Human-like motorway lane change trajectory planning for autonomous vehicles

The human lifestyle can be foreseen to have a tremendous change once the automation of transportation has been fully realised. The majority of current researches merely focus on improving the efficiency performance of autonomous vehicles(e.g. the energy management system, the handling, etc.)instead of putting the human acceptance and preference into consideration, leaving the knowledge gap of achieving the personalised automation. The primary objective of this research is to develop a novel human-like trajectory planning algorithm that is able to mimic the performance of human drivers and generate a feasible trajectory for an autonomous vehicle to complete a motorway lane change, which is the most representative and commonest manoeuvre on the motorway. This thesis can be divided into four main sections. Starting with the part of literature review, which summarises the existing techniques and the associated knowledges that can be taken the advantage of; including the trajectory planning, the driving styles, the lane change manoeuvre and the Model Predictive Control (MPC). An appropriate-designed experiment is then introduced and implemented, with the purpose of constructing a precise and reliable human driving database. This database contains 551 lane changes on the motorway from 12 different male drivers. Through applying data statistics methods, the human characteristics can be mined from the experimental data, showing that the vehicle velocity v, the hand steering wheel angle δhandreal, the longitudinal acceleration aₓ, the rate of hand steering γhandsteer and the rate of longitudinal accelerating γlongAcc are the essential features for the motorway lane change manoeuvre. An off-line constraint table for the three nominated driving styles can be therefore constructed based on these features. Finally, the obtained human information is then fused with the traditional MPC planning technique so as to achieve the proposed human-like trajectory planning algorithm. The main contribution of this study is proposing a novel approach of combining the real human driving data and the traditional planning technique(i.e. MPC) to achieve human-like lane change trajectory planning for autonomous vehicles. An integrated human driving database which contains both the video footages and the vehicle-dynamic-based signals from 12 different participants is built. Moreover, the draft marginal values of the essential parameters for the driving styles while performing a right lane change on the motorway are also presented. Both the collected driving database and the driving styles' constraint table can be seen as distinctive achievements, providing resourceful materials for future researches.