encoder and constitutes, and the STM32 control panel is used for receiving inertial navigation sensor data, and Arduino control panel is used for controlling laser radar automatic lifting device, then carries out serial port communication with STM32 control panel and Arduino control panel and host computer Jetson Nano host computer, and the command layer sends control instruction, and the control layer responds in real time, makes construction vehicle make corresponding change.

The hardware layer is composed of a mowing robot body, a 2D laser radar, an inertial navigation sensor, a monocular camera and the like.

Drawings

FIG. 1 is an overall architecture of an unmanned vehicle;

FIG. 2 is a SLAM system frame of an unmanned vehicle;

FIG. 3 is a control system hardware configuration;

FIG. 4 is an overall framework of the inventive Cartograph algorithm.

Detailed Description

In fig. 1, a Jetson Nano development board is used as a main control module of the system and is responsible for acquiring data of a sensor, wherein the Jetson Nano development board comprises a 2D laser radar, an IMU inertial measurement module and a monocular camera and is connected with 2 control modules, one is a control board based on Arduino and used for controlling an automatic lifting device of the laser radar, and the other is a control board based on STM32 and used for controlling a motor of a construction vehicle.

In fig. 2, the unmanned construction vehicle hardware platform is designed in a layered manner, and mainly comprises a hardware layer, a control layer, an algorithm layer and an instruction layer.

In fig. 3, the control system hardware of the unmanned construction vehicle is formed, when the single-line laser radar of the unmanned construction vehicle scans the obstacle ahead, collected information is transmitted to the jetson nano mini host computer to carry out algorithm judgment, so that whether the laser radar lifting device is changed or not is decided, then decision information is transmitted to Arduino, the Arduino controls a motor to drive a screw rod sliding table, and the single-line laser radar is driven to vertically move, so that an accurate grid map is built. The system mainly comprises a power supply module, a 2D laser radar acquisition module, a Jetson Nano mini host, an Arduino control module and the like.

In fig. 4, an improved Cartographer algorithm framework is provided, so that the problems of lower repositioning accuracy and longer loop detection time are solved, and the traditional Cartographer algorithm is enhanced and improved. The method is improved on the original Cartographer algorithm framework: firstly, loading a dictionary, and extracting features: the method for extracting the characteristic points comprises the steps of firstly extracting the characteristic points in the image in a blocking mode, secondly distributing the characteristic points by using a quadtree, and finally extracting the corresponding largest characteristic points in each block, so that the characteristic points with uniform distribution are obtained. The feature points are then described, i.e., the descriptors are computed, the dictionary is queried to compute the bow vector for the frame, and the bow vector and the image index are transmitted to the back end. And updating the database, namely binding the image index and the nodeid at the current moment together in the back end to be recorded as a node __ to_image, and simultaneously updating the image index and the nodeid into a word-image index database and recording the image index into a KeyFrameDataBase, wherein the database stores the index of the image where each word is located. And (3) inquiring a closed loop, namely searching historical images of the same words in a database according to the words of the images obtained at the current moment. After the image is found, the index of the image is corresponding to the node index, and finally the current laser data is matched with the map at the position of the node by using the current laser data. And (3) saving the database, namely building a map, ending the saving of the map, and saving the database in a hard disk.

Claims

1. An unmanned construction vehicle technique that can pinpoint, its characterized in that: the unmanned construction vehicle overall architecture includes the host computer: a main control system module; a sensor module: 2D lidar, monocular camera, inertial measurement unit; the lower computer: a motion control module; the unmanned construction vehicle platform layering design score is: an instruction layer, an algorithm layer, a control layer and a hardware layer; unmanned construction vehicle SLAM algorithm: a Cartgrader algorithm to enhance positioning.

2. The overall architecture of an unmanned construction vehicle technology capable of being precisely positioned according to claim 1, wherein: the construction vehicle is innovatively designed as a whole, and the construction vehicle is designed from an upper computer, a sensor module and a lower computer; the system comprises a lifting device, a 2D laser radar, a balancing device and the like, wherein an upper computer is a Jetson Nano development board, a sensor module is a 2D laser radar, a monocular camera and an IMU (inertial measurement unit), a personal computer remote real-time monitoring module with a Linux system is mounted, and a lower computer is an STM32 control module and an Arduino control module, wherein the STM32 control module controls a chassis motor to move, and the Arduino control module controls the laser radar to automatically lift.

3. The platform layering design of an unmanned construction vehicle technology capable of being precisely positioned according to claim 1, wherein the platform layering design is characterized in that: the hardware platform is designed in a layered manner aiming at each functional module of the positioning and obstacle avoidance system of the unmanned vehicle platform, and comprises an instruction layer, an algorithm layer, a control layer and a hardware layer; the instruction layer is used for installing a Ubuntu system on the PC and the Jetson Nano host, so that the PC and the Jetson Nano host are under the same local area network; the construction vehicle construction method comprises the steps that an Rviz visualization tool can be opened in an ROS to remotely monitor the construction state and the construction environment of a construction vehicle in real time, and the position of the construction vehicle in the construction environment is monitored in real time; the algorithm layer is mainly composed of a Jetson Nano host and a 2D laser radar, and is mainly used for receiving data of the laser radar and inertial navigation data acquired by an STM32, and local path planning of a construction vehicle platform is realized through SLAM and obstacle avoidance algorithm; the control layer is composed of an STM32 control board, an Arduino control board, an inertial navigation sensor and an encoder, wherein the STM32 control board is used for receiving data of the inertial navigation sensor, the Arduino control board is used for controlling an automatic laser radar lifting device, then the STM32 control board and the Arduino control board are communicated with a Jetson Nano host of an upper computer through serial ports, the command layer sends a control command, and the control layer responds in real time to enable a construction vehicle to change correspondingly; the hardware layer is composed of a mowing robot body, a 2D laser radar, an inertial navigation sensor, a monocular camera and the like.

4. The positioning technology of the unmanned construction vehicle technology capable of being accurately positioned according to claim 1, wherein: the improved Cartographer algorithm framework is used for enhancing and improving the traditional Cartographer algorithm to solve the problems of lower repositioning accuracy and longer loop detection time consumption; the method is improved on the original Cartographer algorithm framework: firstly, loading a dictionary, and extracting features: the method for extracting the characteristic points comprises the steps of firstly extracting the characteristic points in the image in a blocking mode, secondly distributing the characteristic points by using a quadtree, and finally extracting the corresponding largest characteristic points in each block, so that the characteristic points with uniform distribution are obtained; then describing the feature points, namely calculating descriptors, inquiring a dictionary to calculate bow vectors of the frames, and transmitting bow vectors and image index to the rear end; in the back end, the image index is bound with the nodeid at the current moment and recorded as a node __ to_image, and is updated into a word-image index database and recorded as a KeyFrameDataBase, and at the moment, the index of the image where each word is located is stored in the database; inquiring a closed loop, namely searching historical images of the same word in a database according to the word of the image obtained at the current moment; after the image is found, the index of the image is corresponding to the node index, and finally the current laser data is matched with the map at the position of the node by applying the current laser data at the position of the node; and (3) saving the database, namely building a map, ending the saving of the map, and saving the database in a hard disk.