Geofencing and follow-me control method of autonomous golf-cart vehicles by use of open-source map data and ego-motion compensation

Abstract

The present invention generally relates to a driving control technology for implementing geofencing and following driving functions in an autonomous golf cart vehicle based on open source map data (e.g., OpenStreetMap) and ego motion compensation. In particular, the present invention relates to a geofencing and following driving control technology for an autonomous golf cart vehicle based on open source map data and ego motion compensation, configured to perform geofencing and following driving through high-precision real-time estimation of the vehicle state (position, attitude) and distortion-free surrounding perception by real-time interpretation of semantic attribute information inherent in the open source map data to generate geofencing data for the autonomous golf cart vehicle in real time, recognizing and setting a follower target in front of the vehicle using a front LiDAR sensor, and real-time correction of sensing distortion caused by ego motion through the fusion processing of multiple sensing data. According to the present invention, by combining the generation of semantic driving rules based on open source map data and the generation of vehicle control data based on the correction of ego motion distortion, there is an advantage of simultaneously achieving the conflicting goals of low cost and high reliability in the implementation of geofencing and following driving functions of an autonomous golf cart vehicle.

Description

Technology Field

[0001] The present invention relates to a driving control technology that implements geofencing and follow-me functions in an autonomous golf cart vehicle based on open source map data (e.g., OpenStreetMap) and vehicle motion compensation.

[0002] In particular, the present invention relates to open source map data and geofencing and following driving control technology for an autonomous golf cart vehicle based on ego-motion compensation, configured to generate geofencing data for an autonomous golf cart vehicle in real time by interpreting semantic attribute information inherent in open source map data in real time, recognize and set a follower target in front of the vehicle using a front lidar sensor, and perform geofencing and following driving through high-precision real-time estimation of the vehicle state (position, attitude) and distortion-free surrounding perception by correcting sensing distortion caused by ego-motion in real time through fusion processing of multiple sensing data. Background Technology

[0004] Recently, there has been growing interest in autonomous driving (self-driving) technology. Autonomous driving can be applied to general-purpose vehicles (e.g., passenger cars, taxis, trucks, buses, etc.) or special-purpose vehicles (e.g., golf carts, AGVs, agricultural machinery, etc.). When applied to general-purpose vehicles, it is crucial to perceive the surrounding environment more accurately than a human driver and drive safely according to the situation. When applied to special-purpose vehicles, it is important to satisfy the specific requirements of the relevant site.

[0005] In this context, geofencing is a critical requirement for specialized vehicles. Geofencing is a technology that establishes virtual boundaries within a specific geographical area and enables vehicles to perform specific actions when entering or exiting those boundaries. Geofencing technology is becoming increasingly important in various fields for safe vehicle operation, access control to specific areas, and increased operational efficiency.

[0006] Existing geofencing-based vehicle control systems require high-precision GPS, high-definition maps (HD Maps) containing detailed terrain information, expensive composite sensors, and high-performance onboard computers capable of processing complex calculations in real time. Since specialized vehicles are cost-sensitive, it is difficult to apply high-specification systems such as expensive sensors or high-performance processors. Furthermore, there are issues such as the difficulty in securing high-definition map data for areas like golf courses, industrial complexes, and farmlands, as well as excessive maintenance costs. High-definition maps play a crucial role in autonomous driving by precisely capturing static road elements such as lanes, curbs, and signs. However, not only are the construction costs of high-definition maps enormous—reaching thousands of dollars per kilometer—but the update costs to keep them up-to-date are also massive. Applying high-definition map data to the autonomous driving of specialized vehicles presents significant challenges.

[0007] Prior art in this field, U.S. Published Patent No. 2013 / 0332007, discloses a technology in which a mobile device outside the vehicle detects geofence intersections to control vehicle functions (e.g., door locking). However, this technology is unsuitable for functions requiring immediate and stable responses, such as vehicle deceleration, because the reliability of control can be degraded by external factors such as communication delays or interruptions. Another prior art, U.S. Registered Patent No. 9,688,288, discloses a technology that utilizes high-specification sensors and a vast geographic information database to request a switch in driving mode when a semi-autonomous vehicle approaches a no-autonomous driving zone. This technology is limited to high-level driving mode switching and has not presented a technology for precisely performing vehicle control under the constraints of a special vehicle.

[0008] Therefore, technology is required to implement autonomous driving reliably and safely in a constraint environment of low-spec hardware and low-quality map data. Prior art literature

[0010] Republic of Korea Registered Patent No. 10-2848677 "Automatic Map Generation System for Autonomous Driving Robots" Republic of Korea Registered Patent No. 10-2548866 "Method, Server, and Computer Program for Detecting Risk Occurrence Using 3D Geofences" Republic of Korea Registered Patent No. 10-2506396 "System for Providing Equipment and Material Sales Service Combining Smart Logistics and 24-Hour Unmanned Stores" Republic of Korea Registered Patent No. 10-2719983 "Device and Method for Designating Smart Work Areas and Paths Based on Multi-Measurement Terrain Data in a Communication System" Republic of Korea Published Patent No. 10-2025-0011353 "System and Method for Verifying Reliability of Edge Infrastructure for Supporting Autonomous Driving Operation" US Registered Patent US 9,688,288 B "Geofencing for auto drive route planning" US Published Patent US 2013 / 0332007 A "Accessory control with geo-fencing" Japanese Published Patent JP 2023-55695 A The problem to be solved

[0011] The objective of the present invention is to provide a driving control technology that implements geofencing and following driving functions in an autonomous golf cart vehicle based on open source map data (e.g., OpenStreetMap) and vehicle motion compensation.

[0012] In particular, the objective of the present invention is to provide open source map data and geofencing and following driving control technology for an autonomous golf cart vehicle based on ego-motion compensation, configured to generate geofencing data for an autonomous golf cart vehicle in real time by interpreting semantic attribute information inherent in open source map data in real time, recognize and set a follower target in front of the vehicle using a front lidar sensor, and perform geofencing and following driving through high-precision real-time estimation of the vehicle state (position, attitude) and distortion-free surrounding perception by correcting sensing distortion caused by ego-motion in real time through the fusion processing of multiple sensing data.

[0013] The problems solved by the present invention are not limited to this, and other problems can be understood from the description in this specification. means of solving the problem

[0015] To achieve the above objective, the present invention provides a geofencing and following driving control method for an autonomous driving golf cart vehicle based on open source map data and vehicle motion compensation, which is performed by an autonomous driving vehicle control device installed in an autonomous driving golf cart vehicle equipped with an autonomous driving function.

[0016] A geofencing and following driving control method for an autonomous golf cart vehicle based on open source map data and ego-vehicle motion compensation according to the present invention comprises: a first step of generating geofence data by analyzing semantic attribute information included in open source map data and combining geometric feature data for a golf course with vehicle control rules; a second step of recognizing and setting a follower target in front of the golf cart vehicle using a front LiDAR sensor; a third step of collecting multi-sensing data including GNSS data, LiDAR data, and IMU data related to the golf cart vehicle through a multi-sensor suite; a fourth step of correcting ego-vehicle motion distortion through fusion processing of the multi-sensing data and obtaining real-time vehicle status information regarding the position, speed, and direction of the golf cart vehicle; and a fifth step of controlling the operation of the golf cart vehicle according to vehicle control rules when identifying the golf cart vehicle's approach to a geofence boundary based on the geofence data based on the real-time vehicle status information of the golf cart vehicle. and a sixth step of controlling the operation of the golf cart vehicle to follow a target in front of the golf cart vehicle while maintaining a certain distance by recognizing the target using a front lidar sensor; may be configured to include.

[0017] In the present invention, the first step may be configured to include: a step of filtering and extracting geometric feature data having golf course-related attribute tags from the open source map data; a step of converting the filtered and extracted geometric feature data into a pre-set geographic data structure to create a local map database; a step of mapping and setting vehicle control rules according to the golf course environment for the geometric feature data based on attribute tag information stored in the local map database; and a step of generating geofence data by a combination of the geometric feature data and the vehicle control rules.

[0018] In the present invention, the fourth step may comprise: identifying a sequence of lidar points provided by a lidar sensor during a scanning period while the golf cart vehicle is driving; calculating sequential minute attitude changes of the golf cart vehicle using high-frequency motion information of IMU data for each lidar point of the lidar point sequence; calculating attitude information of the golf cart vehicle at a series of time points in which each lidar point is captured by linear interpolation or spline interpolation of the minute attitude changes; generating a distortion-free point cloud in which distortion caused by the movement of the golf cart vehicle is corrected by converting each lidar point to a common coordinate system based on the series of attitude information of the golf cart vehicle; and estimating real-time vehicle state information regarding one or more of the position, speed, and direction of the golf cart vehicle by tightly coupling and fusing the distortion-free point cloud, the IMU data, and the GNSS data using a Kalman filter.

[0019] In the present invention, the fifth step may be configured to include: a step of calculating a driving prediction path of a golf cart vehicle using real-time vehicle status information of the golf cart vehicle and a preset vehicle kinematics model; a step of determining whether there is a potential intersection between the driving prediction path and a geofence boundary according to geometric feature data stored in the geofence data; a step of maintaining the current vehicle driving state if no potential intersection is identified as a result of the determination; and a step of generating and transmitting a vehicle driving command based on the current speed of the golf cart vehicle and the remaining distance to the geofence boundary to correspond to a vehicle control rule corresponding to the geometric feature data if a potential intersection is identified as a result of the determination.

[0020] In the present invention, the fifth step may be configured to include: obtaining a vehicle control rule from geofence data by referring to real-time vehicle status information of a golf cart vehicle; identifying a driving restriction environment around the golf cart vehicle in real time while driving using the distortion-free point cloud; generating a short-term target path in real time by combining the vehicle control rule and the driving restriction environment identified in real time; calculating a vehicle driving control command in real time to achieve the short-term target path; and transmitting the vehicle driving control command to a vehicle driving module.

[0021] In the present invention, the sixth step may be configured to include: a step of performing object recognition processing on the distortion-free point cloud to identify clusters in front of the vehicle and specifying a follower among them; a step of calculating a target speed of the golf cart vehicle and driving control based on a distance error with the follower; a step of calculating a target steering angle of the golf cart vehicle and driving control based on an angle error with the follower; a step of braking the golf cart vehicle when the follower approaches the golf cart vehicle within a preset threshold; a step of braking the golf cart vehicle when the follower deviates from the golf cart vehicle by more than a preset threshold angle; and a step of braking the golf cart vehicle when the follower is lost.

[0022] Meanwhile, the computer program according to the present invention is stored in a computer-readable non-volatile storage medium to execute the geofencing and following driving control method of an autonomous driving golf cart vehicle based on open source map data and vehicle motion compensation as described above. Effects of the invention

[0024] According to the present invention, by combining the generation of semantic driving rules based on open source map data and the generation of vehicle control data based on the correction of vehicle motion distortion, there is an advantage in simultaneously achieving the conflicting goals of low cost and high reliability in the implementation of geofencing and following driving functions of an autonomous golf cart vehicle.

[0025] According to the present invention, since a low-computation kinematics model is used for vehicle position prediction and only simple physical equations are used for deceleration control, stable real-time control is possible even on low-spec hardware. Furthermore, since deceleration commands are calculated by utilizing real-time collected vehicle speed and distance information to a target point, there is an advantage of being able to operate effectively even with low-quality open-source map data. Brief explanation of the drawing

[0027] [Fig. 1] is a block diagram showing the overall configuration of an autonomous driving vehicle control device according to one embodiment of the present invention. [Fig. 2] is a software configuration diagram of an autonomous driving vehicle control device according to an embodiment of the present invention. [Fig. 3] is a flowchart of a geofencing and following driving control method for an autonomous driving golf cart vehicle based on open source map data and vehicle motion compensation according to an embodiment of the present invention. [Fig. 4] is a flowchart of the process of generating geofence data using open source map data in the present invention. [Fig. 5] is an example diagram of mapping vehicle control rules by semantically interpreting attribute tags in the present invention. [Fig. 6] is a flowchart of the process for correcting magnetic motion distortion by multi-sensor fusion in the present invention. [Fig. 7] is a flowchart of a first embodiment of the geofencing control process for a golf cart vehicle in the present invention. [Fig. 8] is a flowchart of a second embodiment of the geofencing control process for a golf cart vehicle in the present invention. [Fig. 9] is a flowchart of the tracking driving control process for a golf cart vehicle in the present invention. Specific details for implementing the invention

[0028] The present invention will be described in detail below with reference to the drawings.

[0029] In describing the present invention, detailed explanations of parts that overlap with the prior art may be omitted.

[0030] [Fig. 1] is a block diagram showing the overall configuration of an autonomous driving vehicle control device according to one embodiment of the present invention.

[0031] The autonomous vehicle control device (100) is a device that controls the driving of an autonomous golf cart vehicle based on open source map data and autonomous vehicle motion compensation, and is configured to include a data collection unit (110), an autonomous driving processing unit (120), a vehicle drive control unit (130), and a user interface unit (140). In terms of hardware, the autonomous vehicle control device (100) is implemented through a Hardware Management Control Unit (HMCU) and a multiple sensor suite installed in the autonomous golf cart vehicle, and the HMCU is composed of a processor (e.g., ARM-based SoC, FPGA, etc.), a storage medium (non-transitory memory) (e.g., flash memory), an input / output interface, etc.

[0032] The data collection unit (110) is a component that collects various data (GNSS data, LiDAR data, IMU data) related to the real-time location, direction, speed, and surrounding conditions of the golf cart vehicle through a multi-sensor suite installed on the autonomous golf cart vehicle. In the present invention, the data collection unit (110) is equipped with a GNSS receiver (111), a LiDAR sensor unit (112), and an IMU sensor unit (113).

[0033] The GNSS receiver (111) is a component that receives signals from multiple satellite systems, such as GPS, GLONASS, and Galileo, to obtain global position information of the golf cart vehicle. It is also called a Global Navigation Satellite System (GNSS) receiver and can be implemented as a cost-effective GPS dual-band receiver or a high-precision Real-Time Kinematic (RTK) receiver. For example, the Waveshare LC29H(AA) GPS module provides an accuracy of approximately 1.5 meters (CEP) and an update rate of up to 10 Hz when positioned alone.

[0034] The LiDAR sensor unit (112) is a component that acquires geometric information about the surroundings of a vehicle by generating more than hundreds of thousands of three-dimensional point cloud data per second with a 360-degree horizontal field of view through a LiDAR (Light Detection and Ranging) sensor. In the present invention, a LiDAR sensor is utilized for a follow-up driving function and for detecting obstacles ahead. To this end, a front LiDAR sensor generates point cloud data to enable the identification of the relative position of objects ahead. For example, the SLAMTEC Rplidar C1 module provides a maximum detection distance of 6 meters and a scanning speed of 8 Hz.

[0035] The IMU sensor unit (113) is a component that provides information on changes in vehicle dynamics by measuring the linear acceleration and angular velocity of the vehicle through an inertial measurement unit (IMU) composed of a 3-axis accelerometer and a 3-axis gyroscope. Since the IMU sensor unit (113) provides an update rate of 100 Hz or higher, the frequency of information update is higher compared to the GNSS receiver unit (111) (e.g., 10 Hz) or the LiDAR sensor unit (112) (e.g., 8 Hz).

[0036] The autonomous driving processing unit (120) is a component that provides the autonomous driving function of a golf cart vehicle. In the present invention, the autonomous driving processing unit (120) is configured to include a map data storage unit (121), a destination driving processing unit (122), a geofencing processing unit (123), and a follow driving processing unit (124).

[0037] The map data storage unit (121) is a component that stores open-source map data on a storage medium (e.g., flash memory). Open-source map data refers to open map data that is crowdsourcing-based and created by volunteers worldwide, and can be freely used, edited, and distributed; for example, OpenStreetMap (OSM) data can be cited. With OSM data, map data of a specific area can be downloaded using a web-based Overpass Turbo tool. Unlike commercial map services, open-source map data has the advantage of not incurring license fees for commercial use; however, due to the nature of the data being built by volunteers, it does not guarantee centimeter-level geometric accuracy, which makes it difficult to utilize in fields requiring precise control (e.g., autonomous driving). In order to implement autonomous driving with a golf cart vehicle as in the present invention, specific environmental features such as cart-only roads, grass around the green, and water hazards must be personality Ina attribute Accordingly, it must be possible to perform complex controls such as deceleration, stopping, entry prohibition, and steering restriction, but open source map data does not provide the geometric accuracy for this. The present invention provides technical means to overcome the limitations of such open source map data.

[0038] The destination driving processing unit (122) is a component that controls the golf cart vehicle to drive itself to a location set as a destination. In a golf course, a travel path is predetermined along 18 holes, and within each hole, a travel path is also defined from the tee shot point to the green point. The autonomous driving golf cart vehicle must drive itself to a destination determined at that time based on this travel path, and in the present invention, autonomous driving is performed along the travel path between the current location and the destination by referring to open source map data.

[0039] The geofencing processing unit (123) is a component that controls the golf cart vehicle to automatically stop when the golf cart vehicle approaches a non-operational area of ​​the golf course defined in open source map data (e.g., hazard area, OB (Out of Bounds) area, green area, bunker area, repair area, etc.). This is a safety function for an autonomous golf cart vehicle. Additionally, it is a safety function that prevents accidents from occurring during the autonomous driving process of the golf cart vehicle by dividing various areas on the golf course and setting vehicle control rules for each area. At this time, the geometric feature information defining the geographical area can be defined as vertex coordinates in the form of a polygon. The geofencing processing unit (123) compares the vehicle location with the map data and, if it corresponds to a non-operational area, immediately decelerates and stops, and if vehicle control rules are set, controls driving accordingly.

[0040] The following driving processing unit (124) is a component that controls the golf cart vehicle to safely follow a target (e.g., caddy, golfer, etc.) in front of the golf cart vehicle by recognizing the target using the front LiDAR sensor unit (112) when the target is set through user menu operation, while maintaining a certain distance. It identifies the target by clustering LiDAR data and controls the driving of the golf cart vehicle by calculating the distance and angle with the identified target.

[0041] At this time, the data generated by the LiDAR sensor unit (112) while the golf cart vehicle is moving is severely distorted, making it difficult to use as is for tracking driving control. Accordingly, the tracking driving processing unit (124) fuses the data generated by the IMU sensor unit (113) and the data acquired by the GNSS receiver unit (111) with the LiDAR data to correct the LiDAR data distortion caused by the vehicle's motion in real time. In one embodiment, the data distortion can be corrected in real time by using an Extended Kalman Filter (EKF) or an Unscented Kalman Filter (UKF) to fuse the IMU data, GNSS data, and LiDAR data in a tightly-coupled manner.

[0042] The vehicle drive control unit (130) is a component that generates drive control commands for a golf cart vehicle according to vehicle control policies (e.g., driving, deceleration, stopping, etc.) generated by the target driving processing unit (122), geofencing processing unit (123), and follow driving processing unit (124), and transmits these commands to the vehicle drive system. Examples of such drive control commands include setting a target velocity or setting a steering angle. In one embodiment, the drive control commands are transmitted to the vehicle drive system (steering, braking, acceleration devices) via a standard communication bus such as a Controller Area Network (CAN) bus. The vehicle's braking device or motor controller receives the CAN message and executes acceleration, deceleration, and steering of the vehicle.

[0043] The user interface section (140) is a component that provides an interface with the user of the golf cart vehicle (e.g., caddy, golf player, etc.). For example, it may provide an input interface to set the vehicle driving mode (e.g., normal driving, autonomous driving, following driving) through the operation of an external physical button. In addition, it may provide an output interface that indicates the driving status of the golf cart vehicle by changing the color of an external indicator (lamp) for each driving mode.

[0045] [Fig. 2] is a software configuration diagram of an autonomous driving vehicle control device according to one embodiment of the present invention.

[0046] The software architecture of the autonomous vehicle control device (100) consists of a device driver (e.g., GNSS driver, LiDAR driver, IMU driver, CAN driver, etc.) installed on an operating system (e.g., Debian 11), and Docker running on top of it. Docker is a container-based virtualization platform and is a technology that allows applications and everything necessary for their execution (code, runtime, system tools, libraries, etc.) to be packaged into an isolated environment called a container, enabling easy distribution, expansion, and management. Referring to [Fig. 2], Docker may include ROS2, an open source software library for building robot applications, a sensor node manager, a follow-up driving module, a destination driving module, a geofencing module, a map data manager, an external integration module, and a command and control module. At this time, the command and control module may be implemented as a message broker (e.g., Mosquitto) based on the MQTT (Message Queuing Telemetry Transport) protocol.

[0048] [Fig. 3] is a flowchart of a geofencing and following driving control method for an autonomous driving golf cart vehicle based on open source map data and vehicle motion compensation according to an embodiment of the present invention.

[0049] Step (S110): An autonomous vehicle control device (100) generates geofence data by analyzing semantic attribute information included in open source map data (e.g., OSM) for a geofence function and combining geometric feature data for a golf course with vehicle control rules. The geofence data includes information on non-driving areas of the golf course (e.g., hazard area, OB area, green area, bunker area, repair area, etc.). In the present invention, geofence is a safety function that automatically stops a golf cart vehicle when it deviates from the normal area of ​​the golf course (e.g., driving path, fairway area, etc.) and approaches a non-driving area (e.g., hazard area, OB area, green area, bunker area, repair area, etc.). Additionally, it is a safety function that prevents accidents from occurring during the autonomous driving process of a golf cart vehicle by dividing several areas in the golf course and setting vehicle control rules for each area.

[0050] Open source map data contains a large amount of geometric feature information that defines geographical areas. Additionally, open source map data includes semantic attribute tags (e.g., highway, maxspeed, leisure, natural, landuse, building, golf, barrier, etc.) that define the nature of each geometric feature. To smoothly process large volumes of open source map data on low-spec processors, it is desirable to configure the system to generate geofence data by filtering only the geometric feature information with predefined semantic attribute tags within the geographical area of ​​the golf course, converting it into a lightweight GeoJSON format to create a local map database for the golf course, and then utilizing the local map database.

[0051] For geofencing, the autonomous vehicle control device (100) generates geofence data corresponding to the geometric structure of the geometric feature and generates a set of vehicle control rules (e.g., normal driving, deceleration driving, stop, no entry) based on semantic attribute tags. At this time, the area corresponding to the set of vehicle control rules for 'no entry' can be set as non-driving area information of the golf course.

[0052] The process of generating geofence data using open source map data in the present invention will be described later with reference to [Fig. 4].

[0053] Step (S120): The autonomous driving vehicle control device (100) preferably recognizes a person in the space in front of the golf cart vehicle through a front LiDAR sensor while the vehicle is stationary, and sets a follower target (e.g., caddy, golf player, etc.) in front of the golf cart vehicle through, for example, user menu operation. In the present invention, following driving refers to a convenience function that recognizes a follower target in front in real time using a front LiDAR sensor and safely follows them while maintaining a certain distance.

[0054] Step (S130): An autonomous vehicle control device (100) preferably collects multi-sensing data related to a golf cart vehicle through a multi-sensor suite installed in the golf cart vehicle. In the present invention, the multi-sensing data is configured to include GNSS data, LiDAR data, and IMU data.

[0055] Step (S140): The autonomous vehicle control device (100) corrects the vehicle motion distortion through the fusion processing of multiple sensing data (GNSS data, LiDAR data, IMU data) and obtains real-time vehicle status information regarding the position, speed, and direction of the golf cart vehicle. At this time, for the fusion processing of multiple sensing data, a tightly coupled fusion method based on an extended Kalman filter (EKF) or an unscented Kalman filter (UKF) may preferably be used.

[0056] When a golf cart is in motion, motion distortion occurs in the LiDAR data, making it difficult to accurately recognize the surrounding conditions of the vehicle. This problem becomes severe when the vehicle is turning. Golf carts often need to turn, especially when following other players, and there is a risk of colliding with other golf players in the vicinity because they cannot follow safely. Accordingly, the autonomous vehicle control device (100) needs to correct the motion distortion of the LiDAR data by utilizing IMU data.

[0057] In addition, to perform destination driving, geofencing, and follow driving for a golf cart, it is necessary to accurately recognize the vehicle's pose in real time while driving. Since IMU data provides information regarding the vehicle's linear acceleration and angular velocity, changes in pose can be accurately recognized, but recognizing the vehicle's pose itself is difficult. LiDAR data provides information about the vehicle's surrounding environment, so it can aid in estimating the pose, but accuracy decreases due to motion distortion during driving. Therefore, it is desirable to configure the system to determine the golf cart's pose information by interpolating LiDAR data using IMU data that is updated much more frequently (i.e., high frequency).

[0058] Then, based on the attitude information determined in this way, a LiDAR point cloud with the ego motion distortion of the golf cart vehicle removed is obtained, and from this, LiDAR odometry information, which is the geographical location information of the golf cart vehicle, can be acquired in real time. The geographical location information acquired in this way can be usefully utilized to correct GNSS data. If a high-precision RTK receiver is installed on the golf cart vehicle, there is no problem in using the GNSS data of the multi-sensing data as is. On the other hand, if a dual-band receiver is installed on the golf cart vehicle for cost efficiency, there is a possibility that errors are included in the GNSS data; therefore, it is desirable to correct the GNSS data of the multi-sensing data using a LiDAR odometry utilizing the distortion-compensated point cloud.

[0059] The process of correcting vehicle motion distortion through the fusion processing of multiple sensing data (GNSS data, LiDAR data, IMU data) in the present invention will be described later with reference to [Fig. 6].

[0060] Step (S150): When the autonomous vehicle control device (100) identifies that the golf cart vehicle is approaching a geofence boundary based on real-time vehicle status information (location information, speed information, direction information) for geofencing, it controls the operation of the golf cart vehicle according to the vehicle control rules for the area. In particular, when the golf cart vehicle is approaching a non-operation area, it controls the operation of the golf cart vehicle to stop automatically.

[0061] To this end, the autonomous vehicle control device (100) predicts the future driving path of the golf cart vehicle using real-time state data (location information, speed information, direction information) and the vehicle's kinematic model, and if a potential intersection with geometric feature data included in the geofence data is expected, it generates commands according to vehicle control rules using a physics-based equation of motion based on the golf cart vehicle's current speed and the remaining distance to the geofence boundary to control driving.

[0062] In the previous process, motion distortion of LiDAR sensor data caused by ego-motion was corrected in real-time at each point using sensor data from an Inertial Measurement Unit (IMU), thereby performing high-precision real-time estimation of the vehicle state (position, attitude) and distortion-free surrounding perception, so geofencing processing can be executed safely and precisely.

[0063] The process of controlling geofencing for a golf cart vehicle in the present invention will be described later with reference to [Fig. 7].

[0064] Step (S160): The autonomous driving vehicle control device (100) recognizes a target to be followed in front of the golf cart vehicle using a front LiDAR sensor and controls the operation of the golf cart vehicle to follow while maintaining a certain distance. In the preceding process, motion distortion of the LiDAR sensor data caused by ego-motion is corrected in real-time at each point using sensor data from an inertial measurement unit (IMU), thereby performing high-precision real-time estimation of the vehicle state (position, attitude) and distortion-free surrounding perception, so the following driving can be executed safely and precisely.

[0065] The technical feature of the present invention described above is that it organically combines and applies two independent algorithms to an autonomous driving golf cart used on a golf course: a driving rule determination configuration based on the semantic attributes of open-source map data and a high-precision vehicle state estimation configuration through the correction of vehicle motion distortion.

[0066] In conventional technology, autonomous golf cart vehicles using a GPS-based geofencing method, which is low-cost and has simple functions, or autonomous golf cart vehicles using a high-resolution map method, which is functionally superior but has high costs and scalability limitations, have been used.

[0067] On the other hand, the present invention adopts a separation-combination method that extracts high-dimensional semantic driving rules regarding 'what' to do from open source map data, and precisely generates information on 'where' and 'how' required for the physical control of the vehicle by fusion processing real-time data from multiple sensors mounted on the vehicle. This structure utilizes the rich semantic attribute information, which is an advantage of open source map data, while solving the problem of lack of geometric accuracy, which is a weakness of open source map data. By solving the problem of 'distortion caused by vehicle motion' that undermines the reliability of real-time sensor data, it is possible to achieve high reliability and precision comparable to expensive high-resolution map systems while being based on low-cost open source map data.

[0068] The process of controlling the following driving of a golf cart vehicle in the present invention will be described later with reference to [Fig. 9].

[0070] [Fig. 4] is a flowchart of the process of generating geofence data using open source map data in the present invention.

[0071] Step (S210): The autonomous vehicle control device (100) filters and extracts geometric feature data having golf course-related attribute tags from open source map data.

[0072] The autonomous driving vehicle control device (100) obtains open source map data for the golf course area, for example, in a compressed PBF (Protocol Buffer Binary Format) format. Since the total data size of the open source map data is in the tens of gigabytes, a high-performance computer device must be installed in the golf cart vehicle to process it as is. To avoid this, the present invention filters and extracts only the geographic data necessary for autonomous driving of the golf cart from the open source map data.

[0073] First, select and extract map data for a target area (e.g., a 5 km radius area around a golf course reference location) from open source map data using command-line utilities such as osmconvert or osmium. Then, select and extract geometric feature data with attribute tags related to the golf course (e.g., highway, leisure, natural, golf, building, etc.) using command-line utilities such as osmosis or ogr2ogr. For example, vector data such as course boundaries, cart paths, and hazard areas (water hazards, bunkers) can be extracted using attribute tags such as leisure=golf_course and highway=path.

[0074] This filtering process drastically reduces the amount of map data that the autonomous vehicle control unit (100) needs to process, enabling real-time processing on a low-spec HMCU. The filtering process (S210) may be performed before the operation of the golf cart vehicle or may be performed periodically during the operation of the vehicle. For example, the entire golf course may be set as a bounding box before the operation of the golf cart vehicle to process open source map data. Alternatively, the open source map data may be processed during the operation of the golf cart vehicle by setting an area of ​​100 meters around the vehicle location as a bounding box.

[0075] Step (S220): The autonomous vehicle control device (100) creates a local map database by converting the filtered and extracted geometric feature data into a preset geographical data structure.

[0076] The filtered and extracted geometric feature data is converted into a lightweight geographic data structure that is easy to process and search, preferably in the GeoJSON format. GeoJSON is an open standard format for representing geographic features and is suitable for the present invention as it can store both geometric information (e.g., points, lines, polygons) and attribute information (properties). The converted GeoJSON file is stored as a local map database in the storage device of the HMCU.

[0077] It is desirable to perform additional data transformation on local map databases, in which geometric feature data has been converted into a geographic data structure, so that they can be processed smoothly even on low-performance hardware. For example, GeoJSON or OSM XML files can be converted into a lightweight coordinate list (CSV or binary) that is easy to use in HMCU. In addition, it is advisable to generate spatial indexes, such as R-trees, for GeoJSON files to prepare the data suitable for real-time location-based queries (Point-in-Polygon, Nearest Neighbor).

[0078] Step (S230): The autonomous vehicle control device (100) sets up vehicle control rules according to the golf course environment for geometric feature data based on attribute tag information stored in a local map database.

[0079] Open source map data includes semantic attribute tags existing in a key=value format (e.g., highway, maxspeed, leisure, natural, landuse, building, golf, barrier). This is a process of semantically interpreting these attribute tags to suit the golf course environment and setting vehicle control rules (e.g., normal driving, deceleration, stopping, entry prohibition, etc.) for geometric feature data (geometric geospatial). At this time, specific vehicle control parameters (e.g., maximum speed, entry allowed or entry prohibited) can be included in the vehicle control rules. In this case, the system can be configured to semantically interpret combinations of multiple attribute tags. [Fig. 5] is an example diagram illustrating the mapping of vehicle control rules by semantically interpreting attribute tags to suit the golf course environment in the present invention. The vehicle control rules can be utilized in geofencing. Conventional geofencing operates based on binary logic to determine whether entry is allowed or prohibited, whereas the geofencing of the present invention enables intelligent control of golf cart vehicles according to the vehicle control rules set in the geometric feature data.

[0080] Step (S240): The autonomous vehicle control device (100) generates geofence data by combining geometric feature data and vehicle control rules.

[0081] In the present invention, geofencing is dynamically generated and activated by a combination of the current location of a golf cart vehicle, the geometric structure of open source map features, and their attribute tags. This forms a multi-layered policy zone in which rules are applied in real time when the golf cart vehicle enters a zone with specific attributes. This configures the local map database to function as a rule database, thereby implementing context-aware intelligent control beyond the simple boundary detection of conventional geofencing.

[0083] [Fig. 6] is a flowchart of the process for correcting magnetic motion distortion by multi-sensor fusion in the present invention.

[0084] LiDAR sensors, which autonomous driving systems rely on, suffer from data distortion caused by the vehicle's movement itself, a phenomenon known as ego-motion distortion. LiDAR sensors take time to complete a single scan (e.g., a 360-degree rotation) (e.g., 100ms). If the vehicle rotates or moves forward during this time, the points captured at the start and end of the scan are measured from different positions and orientations, resulting in the generation of a distorted point cloud. Such distorted LiDAR data leads to inaccurate perception of the vehicle's surrounding environment, and vehicle control based on this can result in safety issues such as deviations from the intended path or collisions. In particular, LiDAR-based autonomous driving can cause significant errors when the vehicle is turning. For instance, when a vehicle turns left, stationary external objects appear to move to the right within the LiDAR coordinate system, leading to unnecessary steering maneuvers.

[0085] For autonomous vehicle control, it is necessary to immediately correct distortions in LiDAR data input at every moment. The present invention utilizes an IMU sensor that provides data at a much higher frequency (e.g., 100-400 Hz) than LiDAR. That is, high-frequency angular velocity and acceleration data (high-frequency motion data) from an Inertial Measurement Unit (IMU) are used to compensate in real time for minute vehicle movements that occur during the short duration of a LiDAR scan. By utilizing the angular velocity and acceleration data from the IMU, virtual movements caused by vehicle rotation and movement are compensated, and the vehicle's 6-DOF pose can be precisely estimated (interpolated) from the exact timestamp at which each LiDAR point was measured. Subsequently, a 'de-skewed' point cloud is generated by converting each LiDAR point to a common coordinate system using this data. This enables reliable surrounding perception and precise position estimation of the vehicle using a combination of low-cost sensors mounted on the vehicle, without relying on the geometric inaccuracies of open-source map data.

[0086] Step (S310): The autonomous vehicle control unit (100) identifies a sequence of lidar points provided by the lidar sensor during the scanning period while the golf cart vehicle is driving.

[0087] A LiDAR point sequence consists of multiple LiDAR points, and each of these LiDAR points generated by the LiDAR sensor is assigned a timestamp corresponding to the time of capture. The autonomous vehicle control device (100) takes into account that each point of the LiDAR data has its own unique timestamp and performs real-time distortion compensation (de-skewing) processing using high-frequency motion data of the IMU sensor as shown below.

[0088] Step (S320, S330): The autonomous vehicle control device (100) calculates sequential minute attitude changes of the golf cart vehicle using high-frequency motion information of inertial measurement unit (IMU) data for each lidar point of the lidar point sequence, and calculates the attitude information of the golf cart vehicle at a series of points in time where each lidar point is captured by linear interpolation or spline interpolation of the minute attitude changes.

[0089] Generally, the LiDAR sensor generates LiDAR data at a low frequency of about 10 Hz, and the IMU sensor generates angular velocity (ω) and acceleration (a) at a high frequency of 100 Hz or higher. The LiDAR data and IMU data are each assigned a timestamp at the time of sensing. The autonomous vehicle control device (100) can very precisely model the vehicle attitude change during the approximately 0.1 seconds of the LiDAR scan by integrating the IMU data (ω, a). If the k-th LiDAR point in the LiDAR scan is measured at timestamp tk, the vehicle attitude at time tk can be precisely interpolated based on the IMU data. Mathematically, the vehicle attitude change can be modeled using a rotation matrix and a translation vector.

[0090] Specifically, the sequential minute attitude changes of the golf cart are first calculated from the IMU data. That is, the IMU data (ω, a) is processed in chronological order to calculate the minute rotation (delta_rotation) and minute translation (delta_translation) between each IMU measurement point. By accumulating these minute rotations and translations over the LiDAR scan cycle, the attitude change of the golf cart relative to the start of the LiDAR scan can be obtained. Subsequently, the relative attitude information of the golf cart at the time of capture is estimated for each LiDAR point in the LiDAR point sequence by linearly or spline interpolating the relative attitude changes of the golf cart calculated based on the IMU data.

[0091] Step (S340): The autonomous vehicle control device (100) generates a distortion-free point cloud in which distortion caused by the movement of the golf cart vehicle is corrected by converting each LiDAR point into a common coordinate system based on the attitude information of the golf cart vehicle calculated above. At this time, converting each LiDAR point into a common coordinate frame can be implemented by applying a homogeneous transformation matrix, which includes a rotation matrix and a translation vector obtained from the attitude information of the golf cart vehicle, to the LiDAR point coordinates.

[0092] This step is a process of generating a distortion-free point cloud by compensating for the movement of the vehicle occurring during the LiDAR scan cycle (e.g., 0.1 seconds) at the individual LiDAR point level.

[0093] First, each LiDAR point is converted into a common coordinate frame. That is, the coordinates of each LiDAR point are values ​​measured in the LiDAR sensor coordinate system moving with the vehicle. For each of these LiDAR points, they are converted into coordinates in the common coordinate system by homogeneous transformation based on the attitude information of the golf cart vehicle at the time of the corresponding scan. The common coordinate system is a fixed coordinate system, which is a 3D Cartesian coordinate system, and can be the vehicle coordinate system or the world coordinate system at the start of the scan (t=0).

[0094] Homogeneous transformation corresponds to a transformation that restores the vehicle's movement at the time of LiDAR sensing, and is a process that transforms all points as if they were measured simultaneously at the start of the scan. Through homogeneous transformation, all LiDAR points acquire coordinates in a fixed coordinate system. This coordinate transformation is performed individually and in real-time for multiple LiDAR points generated by a LiDAR scan. Specifically, if we denote the coordinates of a LiDAR point in the LiDAR sensor coordinate system as PL and the capture time as t, the rotation matrix RIMU(t) and the translation vector TIMU(t) are obtained from the attitude information of the golf cart vehicle at time t, and the coordinates (PW) in the joint coordinate system for the LiDAR point can be obtained by PW = RIMU(t)·PL + TIMU(t).

[0095] Through the above process, when all LiDAR points belonging to the point cloud are converted into a common coordinate system, the distortion is removed (de-skewed) and a geometrically accurate point cloud is generated. This distortion-free LiDAR data represents the geometrically accurate surrounding environment of the vehicle, through which the autonomous vehicle control device (100) can accurately recognize the location of road boundaries, obstacles, etc., with, for example, centimeter-level precision.

[0096] This is crucial for autonomous driving based on open-source map data for golf carts. Open-source map data defines narrow cart paths (highway=path) next to water hazards (natural=water) on golf courses. Uncompensated LiDAR data during sharp turns can perceive the edge of the path differently from its actual location, potentially causing the golf cart to deviate from the path and lead to accidents. By correcting for vehicle motion distortion in real-time, the vehicle's surrounding environment—such as the edge of the driving path—can be accurately sensed within an acceptable margin of error, thereby enabling safe driving.

[0097] Meanwhile, instead of the complex coordinate system transformation method described above, a method of lightweightly fusing LiDAR data and IMU data using a complementary filter can be adopted. This is applicable when considering the tracking driving function and is suitable for low-spec HMCUs due to the low computational load. Generally, a complementary filter is a signal processing technique used to obtain accurate and stable state estimates (e.g., attitude, angle) of a measurement target by fusing data obtained from two or more sensors through the combination of two complementary filters, such as a high-pass filter (HPF) and a low-pass filter (LPF). The complementary filter is configured to ensure the stability of sensor fusion by ensuring that the sum of the transfer functions of the two filters (HPF, LPF) meets a specific condition (ideally 1), and it operates by taking the advantages of two sensors with mutually complementary characteristics and offsetting their disadvantages.

[0098] In this invention, the complementary filter is effective in combining the advantages of LiDAR data and IMU data having different frequency characteristics. The gyroscope of the IMU is accurate for short-term rotational changes (high-frequency components), while LiDAR data provides the absolute position of the target over the long term (low-frequency components). When a golf cart turns, LiDAR data captures virtual movement as if the pursued target is rotating in the opposite direction, leading to steering errors. This problem can be solved by inputting IMU data and LiDAR data using a complementary filter.

[0099] Step (S350): The autonomous vehicle control device (100) estimates real-time vehicle state information regarding one or more of the position, speed, and direction of the golf cart vehicle by tightly coupling and fusing the undistorted point cloud, IMU data, and GNSS data using a Kalman filter.

[0100] Generally, the tightly coupled method fuses the raw measurements from each sensor within a single optimization framework, providing higher accuracy and robustness than the loosely coupled method. The Kalman filter can be composed of an Extended Kalman Filter (EKF) or an Unscented Kalman Filter (UKF). It corrects the accumulated error of IMU data using GNSS data and distortion-free LiDAR data (distortion-free point cloud) and performs stable vehicle position estimation through LiDAR and IMU-based odometry in areas where GNSS signals are weak. Since golf courses often consist of areas surrounded by mountains, sections with weak GNSS signals appear frequently; however, even when using a low-performance dual-band receiver in a golf cart, geographical position errors can be corrected by distortion-free point cloud and IMU data.

[0102] [Fig. 7] is a flowchart of a first embodiment of the geofencing control process for a golf cart vehicle in the present invention.

[0103] In the present invention, geofencing is a safety function that automatically stops a golf cart vehicle when it deviates from the normal areas of a golf course (e.g., driving path, fairway area, etc.) and approaches non-driving areas (e.g., hazard area, OB area, green area, bunker area, repair area, etc.). In addition, it is a safety function that prevents accidents from occurring during the autonomous driving process of a golf cart vehicle by dividing the golf course into various areas and setting vehicle control rules for each area.

[0104] Step (S410): The autonomous driving vehicle control device (100) calculates the predicted driving path of the golf cart vehicle using the previously estimated real-time vehicle state information (position, speed, direction) of the golf cart vehicle and a preset vehicle kinematic model.

[0105] In this case, the vehicle kinematic model can be defined as a kinematic bicycle model, which simplifies the vehicle into a front wheel, a rear wheel, and a rigid body connecting the two wheels. In the kinematic bicycle model, the vehicle's state of motion can be expressed by the following differential equation.

[0106] x′ = v·cos(θ)

[0107] y′ = v·sin(θ)

[0108] θ′ = Lv·tan(δ)

[0109] Here, (x, y) is the vehicle position, θ is the vehicle direction angle, v is the vehicle speed, L is the vehicle wheelbase, and δ is the front wheel steering angle.

[0110] The future driving path of a golf cart vehicle can be calculated numerically by using Euler integration to calculate the vehicle kinematics model represented by the differential equation above. That is, the predicted driving path of the golf cart vehicle is generated by taking the previously estimated current position of the golf cart vehicle as a starting point and repeatedly calculating the future state by numerically integrating the differential equation of the vehicle kinematics model (e.g., Euler integration) at short time intervals Δt.

[0111] Step (S420): The autonomous vehicle control unit (100) determines whether there is a potential intersection between the predicted driving path and the geofence boundary according to the geometric feature data stored in the geofence data.

[0112] Step (S430, S440): The autonomous vehicle control device (100) maintains the current driving state of the golf cart vehicle if no potential intersection is identified as a result of the above judgment.

[0113] Step (S430, S450 ~ S470): When a potential intersection is identified as a result of the above judgment, the autonomous vehicle control device (100) generates a vehicle driving command using a physics-based equation of motion based on the current speed of the golf cart vehicle and the remaining distance to the geofence boundary to correspond to the vehicle control rule corresponding to the geometric feature data.

[0114] For the purpose of geofences, in most cases, the vehicle driving command is a deceleration command. The deceleration command is executed repeatedly until the speed of the golf cart vehicle falls below a threshold according to the vehicle control rules. If the geometric feature data corresponds to a non-operational zone, the threshold is 0, so the deceleration command is executed repeatedly until the golf cart vehicle comes to a stop.

[0115] Referring to [Fig. 5], the threshold of the vehicle speed is set differently according to the vehicle control rules. The control loop measures the current vehicle speed v_current and the remaining distance d_remaining to the geofence boundary in every cycle, and calculates the equivalent deceleration a required at the current moment to stop precisely at the boundary point using the following equation of motion for constant acceleration.

[0116] a = 2·d_remaining - v_current

[0117] In addition, for actual application, a safety factor (e.g., 0.8 to 0.9) can be added to account for system response delays or external factors, and the deceleration command a_cmd can be calculated as follows.

[0118] a_cmd = 2·d_remaining·safety_factor - v_current

[0119] The deceleration command calculated in this way is transmitted to the vehicle drive module to enable actual vehicle operation accordingly.

[0121] [Fig. 8] is a flowchart of a second embodiment of the geofencing control process for a golf cart vehicle in the present invention.

[0122] Step (S510): The autonomous vehicle control device (100) obtains vehicle control rules (e.g., precision deceleration mode; maximum speed 5 km / h) from geofence data by referring to the real-time vehicle status information (position, speed, direction) of the golf cart vehicle estimated earlier. For example, using an R-tree index for a local map database, it searches for all geometric feature data (polygons, lines) that contain or are adjacent to the vehicle's current position (current_pose) and obtains the corresponding vehicle control rules.

[0123] Step (S520): The autonomous vehicle control device (100) uses a distortion-free point cloud to identify driving restriction environments, such as path edges, static obstacles, dynamic obstacles, etc., around the golf cart vehicle in real time while driving.

[0124] Step (S530, S540): The autonomous vehicle control device (100) generates a short-term target path in real time by combining vehicle control rules based on the current location and a driving restriction environment identified in real time, so as to satisfy all of these conditions. Then, it calculates vehicle driving control commands (e.g., target steering angle, target acceleration / deceleration) in real time to achieve the short-term target path. When a driving restriction environment is identified during vehicle driving, the short-term target path is updated immediately, and the vehicle driving control commands are updated accordingly. At this time, if multiple driving policies overlap (e.g., the 'green' area within a 'golf course'), the policy with the highest pre-set priority is selected as the final policy. If the priority is unclear or identical, a policy that is relatively more restrictive (e.g., lower driving speed) is selected to calculate the vehicle driving control commands.

[0125] Step (S550): The autonomous vehicle control device (100) transmits a vehicle drive control command to the vehicle drive module so that actual vehicle drive is performed accordingly.

[0127] [Fig. 9] is a flowchart of the following driving control process for a golf cart vehicle in the present invention.

[0128] In the present invention, following driving refers to a convenience function that controls a golf cart vehicle to safely follow a target in front of the golf cart vehicle while maintaining a constant distance by recognizing the target in real time using a front LiDAR sensor. For example, when a golfer stops to take a shot, the golf cart vehicle can be implemented to smoothly stop on its own at a certain distance (e.g., 2 meters) behind the player, and when the golfer walks along the fairway for the next shot, the golf cart vehicle can be implemented to drive accurately following the golfer's path while maintaining a constant distance (e.g., 4 meters).

[0129] Step (S610, S620): The autonomous vehicle control device (100) performs object recognition processing on a distortion-free point cloud to identify clusters in front of the vehicle and specifies a target to follow among them. At this time, for object recognition processing, for example, a clustering algorithm such as DBSCAN (Density-Based Spatial Clustering of Applications with Noise) can be used. When object recognition processing is performed, high-density areas in the point cloud output by the LiDAR sensor are clustered into object units. Then, based on object characteristics, such as the general size or movement speed of a person, a target to follow (golfer) is specified among the clusters, and the target is managed so as not to be lost through inter-frame tracking.

[0130] Step (S630): The autonomous driving vehicle control device (100) calculates the target speed of the golf cart vehicle and controls the drive based on the distance error with the follower (i.e., the difference between the current distance and the target distance). For example, the target speed is calculated so that the vehicle accelerates smoothly when the follower moves away, and conversely, decelerates smoothly when the follower moves closer.

[0131] Step (S640): The autonomous driving vehicle control device (100) calculates the target steering angle of the golf cart vehicle and controls the driving based on the angle error with the follower (i.e., the difference between the current angle and the target angle (e.g., 0 degrees)). For example, if the follower moves to the left, the steering is controlled smoothly to the left, and conversely, if the follower moves to the right, the steering is controlled smoothly to the right.

[0132] Step (S650 ~ S670): The autonomous driving vehicle control device (100) brakes the golf cart vehicle when the follower approaches the golf cart vehicle within a preset threshold. This is a measure to prevent a collision accident between the follower and the golf cart vehicle.

[0133] Additionally, the autonomous driving vehicle control device (100) brakes the golf cart vehicle when the follower deviates from the golf cart vehicle by more than a preset threshold angle. This is a measure to prevent the control of the steering control from becoming unstable.

[0134] In addition, the autonomous driving vehicle control device (100) brakes the golf cart vehicle when the follower is lost. This is a measure to prevent the golf cart vehicle from operating unpredictably when the follower is lost.

[0136] Meanwhile, the present invention can be implemented in the form of computer-readable code on a computer-readable non-volatile recording medium. Various types of storage devices exist as such non-volatile recording media, such as hard disks, SSDs, CD-ROMs, NAS, magnetic tapes, web disks, and cloud disks. Additionally, the present invention may be implemented in the form of a computer program stored on a medium to execute a specific procedure in combination with hardware. Explanation of the symbols

[0138] 100: Autonomous driving vehicle control unit 110: Data Collection Unit 111 : GNSS receiver 112: LiDAR sensor unit 113: IMU sensor section 120 : Autonomous driving processing unit 121 : Map data storage unit 122 : Purpose-driven driving processing unit 123 : Geofencing processing unit 124 : Tracking driving processing unit 130: Vehicle drive control unit 140: User Interface Section

Claims

Claim 1 A method for geofencing and following driving control of an autonomous golf cart vehicle based on open source map data and ego motion compensation, performed by an autonomous vehicle control device installed in an autonomous driving golf cart vehicle equipped with autonomous driving capabilities, comprising: a first step of generating geofence data by analyzing semantic attribute information included in the open source map data and combining geometric feature data for a golf course with vehicle control rules; a second step of recognizing and setting a follower target in front of the golf cart vehicle using a front LiDAR sensor; a third step of collecting multi-sensing data including GNSS data, LiDAR data, and IMU data related to the golf cart vehicle through a multi-sensor suite; a fourth step of correcting ego motion distortion through fusion processing of the multi-sensing data and obtaining real-time vehicle status information related to the position, speed, and direction of the golf cart vehicle; and a fifth step of controlling the operation of the golf cart vehicle according to vehicle control rules when identifying the golf cart vehicle's approach to a geofence boundary based on the geofence data based on the real-time vehicle status information of the golf cart vehicle. The method is configured to include a sixth step of controlling the operation of the golf cart vehicle to follow a target in front of the golf cart vehicle while maintaining a certain distance by recognizing the target using a front lidar sensor; wherein the fourth step comprises: a step of identifying a lidar point sequence provided by the lidar sensor during the scanning period while the golf cart vehicle is driving as lidar data; a step of calculating sequential minute attitude changes of the golf cart vehicle using high-frequency motion information of the IMU data for each lidar point of the lidar point sequence; and a step of calculating attitude information of the golf cart vehicle at a series of time points in which each lidar point is captured by linear interpolating or spline interpolating the minute attitude changes.A method for geofencing and following driving control of an autonomous golf cart vehicle based on open source map data and ego motion compensation, characterized by comprising: a step of generating a distortion-free point cloud in which distortion caused by the movement of the golf cart vehicle is corrected by converting each LiDAR point into a common coordinate system based on attitude information of the series of golf cart vehicles; and a step of estimating real-time vehicle state information related to the position, speed, and direction of the golf cart vehicle by tightly coupling and fusing the distortion-free point cloud, the IMU data, and the GNSS data using a Kalman filter. Claim 2 A method for geofencing and following driving control of an autonomous driving golf cart vehicle based on open source map data and self-driving vehicle motion compensation according to claim 1, wherein the first step comprises: filtering and extracting geometric feature data having golf course-related attribute tags from the open source map data; formatting the filtered and extracted geometric feature data into a preset geographic data structure to create a local map database; mapping and setting vehicle control rules according to the golf course environment for the geometric feature data based on attribute tag information stored in the local map database; and generating geofence data by a combination of the geometric feature data and the vehicle control rules. Claim 3 delete Claim 4 The method of claim 1, wherein the fifth step comprises: calculating a predicted driving path of a golf cart vehicle using real-time vehicle state information of the golf cart vehicle and a preset vehicle kinematics model; determining whether there is a potential intersection between the predicted driving path and a geofence boundary according to geometric feature data stored in the geofence data; maintaining the current vehicle driving state if no potential intersection is identified as a result of the determination; and generating and transmitting a vehicle driving command based on the current speed of the golf cart vehicle and the remaining distance to the geofence boundary to correspond to a vehicle control rule corresponding to the geometric feature data if a potential intersection is identified as a result of the determination. Claim 5 The method of claim 1, wherein the fifth step comprises: obtaining a vehicle control rule from geofence data by referring to real-time vehicle status information of the golf cart vehicle; identifying a driving restriction environment around the golf cart vehicle in real-time while driving using the distortion-free point cloud; generating a short-term goal path in real-time by combining the vehicle control rule and the driving restriction environment identified in real-time; calculating a vehicle driving control command in real-time to achieve the short-term goal path; and transmitting the vehicle driving control command to a vehicle driving module; characterized by comprising a geofencing and following driving control method for an autonomous driving golf cart vehicle based on open source map data and self-vehicle motion compensation. Claim 6 The method of geofencing and following driving control of an autonomous driving golf cart vehicle based on open source map data and self-vehicle motion compensation according to claim 4, wherein the sixth step comprises: a step of performing object recognition processing on the distortion-free point cloud to identify clusters in front of the vehicle and specifying a follower among them; a step of calculating a target speed of the golf cart vehicle based on a distance error with the follower and controlling driving; a step of calculating a target steering angle of the golf cart vehicle based on an angle error with the follower and controlling driving; a step of braking the golf cart vehicle when the follower approaches the golf cart vehicle within a preset threshold; a step of braking the golf cart vehicle when the follower deviates from the golf cart vehicle by more than a preset threshold angle; and a step of braking the golf cart vehicle when the follower is lost. Claim 7 A computer program stored in a computer-readable storage medium for executing a geofencing and following driving control method for an autonomous driving golf cart vehicle based on open source map data and self-driving vehicle motion compensation according to any one of claims 1, 2 and 4 to 6 on a computer.

Images