WO2021120575A1

WO2021120575A1 - Real-time simulation and test method for control system of autonomous driving vehicle

Info

Publication number: WO2021120575A1
Application number: PCT/CN2020/098279
Authority: WO
Inventors: Jiafeng CHAI; Pingliang HAN; Mingcong LI; Zhishan LI
Original assignee: Suzhou Zhijia Science & Technologies Co., Ltd.
Priority date: 2019-12-16
Filing date: 2020-06-24
Publication date: 2021-06-24
Also published as: CN110888417A

Abstract

An example of the present disclosure includes a real-time simulation and test method for a control system of an autonomous driving vehicle. The example method comprises the following blocks. Providing a vehicle dynamics model (S100). Carrying out a simulation test and a field test (S200). Providing an input scenario library (S300). Providing a vehicle dynamics model library (S400). Designing a control algorithm and verifying a stability of the control algorithm (S500). Using a Monte Carlo method to carry out a traversal test and determining whether analysis simulation results of the automated traversal test meet robustness and stability requirements of the control system (S600).

Description

REAL-TIME SIMULATION AND TEST METHOD FOR CONTROL SYSTEM OF AUTONOMOUS DRIVING VEHICLE

TECHNICAL FIELD

The present disclosure belongs to the field of autonomous driving of motor vehicles, and in particular to a real-time simulation and test method for a control system of an autonomous driving vehicle, such as a truck.

BACKGROUND

Autonomous driving may be implemented at various different levels of automation. The first level is driver assistance, which can achieve transverse (steering) or longitudinal (throttle brake) control. The second level is partially autonomous driving, which can achieve transverse and longitudinal controls. The third level is conditionally autonomous driving, which can realize autonomous driving under certain conditions. The fourth level is highly autonomous driving, which can realize autonomous driving on most road sections, but need a driver to take over driving in some special scenarios. The fifth level is completely driverless, that is, no manual takeover is needed.

As higher levels of autonomous driving are implemented, this places ever greater commands on the vehicle control system. For example, the design of advanced autonomous driving of L4 (level 4) or higher have higher requirements on the control system than the current L2 (level 2) assisted driving. The traditional assisted driving LKA (lane keeping assist) and ACC (full-speed adaptive cruise control) designs pay more attention to the comfort of passengers. Due to the limitation of the use scenarios, the need of a driver to take over driving and other factors, the design of the control system does not need to consider too much stability and robustness requirements. At some control system design boundaries, such as under operation conditions of emergency obstacle avoidance or wet roads, autonomous driving is not allowed. However, advanced autonomous driving which is to operate with limited or no driver assistance has to cover a more comprehensive range of scenarios and operation conditions. This puts ever higher requirements on the robustness and reliability of the design of the vehicle control system. It is a challenge to ensure that the control system is designed to be stable under different operation conditions and able to prevent the vehicle from losing control, especially as the vehicle ages and wears and the operating conditions change.

SUMMARY

Afirst aspect of the present disclosure provides a real-time simulation and test method for a control system of an autonomous driving vehicle, the method comprising the following blocks:

S100: providing a vehicle dynamics model;

S200: using a control algorithm to carry out a simulation test and a field test to determine whether the field test and the simulation test have reached the same performance, if yes, then proceeding to the next block S300; and if not, returning to the previous block S100;

S300: providing an input scenario library that covers a plurality of driving scenarios;

S400: providing a vehicle dynamics model library including the vehicle dynamics model established in S100;

S500: designing a control algorithm and verifying a stability of the control algorithm, wherein the design of the control algorithm and the verification of the stability of the control algorithm are performed according to a vehicle dynamics model under normal operation conditions;

S600: using a Monte Carlo shooting method to carry out an automated traversal test and determining whether analysis simulation results of the automated traversal test meet robustness and stability requirements of the control system, if yes, then proceeding to the next block S700; and if not, returning to block S500; and

S700: carrying out a field test, and if the design requirements are not met, then returning to block S500 to produce a new design for the control algorithm. However, if the design requirements are met, then the design of the control algorithm may be considered to be completed.

The above method helps to design a control system which may be stable under a large variety of operating conditions and may shorten the development cycle of the control system. Each block of the method may be executed by a computing device including a processor for example by executing machine readable instructions stored on a non-transitory computer readable storage medium.

A second aspect of the present disclosure provides a non-transitory computer readable storage medium storing machine readable instructions which are executable by a processor to perform the method of the first aspect of the present disclosure.

Further features and aspects of the present disclosure are provided in the dependent claims.

BRIEF DESCRIPTION OF THE DRAWINGS

Fig. 1 is a schematic diagram of a real-time simulation and test method for a control system of an autonomous driving vehicle according to an example of the present disclosure.

Fig. 2 is a working flowchart of an embodiment of a real-time simulation and test method for a control system of an autonomous vehicle according to an example of the present disclosure.

Fig. 3 is a diagram of an inclusion relationship of a simulation scenario set of a real-time simulation and test method for a control system of an autonomous driving vehicle according to an example of the present disclosure.

Fig. 4 is a diagram of the inclusion relationship of vehicle models of a real-time simulation and test method for a control system of an autonomous driving vehicle according to an example of the present disclosure.

DETAILED DESCRIPTION OF PREFERRED IMPLEMENTATIONS

A real-time simulation and test method for a control system of an autonomous driving truck according to the present disclosure is illustrated below with reference to Figs. 1-4. It should be noted that the embodiments described below with reference to the drawings are exemplary and aim to explain the present invention, and cannot be interpreted as a restriction on the present disclosure.

The working principle of the present disclosure is shown in Fig. 1, which is mainly divided into several parts which may be implemented in parallel:

1. control input scenario library;

2. control algorithm node;

3. dynamics node;

4. Monte Carlo automated test evaluation tool.

The focus of the present disclosure is on how to efficiently verify the design, stability and robustness of an autonomous vehicle control system. The approach described below may help to verify the stability of the control algorithm throughout the life cycle of the vehicle, improve the stability and robustness of the design of the control algorithm, and shorten the algorithm development cycle. Further, the approach described below allows much of the design and verification to be carried out by computer simulation. By this approach, the number of field tests (collecting data from automated driving of the vehicle in the field) may be reduced, thus saving manpower and material resources and reducing the risk of accidents which may occur during such field tests.

First, a relatively accurate vehicle dynamics model may be established. The vehicle dynamics model may include an input error disturbance modeland an output error noise model. An input error disturbance model is a model which simulates unusual or abnormal inputs and thus can test the stability of the vehicle dynamics model over a wider range of inputs. An output error noise model may add noise to outputs from the vehicle dynamics model to simulate such noise in the real world.

A simulation platform may be verified and optimised to achieve a better consistency with the field test. Then, on this basis, a control input scenario library and a scenario library for the vehicle dynamics model are established. These libraries may be enhanced or perfected by incorporating a wide range of inputs or scenarios optimized such that it is enough to cover the changes of model parameters and the scenarios that can be experienced throughout the life cycle of the vehicle. Finally, the Monte Carlo automated test evaluation tool, that is, a Monte Carlo shooting method, may be used to carry out an automated test of the control algorithm covering all the combinations of the scenario libraries and the dynamics library, and the stability and robustness of the control algorithm are judged to determine whether the design requirements are met.

An example of a real-time simulation and test method for a control system of an autonomous driving vehicle, such as a truck, is as shown in Fig. 2. The real-time simulation and test method comprises the following blocks:

S100: providing a vehicle dynamics model; and modeling vehicle dynamics under normal operation conditions. To make the simulation more accurate, an error interference model at the input and a noise model at the output may be additionally provided.

S200: using a control algorithm to carry out a simulation test and a field test, and repeatedly performing iterative optimization to determine whether the field test and the simulation test have reached the same performance, if yes, then proceeding to the next block; and if not, returning to the previous block S100, and re-establishing a vehicle dynamics model (which may include an error interference model and a noise model) .

S300: establishing an input scenario library that covers a plurality of driving scenarios. The input scenario library may be enhanced to include sufficient driving scenarios to cover scenarios that may be experienced in the lifetime of the vehicle. The scope of the scenario library is shown in Fig. 3. In general, a simulation scenario set includes field scenarios, including but not limited to weather, operation condition, and function scenarios, for example:

1. different acceleration-cruise-deceleration scenarios;

2. vehicle following scenarios at different speeds;

3. AEB (Autonomous Emergency Braking System) scenarios;

4. vehicle following scenarios;

5. front vehicle emergency cut-in scenarios;

6. scenarios with different turning radii at different speeds; and

7. emergency obstacle avoidance scenarios.

S400: establishing a vehicle dynamics model library. An example scope of the model library is shown in Fig. 4. In general, the simulation scenario set may include scenarios where the parameters of the real vehicle are changed throughout the life cycle under different operation conditions. In general, the vehicle dynamics model library may include the following situations:

1. the dynamics model under normal operation conditions;

2. a dynamics model for wet roads with different adhesion coefficients;

3. boundaries of a tire dynamics model caused by tire wear, etc. under different loads; and

4. performance boundaries of dynamic models of a brake, a steering wheel wire control system, an engine, etc. throughout the life cycle of the vehicle.

S500: designing and verifying a control algorithm. The design of the control algorithm and the verifying of a stability margin of the control algorithm may be performed according to a dynamics model under normal operation conditions. In some examples, the control algorithm in S500 may be an amended version of the control algorithm of S200. Thus designing the control algorithm in S500 may comprise optimizing, refining or adjusting the control algorithm of S200. For instance the control algorithm may be optimized to make it more suitable for simulation.

S600: using a Monte Carlo shooting method to carry out an automated traversal test to determine whether analysis simulation results meet the robustness and stability requirements of the control system, if yes, then proceeding to the next block; and if not, returning to block S500. The idea of Monte Carlo shooting is used, and for various combinations of all the scenarios and the dynamics models, the control algorithm is used for automated simulation and verification.

S700: carrying out a field test, and if the design requirements are met, the design of the control algorithm is completed, otherwise, returning to block S500 where a new design of the control algorithm is generated. For example on returning to block S500, the control algorithm may be re-designed.

The real-time simulation and test method for a control system of an autonomous driving truck of the present invention can verify the stability of the control algorithm throughout the life cycle of the vehicle, improve the stability and robustness of the design of the control algorithm, and greatly shorten the algorithm development cycle. Moreover, most of the design and verification of the control algorithm are transplanted to simulation and are completed in the simulation, which can save a lot of manpower and material resources and reduce the risk of accidents in field tests.

The above described real-time simulation and test method may be implemented using one or more computing devices. The computing devices may include at least one processor and at least one memory and may have access to a non-transitory storage medium. For example the method may be implemented by a processor reading machine readable instructions stored on a non-transitory machine readable storage medium. The simulations may be implemented entirely on the computing device. The field tests may include a computer system autonomously driving the vehicle and storing and processing data collected by sensors of the vehicle and/or other monitoring devices during the test driving of the autonomous vehicle.

The above method provides a real-time simulation and test method for a control system of an autonomous driving truck, which, may achieve a good consistency between the field test and the control simulation. This good consistency may be achieved by establishing a dynamics model, an interference model and a noise model with sufficiently high accuracies. A dynamics model library and an input scenario library may also be provided. In this way the test control system can cover many or all scenarios, consider many or all dynamics boundaries and incorporate a large number of factors into the design stage of a control algorithm so that the control algorithm can be thoroughly verified. This approach may shorten the development cycle and reduce the number of field tests required.

The above described system and method may, in certain implementations, impart one or more of the following beneficial effects:

1. The robustness of the control algorithm may be fully verified, and the stability of the design of the control system increased.

2. The research and development cycle may be shortened, and the cost reduced.

3. A large number of field tests may be omitted and the accident rate of field test reduced, saving a lot of manpower and material resources.

4. As most of the algorithm design is not affected by the test environment and region, the iteration efficiency may be improved.

The above embodiments are used to explain and describe the present disclosure, and do not to limit the present disclosure. Within the spirit of the present disclosure and the scope of protection of the claims, any modifications and changes made to the present invention shall fall within the scope of protection of the present disclosure.

Claims

A real-time simulation and test method for a control system of an autonomous driving vehicle, the method comprising the following blocks:

S100: providing a vehicle dynamics model;

S200: using a control algorithm to carry out a simulation test and a field test to determine whether the field test and the simulation test have reached the same performance, if yes, then proceeding to the next block S300; and if not, returning to the previous block S100;

S300: providing an input scenario library that covers a plurality of driving scenarios;

S400: providing a vehicle dynamics model library including the vehicle dynamics model established in S100;

S500: designing a control algorithm and verifying a stability of the control algorithm, wherein the design of the control algorithm and the verification of the stability of the control algorithm are performed according to a vehicle dynamics model under normal operation conditions;

S600: using a Monte Carlo shooting method to carry out an automated traversal test and determining whether analysis simulation results of the automated traversal test meet robustness and stability requirements of the control system, if yes, then proceeding to the next block S700; and if not, returning to block S500; and

S700: carrying out a field test, and if the design requirements are not met, then returning to block S500 to produce a new design for the control algorithm.
The real-time simulation and test method for a control system of an autonomous driving vehicle of claim 1, wherein block S100 further comprises establishing an error interference model and a noise model.
The real-time simulation and test method for a control system of an autonomous driving vehicle of claim 1 or 2, wherein the scenario library in block S300 includes the following scenarios:

1) different acceleration-cruise-deceleration scenarios;

2) vehicle following scenarios at different speeds;

3) AEB scenarios;

4) vehicle following scenarios;

5) front vehicle emergency cut-in scenarios;

6) scenarios with different turning radii at different speeds; and

7) emergency obstacle avoidance scenarios.
The real-time simulation and test method for a control system of an autonomous driving vehicle of any one of claims 1-3, wherein the dynamics model library in block S400 includes:

1) the dynamics model under normal operation conditions;

2) a dynamics model for wet roads with different adhesion coefficients;

3) boundaries of a tire dynamics model caused by tire wear, etc. under different loads; and

4) performance boundaries of dynamic models of a brake, a steering wheel wire control system, an engine, etc. throughout the life cycle of the vehicle.
The real-time simulation and test method for a control system of an autonomous driving vehicle according to any one of the above claims wherein block S200 comprises repeatedly performing iterative optimization to determine whether the field test and the simulation test have reached the same performance.
The real-time simulation and test method of any one of the above claims wherein the control system is for an autonomous driving truck.
The real-time simulation and test method for a control system of an autonomous driving vehicle according to any one of the above claims wherein the control algorithm in block S200 is to perform a simulation test of the vehicle dynamics model.
The real-time simulation and test method of any one of the above claims wherein verifying the stability of the control algorithm comprises verifying that the stability of the control margin is within a stability margin.
A non-transitory computer readable storage medium storing instructions which are executable by a processor to perform the real-time simulation and test method of any one of claims 1-8.
A real-time simulation and test method for a control system of an autonomous driving truck, the method comprising the following steps:

S100: establishing a vehicle dynamics model;

S200: using a control algorithm to carry out a simulation test and a field test, and repeatedly performing iterative optimization to determine whether the field test and the simulation test have reached the same performance, if yes, then proceeding to the next step; and if not, returning to the previous step S100;

S300: perfecting an input scenario library to ensure that the scenario library is able to cover sufficient driving scenarios;

S400: perfecting a vehicle dynamics model library;

S500: designing the control algorithm, and the design of the control algorithm and the verification of a stability margin being performed according to a dynamics model under normal operation conditions;

S600: using a Monte Carlo shooting method to carry out an automated traversal test to determine whether analysis simulation results meet the robustness and stability requirements of the control system, if yes, then proceeding to the next step; and if not, returning to step S500; and

S700: carrying out a field test, and if the design requirements are met, the design of the control algorithm is completed, otherwise, returning to step S500 for the next design of the control algorithm.