TR2022021363A2 - AUTONOMOUS AUTOPILOT COMPUTER ARCHITECTURE SYSTEM - Google Patents

Abstract

The invention relates to a highway autopilot computer architecture system that consists of one or more computers that can operate on the vehicle with network access and that can operate the entire process of making and executing decisions autonomously in each cycle, enabling the land vehicle to move autonomously on highways.

Description

DESCRIPTION AUTONOMOUS AUTOPILOT COMPUTER ARCHITECTURE SYSTEM Technical Field The invention relates to a highway autopilot computer architecture system consisting of one or more computers that can operate on the vehicle with network access and that can operate the entire process of making and executing decisions autonomously in each cycle, enabling the land vehicle to move autonomously on highways. Known State of the Art Autonomous vehicle, also known as robot vehicle, driverless vehicle, is a type of automobile that can perceive its environment and move with little or no human input. Autonomous vehicles contain various sensors that use measurement units such as radar, computer vision, Lidar, sonar, GPS, odometer and inertia to perceive their environment. Advanced control systems interpret sensory information to define appropriate navigation paths, obstacles and relevant signs. There are applications in the state of the art where special facilities have been developed for autonomous highway vehicles. However, there are many variables such as newly constructed roads in rapidly urbanizing countries, and the system design needs to be made taking these variables into account. Some of the urban applications can benefit from special infrastructure. The physical infrastructure can include vehicle-to-vehicle and vehicle-to-infrastructure communication equipment, ground-based units for global navigation systems, special facilities comparable to bus and bicycle lanes, on-street parking restrictions, and special road or pavement modifications. The digital infrastructure can include the maintenance of highly detailed highway maps and related traffic operations data. This special infrastructure can be limited to a manageable set of corridors used by a particular urban mobility system, if really necessary. A solution has been proposed for the limitations encountered in the sensor detection system of an autonomous or semi-autonomous vehicle due to the limits of the sensors. The road information obtained from the sensors and the kinematic properties of the surrounding vehicles are shared between autonomous vehicles via a different server environment or via vehicle-to-vehicle (V2V) communication. The use of the obtained data in the autonomous vehicle has been examined under three different tasks. These are; strategic, tactical and operational tasks. Strategic tasks; the decision-making and route-generating mechanisms of the autonomous vehicle, tactical tasks; the determination of the vehicle speed and steering movements throughout the course in order to realize the created route, and operational tasks provide the actuator control that will perform these planned maneuver movements of the vehicle. The environmental data obtained in the server environment or from the surrounding vehicles are used during the performance of these tasks. It has been suggested that V2V messages can be provided with a special short-range wireless communication protocol. It is particularly suitable for use in low power and short-medium range operating conditions. However, in the mentioned system, there is no structure that systematically organizes the data flow in the extraction of static information such as physical properties of the road, lane lines and traffic components and the extraction of trajectory and maneuver estimates related to other moving objects such as vehicles and pedestrians in the environment. In another patent document in the known art with the publication number USZOl70168485Al; vehicle control is performed with iterative methods throughout the course of the autonomous vehicle from its starting position to its destination. The method determines neighboring location points with iterative calculations within the vehicle limits during the vehicle's journey and calculates the vehicle's state transitions to determine which point it will move to. In each iteration, the vehicle determines the point with the optimal result within the state transitions as the target location for the calculation of the most suitable neighboring location for movement. After moving to this new location, it calculates new vehicle state transitions and provides the calculation of new neighboring location points. In determining the optimal location, the optimal time is taken as a basis by using optimization methods and it is claimed that model-based vehicle control is performed by including the vehicle's limits during the calculation. In the mentioned patent application, the vehicle movement planning is not disclosed as a hierarchical system during the autonomous vehicle's journey from the starting position to the destination point and there is no request-approval mechanism aimed at increasing vehicle safety. In the patent document with the publication number USZOO90319112A1, a method has been developed to plan how fast these trajectories should be crossed while avoiding obstacles during trajectory planning while autonomous vehicles are navigating in obstacle-rich environments such as urban areas. In traditional path planning algorithms, the necessary speed information is not produced while the vehicle is traversing these trajectories. It is mentioned that the speed target of the vehicle should be calculated correctly in order to ensure navigation safety and driver (8) comfort. The created system includes one or more sensors, speed planner, path planner, vehicle guidance and control systems and one or more actuators. These created architectures have a modular structure to be used not only for land vehicles but also for an unmanned aerial vehicle. It is claimed that the obstacles around the autonomous vehicle contain universal data types such as location, speed or direction of movement. However, the invention mentioned does not have a feature such as determining the motion profile by making calculations for collision avoidance and obstacle avoidance so that the autonomous vehicle can safely perform lane change or lane following actions. It does not include an architecture that evaluates whether the maneuver decision taken is appropriate in terms of realizability, safety and comfort. Another application in the state of the art is the patent document with the publication number USlO377375B2. The invention mentioned in this file is related to the creation of all working modules required for the autonomous vehicle with a top-down approach. The invention is an architectural module and has an adaptable structure independent of hardware or software manufacturers. Each module component is created to work in a compatible manner with different vehicle control systems and sensor subsystems. The architecture; It includes human-machine interface and vehicle-environment (V2X) interaction systems as well as sensing, planning and control modules. In the invention in question, the "System Controller" module representing the system observation module communicates partially with environmental modules at a lower level and is positioned in a structure far from modularity. This causes the data flow to the user interface to be interrupted in the event of an error that may occur in the "System Controller" module. In the patent document with the publication number USlO955847B2, which is in the state of the art, there are units (nodes) with a real-time data path that collects sensor-level information from sensors in autonomous vehicles and processes independent data to be used in high-level system elements. By abstracting the sensor level with the high-level system, the sensor level is not affected by any change that may occur in high-level algorithms. In order for these algorithms, which are created at a high level, to communicate smoothly with the sensor level, an application programming interface (API) that can broadcast to the real-time data path has been created. When the file is examined, it is seen that it does not contain a feature related to a central highway autopilot computer architecture (1) that covers and manages all the units at the sensor and system level in autonomous vehicles. In the patent document with the publication number U88260498B2, which is in the state of the art, a modular vehicle control architecture has been designed with a top-down approach. In the architecture, it is aimed to create a vehicle control system that observes the driver's behavior in addition to the vehicle's advanced driver support systems (such as lane tracking system, adaptive cruise control) and provides vehicle control according to this behavior. There is a structure that determines the priority of the command sent to the vehicle control according to the commands given by the driver for the driving support systems via the user interface and also by evaluating the driver's gas, brake and steering activities. In the patent document numbered USlO796204B2, which is in the state of the art, it is aimed to create the task, behavior and motion planning modules of the autonomous vehicle with a stochastic approach using artificial neural networks. Artificial neural network models handle the motion planning modules separately in order to determine the most suitable route, plan the behavior for the detected route and realize the planned behavior. Model trainings are done independently for each module within the determined scenarios. Each module contains various functional tasks within itself. The training of the models is done by including these functional tasks. The mentioned system does not have a structure that ensures that the necessary actions are taken so that its performance does not fall below the safe driving limits. This situation will lead to a low success in terms of system security and reliability. An architecture that enables navigation of autonomous vehicles at high speeds has been created. It is claimed that after the environmental data obtained from the sensors are passed through a map-based cost calculation, the obstacles that pose a collision risk are detected and it is stated that after the interpreted sensor data are passed through the fusion process, they are used in the movement plan and cruise speed calculation required for the autonomous vehicle to complete the planned route. The proposed method aims to reduce the computational load required for the determination of the areas where the autonomous vehicle can possibly travel. Purpose of the Invention The purpose of this invention; is a highway autopilot computer designed with a modular architecture that can work with different vehicle types, hardware and sensors, is responsible for running all software modules, connection and management of electronic sensors and equipment, communication between vehicles and peripherals, bringing data from external information sources and processing them, making a decision and then implementing them. Another purpose of the invention is to be a computer structure that is designed in accordance with distributed architecture in order to manage processing power and allows different modules to be loaded on different computers and to operate by scaling. In this way, the necessary calculations can be done remotely by cloud computing-based computers and can be transmitted to the vehicle with fast network access. Another purpose of the invention; It is a computer structure that systematically organizes the data flow in the extraction of static information such as the physical properties of the road, lane lines and traffic components, and the extraction of trajectory and maneuver estimates related to moving objects such as other vehicles and pedestrians in the environment. Another purpose of the invention is to handle the vehicle motion planning throughout the course of the autonomous vehicle from its starting position to its destination in a hierarchical structure within the "Decision Making and Planning" module. The "Decision Making and Planning" module consists of the sub-units "Route Planning", "Decision Maker" and "Motion Planning". The aim of the invention is to reveal the request-approval mechanism established between the "Decision Maker" and "Motion Planning" units and aimed at increasing vehicle safety. Another purpose of the invention is to handle the vehicle's ability to go to the target location without a collision in a modular structure within the autonomous driving architecture. The maneuver decisions that will ensure that the vehicle goes to the target location without a collision within the most appropriate time can be made in the "Decision Making and Planning" module according to the maneuver estimation of the surrounding vehicles and moving objects by using artificial intelligence-based methods such as reinforcement learning, deep learning, decision tree, fuzzy logic or one of the methods in the known state of the art. According to the maneuver decision taken, the "Motion Planning" sub-unit produces the reference motion information necessary for the autonomous vehicle to safely perform lane change or lane tracking actions. Another purpose of the invention is; is to create the "Motion Planning" sub-unit by calculating collision avoidance and obstacle avoidance in order for the autonomous vehicle to safely perform lane change or lane follow actions and according to the maneuver decision information determined by the "Decision Maker" unit. Another purpose of the invention is to evaluate the maneuver determined by the "Decision Maker" unit in terms of realizability, safety and comfort by the "Motion Planning" unit and to transmit it to the "Decision Maker" unit as feedback. Another purpose of the invention is to consider the system observation module, which will monitor system errors and ensure that relevant actions are taken when necessary, in a way that it will be connected to all modules and components while constructing the modular structure. In the system architecture, by designing the "connected information sources" module in accordance with the distributed architecture, a more powerful structure has been proposed for the correct management of the processing power on the vehicle and on the cloud. Another purpose of the invention is to handle the vehicle motion control in a modular structure within the autonomous driving architecture and to realize it in a hierarchical structure in the "Motion Control" module consisting of "Upper Level Motion Control" and "Lower Level Motion Control" sub-units. The handling of the vehicle motion control within the autonomous driving architecture provides a higher module. Thus, it will be ensured that strategic and tactical maneuver decisions are applied as input to the motion control and thus more optimal vehicle control outputs (Steering Reference, Acceleration Reference, Deceleration Reference) are obtained. Another purpose of the invention is; The aim of the invention is to test the autonomous vehicle's task, behavior and motion planning processes only with the outputs of the artificial neural network model by testing the maneuver decision taken within the request-approval mechanism with different algorithms in terms of realizability, safety and comfort. In addition, the "System Observation Module" monitors the current status of all modules in the system and takes the necessary actions to prevent the system performance from falling below safe driving limits. Thus, a much higher success is achieved in terms of system safety and reliability. Another purpose of the invention is; Rather than an interface system for autonomous vehicles, it includes a central highway autopilot computer that covers and manages all units at the sensor and system level in autonomous vehicles, provides communication with peripheral units, uses connected information sources for information sharing purposes or in computational power requirements, and is positioned in a programmable structure with a user interface. In this context, this computer, which manages autonomous driving tasks from a center and is responsible for their monitoring and execution, can easily integrate with other computers in the vehicle (i.e. electronic control units). The autonomous driving architecture is addressed with an innovative approach, aiming to increase the sustainability and reliability of the system. In the structure where the modules communicate with each other and this communication is observed with the system observation module, a structure has been created where the sustainability of the system can be ensured even in cases of partial failures that may occur in the sub-modules. The proposed modular architecture enables faster system development and software reuse with distributed components controlled by the central highway autopilot computer architecture (1). In addition, this architecture allows for rapid adaptation to possible updates that may be encountered during software development or maintenance by designing the components in isolation from each other. As a result of the combination of connected information sources with this modularity, it is possible to keep the endpoint under security. It allows the data groups belonging to each module to be encrypted and stored in cloud computing computers within the External Information Sources and to be used when needed. Explanation of the Figures Figure 1. Highway autopilot computer architecture Figure 2. Modular Architecture of Highway Autopilot Computer Figure 3. Highway Autopilot Computer Perception Module Figure 4. Highway Autopilot Computer Decision Making and Planning Module Figure 5. Highway Autopilot Computer Motion Control Module Explanation of References in the Figures For a better understanding of the invention, the correspondences of the numbers in the figures are given below: 1. Highway autopilot computer architecture Perception module Motion control module System observation module User interface module Related information source 9. Vehicle sensor . On-board computer 11. External information source 12. Autonomous vehicle 21. Analysis unit 22. Road perception unit 24. Estimation unit 31. Route planning unit 32. Decision making unit 33. Motion planning unit 41. Trajectory follower unit 42. Actuator controller unit 91. Radar 92. Lidar 93. Camera 94. GPS/IMU 101. In-vehicle sensor 1 1 1. Digital map 112. Linked units 113. Cloud computing 231. Object filter unit 232. Object tracking unit 233. Localization unit 234. Strip polynomial estimation unit 235. Sensor fusion unit Detailed Description of the Invention Figure 1 shows the schematic of the highway autopilot computer architecture (1). The highway autopilot computer architecture (1) consists of basic modules containing software loaded on multiple computers to perform different autonomous driving tasks. The relationship between these modules, which communicate with each other over an internal communication line, and the modular architecture of the highway autopilot computer (1), which provides a system view, are given in Figure 2. One of the components of the proposed architecture is the perception module (2). The perception module (2) aims to create a hierarchical environmental model by interpreting both the autonomous vehicle (25) information and the status and characteristics of the environment in which the vehicle is located. In this direction, the raw information obtained from the vehicle sensors (9) and vehicle computers (10) is analyzed in the analysis unit (21) and made usable by the relevant modules after being passed through the preprocessing step. Thus, sensor analysis and preprocessing processes are performed in this unit. From the sensor data that has been preprocessed, the components that constitute the driving scene such as roads, traffic signs and traffic actors that are important for driving activity are detected and the previously detected ones are continued to be followed. The traffic components detected by the individual sensors, namely vehicle sensors (9), are brought together at the next level with the sensor fusion unit (235) or any of the methods that can perform data fusion, and a consistent environmental model is obtained. At the highest level of the perception module (2), the maneuvers or trajectories of the traffic components in the environmental model are estimated according to the needs of the planning modules. The first layer of the perception module (2) includes the analysis unit (21). The vehicle internal information such as wheel speed, tire pressure and information about the vehicle's environmental status such as road conditions and traffic actors are obtained as raw data by vehicle sensors (9). The data obtained from the vehicle sensors (9) are transferred to different vehicle computers (10) in a coded form in accordance with the standard data transfer protocols used in the automotive industry (e.g. CAN protocol). In addition to this vehicle sensor (9) data, information about the vehicle's powertrain, chassis, information display, etc. must also be obtained with a specific data transfer method. Since all this information is transmitted in a coded form, the data must be decoded in order to be usable by higher-level perception modules (2). All of this information is used in the creation of vehicle information. In this context, thanks to the analysis unit (21), the vehicle interior information required for the units in the upper layers of the perception module (2) is obtained and sent to be used in higher-level models. The sensors used to perceive the vehicle's surroundings can sense wider areas or further than required for the targeted advanced driving support or autonomous driving activity. For example; For an active blind spot warning system on the highway, the vehicles in the lanes immediately adjacent to the lane in which the vehicle is driving are important, but the visibility of the sensors does not have to be limited to the lane lines. Information beyond what is required for driving activity may cause unnecessary processing load for higher perception modules, and therefore must be preprocessed. Traffic components must be selected from the sensor data that is pre-sorted. Since sensors such as radar (91) or lidar (92) make measurements according to the return of the electromagnetic waves they emit, they produce data in the form of point clusters rather than creating single object information about each traffic component. The data obtained with the camera (93) is in the form of two-dimensional or three-dimensional images obtained by preprocessing two-dimensional images by the camera (93). Traffic components are determined using various computer vision methods from this data in the form of point clusters, two or three-dimensional images. One of the basic components of driving activity is the road infrastructure. Road components such as lane lines and traffic signs constitute the road infrastructure. For safe driving, it is necessary to be able to perceive driving areas, lane lines and traffic signs. While all of the information about the road infrastructure can be obtained by processing the images obtained with the camera (93), some of the components can be detected individually using different sensors (e.g. lidar (92)). In addition, digital maps (1 l 1) can be obtained by means of maps from external information sources (1 1). This information can be matched with the instantaneous road physical properties perceived while creating the world model. These processes are performed by the road detection unit (22). The data obtained from the sensors contain a certain amount of noise. This noise affects the accuracy of information that is important for driving activities such as the positions and speeds of the detected traffic actors. Therefore, the kinematic status data of the traffic actors must be filtered. The data processing unit (23) that provides filtering, tracking and illusion of a sensor that allows the filtering of the noise contained in the data obtained from the sensors to a certain extent is included in the system. Depending on the characteristics of the sensor noise and the movement models of the traffic actors to be filtered, one of the filtering methods in the known state of the art is applied. These processes are performed by the object filter unit (231). The filtered information of the traffic actors is used to create a comprehensive world model. If the world model has been created in the previous cycles of the system, the newly obtained data is matched with the actors in the current world model, and not only the vehicle but also all traffic actors are tracked in the object tracking unit (232). Thus, a perception with a memory beyond instantaneous sensations is created. The creation of all processed sensor information of the vehicle, called vehicle information, and the information containing the internal status of the vehicle is performed by the localization unit (233). The conversion of the lane lines on the road into mathematical expressions for use in the Decision Making and Planning Module (3) is performed by the estimation unit of lane polynomials (234). In addition, in case of a conflict in the fields of view of the sensors according to the selected sensors and sensor placement architecture, the world model can be further improved by using the sensor fusion technique. Sensor fusion can be used to improve the information of traffic actors being tracked by separate sensors by combining them, or it can contribute to the improvement of the filtered information during the tracking phase and then be used to perform the tracking. These operations are performed by the sensor fusion unit (235). In the last layer of the perception module (2), an artificial intelligence-based decision estimation algorithm is used, which takes into account both the movements of the observed traffic actors/objects and their interactions with their environment. The observed traffic data is converted into time series and estimations are performed with a method selected from machine learning methods used in processing time series or artificial intelligence methods in the known state of the art and deterministic methods. These operations are performed by the estimation unit (24), which provides object trajectory and maneuver estimation. The decision-making and planning module (3) consists of computers that determine the route the vehicle will follow, take decisions to be used in the most effective way to follow the route, and create trajectories to be used in the implementation of these decisions, taking into account the driver (8) commands coming from the user interface and the data received from the perception module (2) (Figure 4). The route planning unit (31) is responsible for planning the route that will ensure that the vehicle reaches the target determined by the user in the most effective way. The route information received from the driver (8) via the user interface module (6) is processed in this unit and, as in current route planning, the user is presented with the most suitable route options to meet criteria such as distance, fuel economy, road options (toll or free roads) and driving time, as well as criteria suitable for the features that the vehicle will have in the future. Again, the driver (8) is given approval for the selected route among the suggested route options via the user interface module (6) and this selected route is recorded for use in the decision-making unit (32) for behavior planning. The decision-making unit (32) for behavior planning is responsible for following the vehicle route determined by the route planning unit (31) and making maneuver decisions to meet the safety, comfort and requests of the passengers. In highway driving, these maneuvers include non-mandatory lane changes for the purpose of ensuring the speed target selected by the driver (8) or the safe distance between vehicles or for performance improvement, and mandatory lane changes required for route tracking. These maneuvers can be performed with artificial intelligence-based methods such as reinforcement learning, deep learning, deterministic methods such as decision trees or fuzzy logic-based methods, and one of the methods in the state of the art. The lane change determined as a highway driving maneuver and the speed change decisions related to it are transmitted to the motion planning unit (33). In addition, a request-approval mechanism has been defined between the decision-making unit (32) and the motion planning unit (33) for behavior planning in order to increase vehicle safety. In this context, if the required movement actions to perform the desired maneuver (e.g. lane change) are not feasible or safe considering the current vehicle status and environmental conditions, the maneuver is not performed and the information about this situation is fed back to the decision-making unit (32) for behavior planning by the motion planning unit (33). In the motion planning unit (33), the reference planning information required for the autonomous vehicle (25) to safely perform the lane change or lane following actions is generated according to the maneuver decision information determined by the decision-making unit (32) for behavior planning. If the maneuver poses a risk for any reason as a result of the evaluation of the maneuver information, new maneuver information is requested from the decision-making unit (32) for behavior planning. According to the determined maneuver command, if the vehicle is to continue lane following, movement trajectories are created in a way that it will stay close to the center line of the lane it is in. During this time, the vehicle is also controlled against the risk of going out of its lane. If the vehicle is going to change lanes, the lateral trajectory points are calculated for each calculation cycle in accordance with the relevant lane change decision, considering the road slope and object information for the target lane and the risk status of the other lanes. During this calculation, the vehicle's collision with other objects in the environment and whether the acceleration movement that the vehicle will apply during the lane change will create a rollover risk are checked in the calculations. As a result of the risk analysis, feedback on the enforceability of the decision and whether it is appropriate in terms of safety is shared with the decision-making unit (32) for behavior planning. If the given decision does not meet the risk analysis criteria, the decision is not implemented and the previous movement mode is continued. The reference planning information for the control of the vehicle is sent to the motion control module (4) of the autonomous vehicle (25). The motion control module (4) is responsible for generating information to be shared with the vehicle computer (10) that controls the actuators to be applied to the vehicle for tracking the vehicle motion information obtained from the decision-making and planning module (3) (Figure 5). The trajectory follower subunit (41) is responsible for converting the reference planning information obtained from the decision-making and planning module (3) into the reference wheel angle and reference resultant acceleration information required for trajectory tracking. Due to the nature of vehicle dynamics, lateral and linear motion dynamics are interdependent and require a complex and non-linear system model for vehicle control. Considering this situation, the algorithm used for vehicle control performs both lateral and linear control of the autonomous vehicle together, thus providing effective and safe autonomous driving. In this way, while the autonomous vehicle is provided with the possibility of highway driving, it becomes possible to perform low-speed driving such as urban cruising or autonomous parking. The obtained reference steering angle and reference resultant acceleration values are sent to the actuator controller unit (42). The actuator controller unit (42) converts the reference wheel angle and reference acceleration values coming from the orbit follower unit (41) into actuator reference information to be realized in the vehicle's motion drive units with various calculation methods and transmits them to the vehicle computers (10). The autonomous vehicle (25) continues its acceleration or deceleration movement until it reaches the reference motion speed, and similarly, the steering reference calculation continues until the determined reference front wheel angle of the vehicle is obtained. Depending on the interface with the vehicle's drive units and the data transfer method, the information transfer of these actuator references is carried out in different ways. For example; gas pedal position, brake pedal position, acceleration required for acceleration / deceleration, torque required for acceleration / deceleration, steering angle, steering motor torque, etc. reference values are transmitted to the vehicle's drive units. The system observation module (5) is the module that has the responsibilities of observing and monitoring the highway autopilot computer architecture (l) components, generating alarms according to warning indicators, determining the necessary rescue actions and performing these actions when conditions require and monitoring performance. All signals produced and used by all modules in the system (perception module (2), decision-making and planning module (3), motion control module (4), user interface module (6) and related information sources (7)) or vital modules in the system architecture are accessible. Through this access, the current status of the modules and the system is monitored and the performance of the system is observed. The health of the hardware on which the system operates, the rate of used and ready-to-use processing power, the rate of usable random access memory, and hardware-related conditions such as fluctuations in power levels are also monitored. Thanks to the feedback mechanism that transmits the current status of the system to all modules, the relevant modules are enabled to receive error reactions or take individual rescue actions. The current status of the system is also transmitted to the driver via the user interface module (6). The explanations of the outputs of all artificial intelligence-based methods used to increase the confidence of the driver (8) and/or passengers in the system and to detect possible errors depending on the level of autonomy and to provide retrospective event logging and queries are created in the system observation module (5) and are also transmitted to the driver via the user interface module (6). While the artificial intelligence models used for this purpose can be developed with naturally explainable techniques; for artificial intelligence models that produce black box models where this is not possible, the known model explainability methods of the technique that reveal the factors affecting the models are also applied in the system observation module (5). Model explainability methods are used to separate the data that most dominantly affects the outputs of artificial intelligence models and present them to the driver. In this context, when determining the vehicle classes (motorcycle, passenger car, truck, etc.) in the images obtained with the camera (103), it can be specified and shown which parts of the image are taken into consideration in the classification. Error conditions in the system can be captured with two different approaches; l) Application of various error detection methods over the signals coming to the module and produced by the module: These methods can be the compatibility of the sampling frequency of the module sending the message with the sampling frequency of the incoming message, the message counter indicating that the connection with the module is continuous, and the incoming message being within the appropriate value range. 2) Processing of individual error messages coming from other modules: These error messages can be internal conformity checks performed by the relevant modules or error conditions in the automaton diagrams. In case the system performance falls below the safe driving limits and the modules cannot take the necessary rescue actions; By shutting down faulty modules, the system is allowed to operate with limited performance in service mode. An action plan is defined for all critical modules that the system monitors. When a problem occurs in the relevant module, the system implements the action plan. In the highway autopilot computer architecture (1), there can be more than one computer for each module and it is monitored and operated by the system monitoring module (5) with the principle of high availability. Thus, when a response cannot be received from any module's computer within the limits, the other computer will be switched to and the sustainability of the system will be ensured. Depending on whether the relevant module is critical or not, the module can be shut down, restarted or the entire chain of interconnected modules can be shut down. All errors in the system are recorded with module labeling and error codes determined specifically for the relevant module. The decisions taken by the system, the maneuvers it implements and the action plans it executes are all recorded instantly and stored as read-only without compromising time-data integrity. All this information can be stored as read-only in the highway autopilot computer architecture (l) and/or in the vehicle computer (10) for diagnostic and debugging purposes. The user interface module (6) allows the activation, deactivation and management features of the entire system or certain sub-components on the main screen. The user interface includes screens for the system settings. It hosts the user settings of the system components and offers the opportunity to change settings on a component basis. It offers the user the opportunity to easily change by switching between pre-created ready-made settings. The user interface module (6) displays the locations of other moving and stationary actors and objects determined by the perception module (2) and the moving actors and objects are also dynamically updated on the interface. Signals and information from traffic signals, imaging and connected information sources (7) are processed instantly and displayed on the interface. The explanation of the decisions made by the decision-making and planning module (3) can be accessed via the decision information detail. In this case, the reasons and confidence of the decision-maker are transparently conveyed to the user with visuals and infographics that the user can understand. In addition, the planned and realized movement trajectories by the motion control module (4) are continuously visualized on the user interface on the digital map (1 l 1) base obtained by the connected information sources module (7). Situations such as malfunctions, faulty operation, and difficulties in perception in physical equipment, especially sensors, are also conveyed to the driver (8) via warning screens based on the information provided by the system observation module (5). In case of occurrence of elements threatening the security of subcomponents of the system, the system notifies the user about the messages coming from the observation module (5) and transmits the emergency aid warning to the previously determined authorities (location, speed, status, road conditions, list of last performed maneuvers, etc.) using the communication channels of the connected information sources module (7). Within the scope of recording and backup; information such as decisions, maneuvers, errors and complete data are stored in a sequential and read-only and unchangeable manner with a time stamp. Methods such as block chain in the known state of the art can be used to store this information. It allows retrospective review of the data or playback in simulators. This information can be backed up to the connected cloud system (113) at certain intervals or sent to the systems of the relevant traffic authorities via connected units (112). Location-based information is obtained from digital maps (111) or high-resolution digital maps (111) and presented dynamically in the user interface module (6). Location-based information on maps includes information such as location, road type and class, slope, height, traffic signs and controllers with centimeter precision. Dynamic information about the location is provided through the digital map service. These may include traffic conditions in the state of the art, points of interest (e.g. parking lots, gas and charging stations, etc.), weather information if any, and all other information that will be provided through these services in the future. Positioning can be done in the case of using high-resolution maps. Lane information and precise location can be provided as input to the decision-making and planning module (3) that determines movement plans. Since map data provides instantaneous data like an external sensor, it can also be provided as input to the perception module (2) and the sensor fusion method. This information is used to more precisely determine the location of other mobile or stationary actors on the road. The continuous connection status of the vehicle is created by using the methods in the state of the art (e.g. 802.11p, etc.) and the methods in the state of the art for wide mobile networks (e.g. 4G, 5G and later) together or separately, especially the short-range communication protocols for instant information exchange between vehicles (V2V) and between vehicles and smart devices installed in the environment (V21). Information about the environment may include time-dependent events that occur instantly on the road. This information may include situations such as instant road conditions, accidents, closed lanes, maintenance work, special speed definitions depending on other conditions at tunnel / bridge entrances and exits, etc. Information about the environment can be transmitted from smart devices installed in the environment at position 0, as well as from other vehicles coming from the same lane or the opposite lane on the road using communication methods. The information at the moment, the situations ahead of the road are transmitted to the vehicle in this way and presented as input to the decision-making and planning module (3) for evaluation. Information from connected vehicle technologies can also serve as a sensor and create input to the perception module (2) for use in sensor vision methods. With the transmission of information formed on the road, it is possible to expect that the information will be transmitted from the front in order to provide convenience to vehicles with privileged passage (e.g. ambulance) on a lane basis and to change lanes accordingly. It is possible to prevent or reduce such situations by adjusting and optimizing the vehicle speed in advance in traffic lights and intersection control situations based on instantaneous situations on the road. In highway use, these communication methods in the state of the art can be used together for the convoy use of vehicles and in the case of a convoy, the maneuvers of the pilot vehicle can be monitored and imitated by the following vehicles. Application of the Invention to Industry This invention is a highway autopilot computer architecture system consisting of one or more computers that can operate on the vehicle with network access and can operate the entire process for autonomously making and executing decisions in each cycle, enabling the land vehicle to move autonomously on highways and consists of units and parts applicable to industry. The invention is not limited to the above example applications, and a person skilled in the art can easily reveal different applications of the invention. These should be evaluated within the scope of the protection claimed by the invention.TR

Claims (1)

1.SYSTEMS. It is a computer architecture system, its feature is; at least one perception module (2) that provides the creation of a hierarchical environmental model by interpreting both the vehicle information and the status and characteristics of the environment in which the vehicle is located; at least one decision-making and planning module (3) that provides the determination of the route to be followed by the vehicle by considering the driver commands coming from the user interface and the data received from the perception module (2), the taking of decisions to be used in the most effective way to follow the route and the creation of trajectories to be used in the realization of these decisions; at least one motion control module (3) that is responsible for the creation of information to be shared with the vehicle computer (10) that controls the actuators in order to be applied to the vehicle in order to follow the vehicle movement information obtained from the decision-making and planning module (3); observing and following the highway autopilot computer components, generating alarms according to warning indicators, determining the necessary rescue actions and It is characterized by being an autonomous autopilot computer architecture system having at least one system observation module (4) which is responsible for performing these actions when conditions are necessary and monitoring the performance, at least one user interface module (6) which provides the display of the locations of other mobile and stationary actors and objects determined by the perception module (2) and the dynamic updating of the moving actors and objects on the interface, and at least one connected information source (7) which allows the transfer of the current information and the situations ahead of the road to the vehicle in this way and to present it as input to the decision making and planning module (3) for evaluation. . It is a perception module (2) mentioned in claim 1 and its feature is; It is located in the first layer of the module and the vehicle; It is characterized by having at least one analysis unit (21) that obtains internal vehicle information such as wheel speed, tire pressure and information about the environmental status of the vehicle such as road condition and traffic actors as raw data from sensors and provides the analysis of these data; at least one road perception unit (22) that provides the creation of the road infrastructure from components belonging to the road such as lane lines and traffic signs; at least one data processing unit (23) that provides the filtering of the noise contained in a certain amount of the data obtained from the sensors, the filtering, tracking and fusion of the sensors; at least one estimation unit (24) that provides the use of an artificial intelligence-based decision estimation algorithm that takes into account both the movements of the observed traffic actors / objects and their interactions with their environment, located in the last layer of the module. It is a data processing unit (23) according to claim 2 and its feature is; It is characterized by having at least one object tracking unit (232) that enables tracking not only the vehicle but also all traffic actors by matching the newly obtained data with the actors in the current world model, at least one localization unit (233) that ensures the creation of all information including all processed sensor information of the vehicle and the internal state of the vehicle, at least one lane polynomial estimation unit (234) that enables the translation of lane lines on the road into mathematical expressions to be used in decision making and planning algorithms, at least one sensor fusion unit (235) that can be used to combine and improve the information of traffic actors being followed by separate sensors and contribute to the improvement of the information filtered beforehand during the tracking phase and then to perform the tracking. It is a decision making and planning module according to claim 1 and its feature is; It is characterized by having at least one route planning unit (31) which is responsible for planning the route that will ensure that the vehicle goes to the target determined by the user in the most efficient way, at least one decision-making unit for behavior planning (32) which is responsible for following the vehicle route determined by the route planning unit (31) and making maneuver decisions to meet the safety, comfort and requests of the passengers during this time, and at least one motion planning unit (33) which provides the production of reference planning information necessary for the autonomous vehicle (25) lane change or lane following actions to be performed safely according to the maneuver decision information determined by the decision-making unit (32) for behavior planning. 5. A motion control module (4) according to claim 1, and its feature is; - It is characterized by having at least one orbit follower unit (41) which is responsible for converting the reference planning information obtained from the decision making and planning module (3) into the reference wheel angle and reference resultant acceleration information required for orbit tracking, - At least one actuator controller unit (42) which enables the reference wheel angle and reference acceleration values coming from the orbit follower unit (41) to be converted into actuator reference information to be realized in the vehicle's motion drive units 10 by various calculation methods and transmitted to the vehicle computers (10).