To solve the problems of trajectory prediction and obstacle avoidance of
self-vehicles in autonomous driving, a obstacle avoidance algorithm that
combines trajectory prediction and vehicle motion planning is proposed. Firstly,
in this paper, Unscented Kalman filter and constant acceleration model, namely
UKF + CA, as well as Hidden Markov model, namely HMM, are combined together.
Predict the trajectory of the vehicle in front and integrate the prediction
results obtained by these two methods, which can improve the accuracy of the
prediction. Then, in the Frenet coordinate system, this paper adopts the methods
of dynamic programming and quadratic programming to generate the planning
trajectory of the self-aircraft. After that, this paper conducts collision
detection between the fusion trajectory of the preceding vehicle and the
planning trajectory of the self-vehicle. If there is a risk of collision, a
virtual obstacle will be generated and the path will be re-planned to avoid the
obstacle. The simulation results show that, whether in the straight-going
scenario or the lane-changing scenario, this method can effectively improve the
accuracy of predicting the trajectory of the vehicle in front, effectively avoid
dynamic obstacles, enhance the driving efficiency, and also improve the safety
and comfort of driving.