A 10,000-word article on automobile control in the intelligent era

In the future, the hot scientific research on human-machine co-driving may revolve around the following points: 1) Modeling and prediction of the driver's driving status, habits, and skills; 2) The theory of motion stability and collision safety of human-machine co-driving vehicles; 3) The driver's perception and cognition of human-machine collaboration in the loop; 4) The interaction and collaboration between humans and machines in decision-making planning and control execution; 5) Personalized human-machine co-driving system; 6) Human-machine co-driving system for conditional automation/high automation; 7) Verification platform and test evaluation method of human-machine co-driving system.

3.5.2 Autonomous Driving Vehicle Control at Different Penetration Rates

Autonomous driving will be a process of gradual development and popularization. For a long time in the future, the traffic participants on the road will be a mixture of autonomous driving vehicles and human-driven vehicles, which will bring more complexity and uncertainty to the already complex traffic environment. Therefore, intelligent driving control that can adapt to different penetration rates and its energy consumption and safety assessment will be a hot topic in the future. Reference [194] studies the impact of vehicle merging and cutting on the energy consumption of autonomous driving vehicles in response to the needs of micro-mixed traffic environment modeling and vehicle energy consumption assessment. This paper uses the Safety Pilot Model Based on the SPMD big data, a random vehicle merging modeling method is proposed. The important anthropomorphic parameters such as "courtesy coefficient" and "patience coefficient" are used to characterize the driver's merging behavior and calibrate the probability distribution. The established merging model is applied to mixed traffic environments, revealing the impact of manned vehicle merging behavior on the speed trajectory planning of autonomous driving vehicles under economic driving requirements, and preliminarily evaluating the potential energy consumption improvement of autonomous driving vehicles. Reference [195] studies the energy consumption optimization problem of vehicles merging into the main road from the ramp, establishes a mixed traffic simulation environment, analyzes the impact of traffic environments with different mixed rates on vehicle energy consumption under autonomous driving control, and provides ideas for the control of autonomous driving vehicles in mixed traffic environments. It can be foreseen that the research on autonomous driving adaptive control methods under different penetration rates in complex traffic environments will be a hot topic for a long time in the future.

3.6 Virtual Testing and Evaluation Technology of Autonomous Driving Control Systems

The design and development of autonomous vehicles is a rigorous system engineering project[202], in which testing and evaluation are key elements. To ensure that the information security and functional safety of autonomous driving control systems are reliable, mileage tests of hundreds of millions to hundreds of billions of miles are required[203]. Currently, the safety assessment of autonomous vehicles is mainly based on road tests[204]. To complete the given mileage tests, both monetary and time costs are challenges[205]. Google proposed a combination of actual road testing and virtual testing in closed test sites, completing 3 million miles of urban road testing in 7 months, greatly increasing the testing speed. Therefore, establishing a virtual test platform that can reflect real and complex traffic environments and developing reliable supplementary test methods (such as virtual testing and scenario testing) can alleviate the reliance on actual road testing, which is crucial for the testing and evaluation of autonomous driving control systems and is a future development trend.

Emerging testing methods such as virtual testing and accelerated testing are gradually becoming a hot topic in current research. In order to build a smart driving simulation test field, the University of Michigan, with the support of the local government, built the world's first virtual town Mcity specifically for testing smart cars. In a white paper released in 2017, Mcity disclosed the methodology of rapid testing of autonomous driving. Its main idea is to use big data to establish a random model of possible dangerous situations, and extract the response results of autonomous driving vehicles through a learning-based method for importance sampling, thereby completing the accelerated reliability evaluation of autonomous driving in dangerous situations. Compared with traditional methods, this method can greatly speed up the test and reduce financial costs [196−197]. New testing methods need to be coordinated with the research of virtual technology, such as the construction of the virtual perception system of the vehicle in the virtual test, the construction of the virtual scene and the acquisition of the data set. The University of Michigan Transportation Research Institute UMTRI integrates video game technology and other virtual technologies, combines the real world with the virtual world, and uses augmented reality technology to create a faster, more effective and more cost-effective CAV (Computer aided Reference [206] points out that in driving simulation systems, virtual intelligent vehicles have high requirements for the realism of the system and the reliability of the experiment, and uses database technology and collision detection algorithms to construct a virtual intelligent vehicle testing method. Reference [207] uses the available information of public data sets to synthesize a simulator based on event camera data based on a new visual sensor, combining traditional cameras with event-based sensors, which has the advantages of low latency, high temporal resolution and high dynamic range. In addition, some commercial simulation software is also used to construct simulated driving scenarios, such as SIVIC, SCANeR and preSCAN [208]. Recently, a virtual testing method based on parallel theory has been proposed [199−200]. This method uses a virtual-reality interactive vehicle intelligent evaluation method to test and verify the ability of unmanned vehicles to understand complex traffic scenarios and make driving decisions, which can help promote the further development of unmanned driving technology.

Although virtual testing technology plays a big role in the testing of autonomous driving control systems, the testing of autonomous driving control is ultimately inseparable from real vehicle or hardware-in-the-loop bench testing. At present, many new control modules for autonomous driving need to interact with the original underlying actuator electronic control unit for signal exchange, and the open permissions of the underlying electronic control unit are mostly in the hands of vehicle manufacturers or electronic control suppliers. Considering factors such as product confidentiality and competition in the same industry, it is not easy to obtain open permissions, so some important tests can only remain in the simulation stage. Therefore, in order to open up the complete autonomous driving test chain, the opening of the electronic control unit interface will be a problem that needs to be solved in the future.

4 Conclusion

Throughout the development history of the automobile industry, the process of automobile driving automation has been ongoing, and autonomous driving technology has gradually moved from assisted driving to fully automated operation of the steering wheel, accelerator, brake and gear. The advent of the era of automobile intelligence highlights the importance of electronic control. Independent research and development capabilities at the control decision-making and execution levels are the premise and guarantee for the mass production of highly automated automobiles. In the intelligent era, there are unprecedented opportunities as well as great challenges in the process of automobile automation. There are still many unknown problems waiting for automobile control workers to explore. With the mutual integration of advanced algorithms and more information to derive more new systems and new functions, cars will become more and more intelligent, and will also bring us safer, more economical, more convenient and more comfortable smart travel.