WO2023163685A1

WO2023163685A1 - Autonomous track tensioning mechanism and operation method for unmanned ground vehicles

Info

Publication number: WO2023163685A1
Application number: PCT/TR2023/050172
Authority: WO
Inventors: Ahmet Semih PARLAK; Halil AKBULUT; Necati Mert CELEN; Muhittin SOLMAZ; Mehmet Onur OZCELİK
Original assignee: Havelsan Hava Elektronik Sanayi ve Ticaret AS
Current assignee: Havelsan Hava Elektronik Sanayi ve Ticaret AS
Priority date: 2022-02-23
Filing date: 2023-02-22
Publication date: 2023-08-31
Anticipated expiration: 2024-08-23

Abstract

The invention relates to an autonomous track tensioning mechanism that enables the unmanned ground vehicle to have optimum driving performance in different terrain (stony, sandy, muddy, grassy, wet ground, etc.) and environmental conditions in the missions performed by UGVs, and to an operation method for the autonomous track tensioning mechanism. The track tensioning mechanism is a frequently used subsystem in UGVs. In order to successfully complete different missions such as logistics deployments, search and rescue, autonomous patrolling, reconnaissance and surveillance; the tension of the rubber track in the vehicle is adjusted autonomously by using of camera (2), sensors, artificial intelligence algorithms, machine learning-based classification algorithms and deep learning-based object detection algorithms according to the terrain conditions during its deployment to the mission area, thereby ensuring that the UGV has optimum driving performance on different terrians and environmental conditions.

Description

AUTONOMOUS TRACK TENSIONING MECHANISM AND

OPERATION METHOD FOR UNMANNED GROUND VEHICLES

Technical Field

The invention relates to an autonomous track tensioning mechanism that enables the unmanned ground vehicle to have optimum driving performance in different terrains (stony, sandy, muddy, grassy, wet ground, etc.) and environmental conditions in the missions performed by unmanned ground vehicles, and to an operation method for the autonomous track tensioning mechanism.

Track tensioning mechanism is a subsystem that frequently used in unmanned ground vehicles. In order to successfully complete different missions such as logistics deployment, search and rescue, autonomous patrolling, reconnaissance and surveillance; the tension of the rubber track of the vehicle is adjusted autonomously by using camera, sensors, artificial intelligence algorithms, machine learning-based classification algorithms and deep learning-based object detection algorithms according to the road conditions during its deployment to the mission area, thereby ensuring that the unmanned ground vehicle has optimum driving performance on different terrains and environmental conditions.

Background

Today, the track tensioning mechanism in tracked vehicles is installed in two different ways. These methods are hydraulic and mechanical track tensioning mechanisms. Hydraulic track tensioning mechanism is used in vehicles that require larger tension force (e.g., Heavy Tracked Vehicles, Main Battle Tanks, etc.). Mechanical track tensioning mechanisms are used in vehicles where the tension force is smaller (Small and Medium Class Unmanned Ground Vehicles, Explosive Ordinance Disposal Robots, etc.). Hydraulic and mechanical track tensioning mechanisms have variable strokes and for different terrain conditions, the tracks must have different tension forces. The track tension value changes the distribution of the force transmitted between the power pack and the track force of the vehicle and the transmitted force from the tracked vehicle to the ground. The track tensioning operation can be performed by the operator of the vehicle before the operation, or it can be adjusted during the operation. In mechanical track tensioning mechanisms, in general, stroke can only be adjusted by the operator before the operation.

Hydraulic track tensioning mechanisms tend to have leakage problem due to working with high-pressure fluids contained in hydraulic pistons, thereby making the track tensioning mechanism susceptible to dust and dirt, accumulated dust and dirt directly or indirectly affect the mobility of the mechanism.

In mechanical track tensioning mechanisms, the mobility of the mechanism is directly or indirectly affected due to dust and dirt accumulated in the screw threads.

The hydraulic track tensioning mechanism is manually operated by the operator before the operation by means of hydraulic pumps in the vehicle and the track tensioning process is performed manually. The track tensioning process in the hydraulic tensioning mechanism can also be performed during the operation. The mechanical track tensioning mechanism can be manually adjusted by the operator only before the operation. Both solutions mentioned herein rely on the operator's knowledge of the operation site and his operational experience. However, as it is known, most of the errors that occur during the operation are human-based. An autonomous control mechanism created by means of reliable tests will minimize the vulnerabilities that may be caused by the human factor.

The track tensioning mechanism is an open-loop operation and its effect on vehicle maneuverability and driving ability is evaluated by user experience. However, tensioning the track system at the wrong value will negatively affect vehicle performance. For this reason, the track tensioning process should be kept within a closed control loop in order not to cause errors in vehicle maneuverability and off-road performance.

In the International patent document numbered W02019109191A1, which is in the state of the art, it is mentioned about systems and methods in which information about a vehicle (for example, an agricultural vehicle or another offroad vehicle) comprising a track system is acquired by monitoring the vehicle. The document also describes controlling the tensioning of the track system. The sensors mentioned in the document are used to detect the level of wear on the track, the amount of wear and possible failures in the track suspension area by using image processing and machine learning algorithms. In order for the algorithms to perform this process, they need to be trained with images of intact tracks, worn tracks, regional images of worn tracks, and images of possible failure scenarios as learning dataset. In order for the algorithms to work in real-time and faultless, they need to be optimized according to the dataset used in learning and the main purpose to be achieved. Generally used machine learning algorithms that are not optimized for the purpose can be applied to any data but will not give reliable results. It may produce results such as failure to detect track wear and suspension failures, incorrect detection, and misgrading of track wear zones. Thus, machine learning algorithms must be customized. Similarly, the algorithm trained for track damage detection and fault detection will not perform detection of terrain type. So, the learning dataset should be trained with images of different terrains and the data should be filtered. So that it can perform more reliable manner under different outdoor conditions.

In the Korean patent document numbered KR20200122800 A (D2), which is in the state of the art, the systems and methods developed for controlling the tension of the track system are mentioned. In the United States patent document numbered US10099735B2, which is in the state of the art, a system for monitoring track tensioning for a track assembly of a heavy duty vehicle may comprise a track tensioning assembly has a fluid-driven actuator.

In the United States patent document numbered US2020114992A1, which is in the state of the art, a track tension adjustment system is mentioned. In the related document, it is stated that the track tensioning can be adjusted by means of single board computer.

In the Chinese patent document numbered CN108297634A, which is in the state of the art, the invention relates to the technical field of traffic equipment, in particular to a multi-functional smart railway platform consists a tracked powertrain system. In this document, it is mentioned that the environmental measurement system may comprise an inertial navigation system and camera.

When the existing studies are examined, there is a need to realize a track tensioning mechanism that enables the unmanned ground vehicle to have optimum driving performance in different terrains and environmental conditions as well as an autonomous operation method of tensioning mechanism.

Objective of the Invention

The objective of this invention is to realize an autonomous track tensioning mechanism that enables the unmanned ground vehicle to have optimum driving performance in different terrains (stony, sandy, muddy, grassy, wet ground, etc.) and environmental conditions in the missions performed by unmanned ground vehicles as well as an autonomous operation method of track tensioning mechanism.

Another objective of the present invention is to realize an track tensioning mechanism as well as an autonomous operation method of track tensioning mechanism enable the use of vehicle performance parameters in the planning of operations in different geographies and terrain types in subsequent operations by obtaining important information for the logistics and sustainment of an operation planned after the operation since vehicle performance (vehicle speed, motor speed, power consumption, etc.) on different terrains is recorded.

Detailed Description of the Invention

The autonomous track tensioning mechanism realized for achieving the objectives of the present invention is shown in the attached figures.

These figures are as follows;

Figure 1: An exploded view of the autonomous track tensioning mechanism subject to the invention on the unmanned ground vehicle.

Figure 2: A perspective view of the autonomous track tensioning mechanism subject to the invention on the unmanned ground vehicle.

Figure 3: A top view of the unmanned ground vehicle when the autonomous track tensioning mechanism subject to the invention is on the unmanned ground vehicle.

Figure 4: B-B cross-sectional view of the unmanned ground vehicle when the autonomous track tensioning mechanism subject to the invention is on the unmanned ground vehicle.

The parts in the figure are numbered one by one and the corresponding numbers are given below.

1. Vehicle body

2. Camera

3. Track tensioning mechanism 4. Linear actuator

5. Inertial navigation system

6. Motor

7. Motor driver

8. Encoder

9. Artificial intelligence computer

B. B axis

The autonomous track tensioning mechanism developed for unmanned ground vehicles subject to the invention comprises the following parts:

A camera (2) which is located in front of the vehicle body (1) and performs the function of acquiring an image of the ground,

A linear actuator (4) which is located inside the track tensioning mechanism (3) of the vehicle and used for tensioning the track of the unmanned ground vehicle,

An inertial navigation system (5) which is located on the vehicle body (1) and measures the linear (x, y, z) acceleration and angular (roll, pitch, yaw) acceleration data of the vehicle in three directions by means of sensors located thereon,

An unmanned ground vehicle motor (6) which is located inside the vehicle body (1) and provides instant power transmission of the vehicle,

An motor driver (7) which is located inside the vehicle body (1) and performs the task of feeding the motors (6) of the vehicle with the required force values in consideration of the commands given by the remote-control unit,

An encoder (8) which is located on the motor (6) and measures the instantaneous track speed of the vehicle,

An artificial intelligence computer (9) which is located inside the vehicle body (1), processes sensor data from the inertial navigation system (5), the unmanned ground vehicle motor (6) and the encoder (8) and applies the required tension force to the track of the unmanned ground vehicle by means of a linear actuator (4).

The autonomous track tensioning mechanism subject to the invention is integrated into the unmanned ground vehicle. The camera (2) delivers the image of the terrain on which the unmanned ground vehicle is located the artificial intelligence computer (9) of the vehicle, and also sensor data from the inertial navigation system (5), the motor driver (7), the unmanned ground vehicle motor (6) and the encoder (8) are delivered to the artificial intelligence computer (9). By means of a deep learning-based object detection algorithm, machine learning algorithm, artificial intelligence algorithm and sensor fusion technique in the artificial intelligence computer (9), the ground on which the unmanned ground vehicle is located is identified and the terrain type-tension force table prepared in advance is checked. The required tension force is calculated by the artificial intelligence computer (9). The determined force is delivered to the linear actuators (4) by the artificial intelligence computer (9). After this process, the sensor data from the motor (6), the encoder (8), the inertial navigation system (5) and the motor driver (7) of the vehicle are also delivered to the artificial intelligence computer (9) together with the vehicle orientation and motor current data. In consideration of these data, the artificial intelligence computer (9) adjusts the tension force by means of the linear actuator (4) if necessary corrections are needed in the tension force.

The method of operation of the autonomous track tensioning mechanism subject to the invention comprises the following steps:

- Acquiring the ground image from the vehicle camera (2),

- Processing the acquired terrain image by an artificial intelligence computer (9), which includes machine learning, deep learning and artificial intelligence methods, - Detecting the terrain type information on which the unmanned ground vehicle is located by a deep learning-based object detection algorithm by using a convolutional neural network,

Determining the required tension force from the terrain type-tension force table contained in the artificial intelligence computer (9) by a machine learning algorithm,

Transmitting the determined tension force to the linear actuators (4) located in the track tensioning mechanism,

- Moving the vehicle by means of a command given by a remote-control unit,

- Delivering the rotational speed information from the inertial navigation system (5), the UGV motor (6) and the encoder (8) as well as the current data from the motor driver (7) to the artificial intelligence computer (9), Comparing the commands given by the remote-control unit in the artificial intelligence computer (9) and the results of the fusion of sensor data with the artificial intelligence algorithm,

Sending the correction signals to the linear actuators (4) in the track tensioning mechanism (3) that they need to apply as a result of the comparison.

In the method subject to the invention, the dynamics and kinematics of the unmanned ground vehicle are evaluated by means of the sensors located on the vehicle, the environment in which it is located is detected, a meaningful conclusion is drawn by fusing all the inferred data, and by this conclusion, a solution to the performance problem in different terrains is provided.

In the method subject to the invention, the terrain type detection can be successfully determined by training a deep learning-based object detection algorithm by using the camera (2) images collected from different terrains in order to perform the terrain type information detection process. These processes are, respectively, collecting images from the vehicle camera (2), marking the related terrain from the collected images, preparing the convolutional neural network for object detection after the marking stage, and finally training the algorithm.

The satellite positioning system module located in the inertial navigation system (5) can measure the position of the vehicle on the earth and the linear acceleration of the UGV can be measured with the accelerometer. By integrating the linear accelerations, the linear speed can be confirmed, and by integrating this linear speed, the position of the vehicle can be confirmed. The angular speed of the UGV can be detected with the gyroscope located in the inertial navigation system (5). The angular speed of the vehicle and the orientation information of the vehicle obtained by means of the satellite positioning system module and magnetometer can be monitored instantaneously. The instantaneous vehicle position and orientation obtained from the sensors on the vehicle are compared with the path given by the remote-control and the amount of deviation of the UGV from the path is controlled.

In the method, data from sensors such as the inertial navigation system (5), the encoder (8) etc. are fused by using of an artificial intelligence algorithm. This generated data is compared with the commands given by the remote-control unit. As a result of the comparison, it is observed whether the vehicle follows the given commands or not and the tension of the track tensioning mechanism is adjusted by linear actuators (4).

In the method, the encoder (8) connected to the shaft of the vehicle's electric motor (6) measures the speed of the traction gear that drives the track. During the slippage between the ground and the track, the spur gear and the track move but the UGV cannot move on the ground and the track turns idle. Since the position information of the UGV is obtained from the sensors on the vehicle, it can be checked whether there is any slippage when this information is compared with the information of the road the vehicle should have proceeded, which is calculated with the motor (6) revolution information obtained from the encoder (8). Similarly, by comparing the orientation information received from the vehicle with the orientation information of the vehicle calculated from the spur gear revolution information from the encoder (8), it is possible to obtain information about whether there is slippage on the right or left track.

As a result of the tests performed in the method and the data collected, an optimum terrain type-tensioning force table was created. The required tension force is determined by comparing the terrain type information generated during the operation with the terrain information data in the table by means of a machine learning algorithm. In order to use machine learning algorithms more efficiently, the following processes are first performed on the collected data: filtering missing or empty data, calculating the direct proportions and inverse proportions of the data to each other by performing exploratory data analysis, updating the information related to this calculation, scaling the data, filtering outlier data and normalizing the data. After these processes are applied, the machine learning algorithm will give better results.

The power consumed by the UGV on different terrain types and conditions (slope, surface profile, etc.) and the currents consumed by the electric motors are determined by performed tests. These data are recorded under the terrain type label. During any operation, the UGV detects the terrain on which it is traveling, while the slope information of the terrain is obtained from the inertial navigation system (5) as well as the current and power consumed from the motor drivers (7). The obtained information is compared with previously performed tests. If the UGV consumes a higher current than usual for similar terrain type and structure, there is a possibility that the amount of track tensioning is excessive. For this reason, the motors (6) are overloaded in terms of torque while the vehicle is traveling. This information is confirmed by the amount of deviation from the path, thereby enabling the optimal amount of track tensioning to be determined so as not to deviate from the path. The determined amount of track tensioning is realized by the linear electric actuator on the UGV. If there is no deviation from the path after the track tension correction has been made, the track tension amount realized for the instantaneous terrain type and structure is accepted as the correct parameter. Thus, the algorithm and data that determine the amount of track tensioning according to the terrain type and structure are not static and constant; they dynamically adjust the amount of track tensioning according to changing conditions.

Claims

CLAIMS The invention relates to an autonomous track tensioning mechanism developed for UGVs, characterized in that it comprises the following parts: an inertial navigation system (5), which is located on the vehicle body (1) and measures the linear acceleration and angular acceleration data of the vehicle in three directions by means of sensors located thereon, a motor driver (7), which is located inside the vehicle body (1) and performs the task of feeding the vehicle in accordance with the commands given by the remote-control unit, an encoder (8), which is located on the motor (6) and measures the instantaneous track speed of the vehicle, an artificial intelligence computer (9) which is located inside the vehicle body (1), processing sensor data obtained from the inertial navigation system (5), the UGV motor (6) and the encoder (8) and applying the required tension force to the track of the UGV by using of an electric powered linear actuator (4). The invention relates to the autonomous track tensioning mechanism developed for UGVs according to claim 1, characterized in that it comprises a camera (2) which is located in front of the vehicle body (1) and performs the function of acquiring an image of the terrain. The invention relates to the autonomous track tensioning mechanism developed for UGVs according to claim 1, characterized in that it comprises a linear actuator (4) which is located inside the track tensioning mechanism (3) of the vehicle and used for tensioning the track of the UGV. The invention relates to the autonomous track tensioning mechanism developed for UGVs according to claim 1, characterized in that it comprises an UGV motor (6) which is located inside the vehicle body (1) and provides instant power transmission of the vehicle. The invention relates to an operation method for an autonomous track tensioning mechanism developed for UGVs, characterized in that it comprises the following steps:

Acquiring the terraib image from the vehicle camera (2),

- Processing the acquired terrain image by an artificial intelligence computer (9), which includes machine learning, deep learning and artificial intelligence methods,

- Detecting the terrain information on which the UGV is located by a deep learning-based object detection algorithm by using a convolutional neural network,

Determining the required tension force from the terrain type - tension force table contained in the artificial intelligence computer (9) by a machine learning algorithm,

- Moving the vehicle by using a remote-control unit,

- Delivering the inertial navigation system (5), the UGV motor (6) and the rotational speed of motor from the encoder (8) as well as the consumed current data from the motor driver (7) to the artificial intelligence computer (9),

Comparing in the artificial intelligence computer (9), the commands given by the remote-control unit and the results of the fusion of sensor data with the artificial intelligence algorithm, Sending the correction signals to the linear actuators (4) in the track tensioning mechanism (3) that they need to apply determined as a result of the comparison.