CN119271485A - A model for online safety monitoring and safety management of intelligent vehicles - Google Patents

Abstract

The invention discloses an intelligent automobile online safety monitoring and safety management model, which is used for forming a safety brain MEIA model prototype by monitoring the necessary functions of safety assurance such as M, E, I, A after improvement and the like. The monitoring of various dynamic and static targets, traffic flows and other states in the operation design domain ODD and the monitoring of AVSO, AVPO, AVCO and other module states also comprises the monitoring of AVMO self-states, self-monitoring, self-evaluation, active safety intervention and self-processing of self-safety states, so that essential equipment for intelligent automobile safety is rearranged in principle.

Description

Intelligent automobile online safety monitoring and safety management model

Technical Field

The invention relates to the field of intelligent automobiles, in particular to an intelligent automobile online safety monitoring and safety management model.

Background

Sensor defects and AI algorithm defects exist in the current automatic driving system, and the defects cannot fully consider the influence of the surrounding environment on safe driving. Furthermore, current architectures suffer from inherent drawbacks in terms of structure and function. In order to ensure traceability of the whole driving safety process, traceability of process management and definiteness of responsibility attribution;

Therefore, it is necessary to redesign the architecture, and propose a security monitoring operation (Monitor Operation) module and a security redundancy architecture to make up for the structural drawbacks of the current sense-program-control serial architecture.

Disclosure of Invention

In order to overcome the defects in the prior art, the embodiment of the invention provides an intelligent automobile online safety monitoring and safety management model, which solves the problem of the structural defects of the current perception S-planning P-control C serial architecture.

This model takes into account the mandatory and safety-related interactions of policy/regulations. Through redesigning architecture and real-time monitoring, evaluation and intervention, the safety and operability of the autopilot system are fully supervised to ensure safety, traceability and integrity. The design purpose of the model is to ensure driving safety and solve the safety problem of the intelligent automobile.

In order to ensure that the Level 3 and above autopilot system (AD SYSTEMS) is safe when put on the market, a set of socially acceptable, technically reliable and complete safety assurance technology system must be established. This constitutes a general goal of security assurance in the development of AD systems. Under this framework, security is considered as a legal concept for AD systems.

In order to achieve the purpose, the invention provides the following technical scheme that an intelligent automobile online safety monitoring and safety management model, namely a safety brain MEIA model, comprises the following functional characteristics:

The monitoring function covers three key aspects of scene awareness, driving awareness and automatic driving awareness. In scene awareness, knowledge of scene clarity, scene complexity, and scene intensity is included. Driving awareness relates to awareness of road conditions, vehicle conditions, driving conditions, and occupant/cargo conditions. While autopilot awareness includes awareness of sensor status, operational status, driving intent execution, and current driving intent.

The assessment function covers four key aspects of scene risk assessment (SRE), compliance assessment of intent (Compliance Evaluation of Intention), security assessment of intent (Safety Evaluation of Intention), and behavioral security assessment (Safety Evaluation of Behavior). Specifically, scene risk assessment includes ODD (Operational Design Domain) identification, risk intuition judgment, and security entropy evaluation. The evaluation of intent then includes a double consideration of compliance and safety, concerning compliance evaluation of traffic regulations and safe driving rules, respectively. In addition, behavioral safety assessment focuses on safety assessment of existing behaviors of the vehicle, thereby ensuring that the behavior of the automated driving system in various situations meets safety standards.

The intervention functions include Constrained decision (Constrained Plan) and Safe parking (Safe Stop). Constraint decision by determining a safety decision boundary, a series of constraint conditions are provided for an autopilot system to ensure that the behavior of the system is within a safe range. These constraints may include avoidance of obstacles, adherence to traffic rules, speed limitation, and the like. Under the condition that the vehicle needs to be parked, the safe parking meets the controllable parking condition and can implement emergency parking. Through reasonable parking decision and control instruction, the safety parking function can ensure the safety and stability of the vehicle in the parking process.

The post-treatment functions comprise key aspects such as security event data arrangement and uploading, sending alarm and rescue signals, high-voltage electric safety management, low-voltage electric safety management, passenger escape support and the like. Specifically, the arrangement and uploading of security event data encompasses the arrangement, uploading and background retrieval of behavioural security event data and incident data. In transmitting alarm and rescue signals, important techniques include accident assessment, ensuring that accident conditions can be assessed rapidly in the event of an accident. High voltage electrical safety management includes efficient management of power cells, high voltage electrical power outage management, and condition monitoring. Low voltage electrical safety management involves the management of a low voltage electrical power supply to ensure the safety of the system in terms of low voltage electrical power. In the aspect of passenger escape support, the technology covers comprehensive management of states of door locks, door and window glass, fans, safety belts and the like so as to support the safety escape of passengers in emergency situations.

The invention has the technical effects and advantages that:

The monitoring, evaluating, intervening and post-improvement safety guarantee functions are integrated through the modules so as to realize the whole-course supervision of SOTIF (safety, operability, traceability and integrality). By constructing such a model prototype of the safe brain (MEIA), an omnidirectional protection can be provided for safe driving.

Drawings

FIG. 1 is a block diagram of an intelligent car on-line safety monitoring and safety management model structure according to the present invention.

FIG. 2 is a schematic flow chart of an intelligent automobile online safety monitoring and safety management model according to the present invention.

FIG. 3 is a block diagram of an intelligent vehicle on-line safety monitoring (AVMO) according to the present invention.

FIG. 4 is a simulation diagram of an example of online safety monitoring and safety management model monitoring of an intelligent automobile according to the present invention.

FIG. 5 is a simulation diagram of an intelligent vehicle online safety monitoring and safety management model evaluation according to the present invention.

FIG. 6 is a simulation diagram of the intelligent automobile on-line safety monitoring and safety management model intervention according to the invention.

FIG. 7 is a simulation diagram of an intelligent vehicle on-line safety monitoring and safety management model according to the present invention.

Detailed Description

The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

1-7, Referring to FIG. 1, there is shown a structural block diagram of an intelligent car on-line safety monitoring and safety management model of the present invention, the safety brain MEIA model includes a series of technical penetration actions of monitoring M, evaluating E, intervening I, and post-treatment A;

the Monitoring is used for acquiring the overall road surface condition, predicting the running condition of the vehicle and defining dynamic constraints, and extracting key information;

the Evaluation is used for evaluating scene risk, intention compliance, intention safety and behavior safety;

The Intervention is used for determining a safety decision boundary, providing a series of constraint conditions and implementing reasonable parking decision and control instructions;

The post-processing AFTERTREATMENT is used for recording and uploading data, constructing dangerous situations in an automatic driving scene, and carrying out post-processing on accidents or accidents.

Referring to fig. 2, a schematic flow diagram of an intelligent automobile online safety monitoring and safety management model is shown, in the running process of a safety brain MEIA model, a monitoring function covers three key aspects of scene awareness, driving awareness and automatic driving awareness, comprehensive cognition driving scenes, scene definition, complexity, severity, road surface condition, vehicle condition and the like, cognition on sensor states, operation execution states and the like, an evaluation function covers four key aspects of scene risk evaluation, intended compliance evaluation, intended safety evaluation and behavior safety evaluation, and ensures compliance with traffic regulations and safety driving rules, an intervention function comprises constraint decision and safety parking, and ensures that system behaviors are in a safety range by setting decision boundaries and implementing safety parking, and a post-processing function comprises key aspects of safety event data arrangement and uploading, sending alarm and rescue signals, high-voltage electric safety management, post-voltage electric safety management, passenger escape support and the like, and ensures timely post-processing of a system after an accident occurs.

Referring to fig. 3, the vehicle is monitored according to AVMO module functions of the united kingdom PAS standard, one type of behavior is monitoring of an Operational Design Domain (ODD), including geographic location, road characteristics, weather illumination, etc., and the second type of behavior is monitoring of vehicle operation, including sensing, decision making, control states. Three types of behavior are self-monitoring, including identifying vehicle faults, and four types of behavior generate parking instructions requesting AVPO to park as soon as possible based on reduced risk of passengers and other traffic participants based on information derived from the beacon system. The five types of behaviors generate parking control, and can directly control the vehicle to park as soon as possible.

Referring to fig. 4, the specific functions of the scene awareness SA are scene clarity cognition, scene complexity cognition, and scene intensity cognition, air conditions (rain, snow, fog, light, wind), traffic guidance signals, road markings, communication connection capability, road network, traffic participation vehicles, weak traffic participants, directions, speeds, distances of various traffic participants, and the like, which are the M-monitoring modules in the MEIA model. By monitoring the above elements, regional and climatic difference conditions can be recognized, on-line self-learning can be realized, and scene definition and element integrity can be enabled by combining a communication connection means. The complexity of the scene is estimated according to general traffic rules and social constraint habits by combining the traffic types, the density, the relative speed, the relative distance and the like, a quantized thermodynamic diagram index is formed, the change degree of a trafficable domain of the scene is judged by combining the change rate of 1D-2D TTC, and a quantized scene intensity expression index is formed and used as an evaluation index for measuring the adaptability of the AV vehicle to the scene under different ODDs. Specific functions of the driving awareness DA include cognizance of road conditions, vehicle conditions, driving conditions, and occupant/cargo conditions. The method comprises the steps of comprehensively sensing road conditions by adopting multi-dimensional modes such as stereoscopic vision, radar reflection, chassis suspension vibration points and the like, taking the road conditions as a basis for adjusting the speed of a vehicle, pre-judging abnormal noise sources and functional safety of the abnormal noise sources in the on-line running conditions of the vehicle, defining a safety control boundary of dynamic constraint of the chassis of the vehicle, pre-judging the relative position and change area of a current working point in the whole safety domain, timely predicting and pre-controlling, monitoring and safety warning the riding gesture of passengers, evaluating the influence form of passengers/cargoes on the dynamics of the vehicle, and evaluating riding comfort and behavior range of the influence of the dynamics of the vehicle. The autopilot awareness AA specific functions are the cognitive sensor state, the operation execution state, the driving intention execution situation, and the current driving intention. The sensors can perform periodic and online self-diagnosis, self-calibration and self-calibration according to absolute calibration objects, mutually verify, recognize track following deviation, TTC deviation, distance deviation and the like, recognize current transverse and longitudinal driving intention, evaluate planning capacity and re-energize.

Referring to FIG. 5, E-assessment in a model of a safety brain MEIA, ODD identification from M-monitored data, risk intuitionistic judgment, safety entropy assessment. And carrying out driving intention compliance assessment based on the AA perception information and intention, carrying out safety driving rule compliance assessment based on SA, DA and AA data, and carrying out safety assessment based on the DA and AA data.

Referring to fig. 6, I-intervention in the model of the safety brain MEIA, safety traffic envelop boundary validation is performed based on SA, DA, AA data in vehicle constraint decisions, boundary constraints are provided for safety decisions, traffic environments environment envelop and safe operation envelop of safety traffic are evaluated and boundary determined, and boundary output is provided. When the safe parking is needed, if the controllable parking condition is met, the current vehicle state, ODD scene state, vehicle goods and passenger state, even the influence on traffic flow in the minimum risk position process, are considered, the optimal parking safety strategy is selected, when the danger is unavoidable and upgraded, and the controllable parking condition is not met, the emergency parking measure is adopted, and the safe parking is carried out by combining a plurality of ISO/ECE/NHTSA rules and scene perception results.

Referring to fig. 7, the a-goodwill in the safety brain MEIA model includes the data arrangement and uploading of safety event, the sending of alarm and rescue signals, the high voltage electrical safety management and the passenger escape support. The method comprises the steps of managing and uploading safety event data, managing and uploading the safety event data, managing, recording, uploading and searching related events according to the fluctuation degree of driving safety indexes, evaluating the rationality of driving strategies and the rationality of reaction capacity, sending alarm and rescue signals for evaluating accident states of passengers inside and outside a vehicle and vulnerable road users, sending alarm and rescue signals in time, judging possible positions of high-voltage electric leakage sources according to empirical data such as collision speed, angle and the like by high-voltage electric safety management, timely controlling or orderly cutting off the possible positions, timely arranging evacuation traffic flows according to collision damage estimation and vehicle positions, driving and moving to the safety ground, ensuring basic communication, necessary sensing and available signal lamps by low-voltage electric safety management, supporting intelligent control of door locks for passenger escape by passenger escape, necessary breaking of window glass, automatic management of fan poisoning gas (including necessary environmental perception before collision), automatic release of a safety belt or pre-tightening for preventing secondary collision and the like.

Claims (5)

1. A smart car online safety monitoring and safety management model, characterized in that: the model includes monitoring M, evaluation E, intervention I, and aftermath A, and obtains a safety brain MEIA model of self-monitoring, self-evaluation, active safety intervention and self-processing of its own safety status; The monitoring is used to monitor the various dynamic and static targets, traffic flow status, and the status of modules such as AVSO (autonomous driving vehicle perception system), AVPO (autonomous driving vehicle path planning system), and AVCO (autonomous driving vehicle control system) in the operation design domain (ODD) by AVMO (autonomous driving vehicle safety brain); The Evaluation is used by AVMO (autonomous vehicle safety brain) to perform low-level self-monitoring and self-assessment of its own safety status. It also includes the assessment of some operational design domain (ODD) boundaries to ensure the safety and effectiveness of the system in various situations. The Intervention is used to ensure the vehicle's own safety status by proactively intervening in safety to deal with various situations based on low-level self-monitoring and self-assessment by AVMO (autonomous vehicle safety brain); The aftertreatment is used for the ability of AVMO (autonomous driving vehicle safety brain) to self-monitor and self-process its own safety status, and also includes the follow-up processing after an accident or unexpected event. 2. According to claim 1, an online safety monitoring and safety management model for smart cars is characterized by: the monitoring is to conduct comprehensive cognition of the clarity, complexity and intensity of driving scenes, improve the ability of autonomous driving vehicles to cope with different scenes, and at the same time, enhance the driving awareness and safety performance of vehicles through multi-dimensional perception of road conditions, prediction of vehicle operating conditions and definition of dynamic constraints. In addition, the awareness and safety performance of the autonomous driving system are improved through self-diagnosis and calibration of sensors, redundant safety mechanisms, and real-time monitoring and evaluation; Specifically, it can firstly perceive and evaluate air conditions (rain, snow, fog, dust, light, wind) to understand the impact of different air conditions on sensors and perception models. Secondly, it can also understand traffic signals and signs, including new types of signals and regional differences. It also has the ability of extended understanding and semantic reasoning to cope with the differentiated road markings and signs in different regions. In order to achieve the transmission and understanding of information, communication connection means are combined to realize the information transmission between vehicles, including the transmission of V2V information. In addition, the ability to identify different types of traffic participants, such as vehicles, pedestrians, and bicycles, and understand their intrinsic properties, such as the type of goods transported and their explosiveness, can also estimate the size and movement ability of traffic participants, providing more accurate information for the model. In terms of traffic flow, the density and speed of traffic flow are identified to provide support for the assessment of traffic conditions, and the complexity of the scene is quantified to form a heat map indicator to intuitively express the complexity of the scene, and this indicator is used as one of the capability evaluation indicators of AV vehicles. 3. According to claim 1, the online safety monitoring and safety management model for smart cars is characterized in that: the evaluation includes scenario risk assessment, intention compliance assessment, intention safety assessment and behavior safety assessment to improve the safety performance of the autonomous driving system; Specifically, the results of SA (scenario awareness), DA (driving awareness) and AA (autonomous driving awareness) are integrated to conduct a comprehensive assessment of the current scenario to determine whether it exceeds the response range of the autonomous driving system and calculate the scenario's danger index. This involves the judgment of scenario risks and the assessment of resilience boundaries, the assessment of intention compliance and safety, the formulation of a safe driving rules compliance assessment strategy, and the assessment of behavioral safety performance bandwidth. At the same time, the compliance of driving intentions is assessed using the perception information and intentions provided by AA. In addition, based on the data from SA, DA and AA, an assessment of safe driving rules compliance is conducted to ensure that the system meets safety requirements during driving. 4. According to claim 1, the smart car online safety monitoring and safety management model is characterized in that: the intervention is based on evaluating and determining the traffic environment and safe operation range, combining rules and scene perception to achieve safety constraint decision-making and safe parking; Specifically, based on SA (scenario awareness), DA (driving awareness) and AA (autonomous driving awareness) data, the boundaries of the safe driving envelope are confirmed to provide boundary constraints for safety decisions. When confirming the boundaries, the current vehicle status, ODD (operational design domain) scenario status, and the status of the vehicle's cargo and passengers are considered. In addition, the impact on traffic flow during the minimum risk position process must also be considered. When the danger is unavoidable and escalates, and the controllable parking conditions are not met, emergency parking measures are taken. This measure is intended to ensure safety in emergency situations and reduce potential risks. 5. According to claim 1, a smart car online safety monitoring and safety management model is characterized by: the aftertreatment, recording and uploading data, the system can construct dangerous situations in autonomous driving scenarios and evaluate their distribution inside/outside the operational design domain, using these new scenario cases to improve the safety of the autonomous driving system and ensure the escape support of the occupants; Specifically, the data elements of the ADS data recorder based on VMAD-SG3 can store incident-triggered data for efficient reconstruction of standard format scenes and automatic assessment of accident severity. Data analysis and post-processing can enhance corner case discovery and model updates, and form new safety rule recommendations. In terms of accident response, based on empirical data such as collision speed and angle, the possible location of the high-voltage electrical leakage source can be determined, and corresponding control or orderly cut-off measures can be taken. At the same time, based on the collision damage estimate and vehicle location, timely arrangements can be made to evacuate traffic flow and drive the vehicle to a safe ground. To ensure the safety of passengers, intelligent door lock control to support passenger escape, window glass breaking when necessary, automatic management of fan poisoning gas (including environmental perception before collision), and automatic release of seat belts or pre-tightening of seat belts can be implemented to prevent secondary collisions. After an accident or accident, the safety brain of the autonomous driving system can self-monitor and process, and perform necessary data storage, uploading and feedback to minimize the occurrence of unknown risks.