Research On Human-Like Decision Making And Motion Control Of Autonomous Vehicles During Curve Driving

Autonomous vehicles are an emerging industry under the background of scientific and technological revolution and have become a hot issue in the research field.Curves are the sections where major safety accidents occur frequently.When a vehicle enters a curve from a straight road,its speed is usually high.Therefore,it is difficult to coordinate the lateral and longitudinal movements of the vehicle at the same time.At present,there are mainly the following scientific problems to be solved for autonomous vehicles driving on curves.Firstly,existing decision-making research tends to be safe and conservative,with a large gap between the decision-making of real drivers.This leads to lower user acceptance and an impact on vehicle traffic efficiency.Secondly,the collaborative planning ability for the lateral and longitudinal movement still needs to be improved.It is difficult to balance the sportiness and ride comfort of an autonomous vehicle.Thirdly,there is a lack of digital and quantifiable objective evaluation methods for autonomous vehicles driving in curves.To solve the above three problems,this study combines real-car experiments and driving simulator experiments to explore the decision-making mechanism of real drivers in difficult situations,and proposes a humanized decision-making method that balances safety,comfort,and ethics,enabling the autonomous system to achieve humanized decision-making.A human-like vector control method for autonomous vehicles is proposed based on the coordination mechanism of skilled drivers;Finally,a quantifiable and objective bioelectrical evaluation method for autonomous vehicles was designed,which can intuitively demonstrate the comfort level of the riding experience.The main research content is as follows:(1)In response to the problem of existing decision-making that leans towards safety conservatism and has a significant gap with the decision-making of real drivers,a human-like decision-making method that considers safety,comfort,and ethics is proposed based on the decision-making mechanism of real drivers in dilemma situations.This method considers the safety of road participants while ensuring the safety of the vehicle from collisions,and solving the trade-off between safety,comfort,and ethics in decision-making.(2)To solve the problem that the lateral and longitudinal motion control of autonomous vehicle in curve conditions affects each other,and the sportiness and ride comfort restrict each other,a humanized acceleration vector control method for curve driving is proposed.Through actual vehicle experiments,it was found that under the control of skilled drivers,the lateral acceleration and longitudinal acceleration of the car vary synergistically,resulting in a smooth change in the absolute value of the synthesized acceleration.This coordination method essentially increases the value of the car’s longitudinal acceleration to reduce the peak lateral acceleration in the curve,making the vehicle’s movement more stable.In addition,a coordination strength coefficient is designed in the control method,which can be adjusted to enable the vehicle to perform acceleration and braking interventions of different intensities.(3)Aiming at the problem of the lack of objective and digital evaluation methods for evaluating the experience of autonomous driving on curved roads,a digital bioelectrical information is proposed.By utilizing the bioelectrical information of passengers,the muscle groups that are greatly affected by vehicle lateral motion were explored through real vehicle experiments.Then,the correlation analysis between muscle activation and the root mean square value of vehicle lateral acceleration was conducted.The Pearson correlation coefficient between the activation of the sternocleidomastoid muscle and the root mean square value of the lateral acceleration of the vehicle is greater than 0.8,indicating a strong correlation level.The strong consistency between the activation of the sternocleidomastoid muscle and the smoothness of vehicle movement can be used to evaluate the comfort of cornering conditions.Finally,the human-like vector control method for autonomous vehicle driving on curves was tested on a real vehicle,and the method was installed on a real vehicle test platform.The effect of the method was verified in the curve condition.The vehicle data and passenger bioelectric information during the real vehicle test were collected.Compared with the experimental results of ordinary drivers driving vehicles,the humanlike acceleration vector control formed by imitating the cooperative mechanism of skilled drivers can significantly reduce the root mean square value of lateral acceleration,lateral acceleration variance,and root mean square value of combined acceleration during the curve process.and can better coordinate lateral and longitudinal acceleration simultaneously.In addition,it also coordinated the lateral and longitudinal acceleration at the same time.From the perspective of bioelectric signal evaluation methods for the experience of driving on a curve,this method can significantly reduce the activation level of the left and right sternocleidomastoid muscles of the occupants during the two stages of the curve,and the reduction trend is better with the increase of vehicle speed.The activation level of the left and right sternocleidomastoid muscles is reduced by more than 20%.Therefore,it can be proved that the human-like acceleration vector control method for curve driving can significantly improve the ride experience.The test results show that the method has reached and exceeded the level of motion coordination of ordinary drivers,making autonomous vehicles more human-like.In addition,the actual ride comfort of autonomous vehicles equipped with this method has been improved.