Abstract: To cope with the problem of real-time trajectory dynamic planning for intelligent vehicles in complex traffic settings, this paper proposes a local trajectory planning algorithm based on Frenet coordinates. Firstly, the road center line is fitted to obtain the road reference line by utilizing a cubic spline curve. Then, according to the initial and target configuration, the sets of quartic and quintic polynomial candidate trajectories are generated by sampling. Additionally, multi-objective cost functions containing safety, trajectory comfortability and trajectory offset are designed to evaluate the trajectory cost by considering the effects of vehicle safety, the lateral and longitudinal acceleration changes of candidate trajectories and the global trajectory tracking ability. Ultimately, the simulation results reveal that the proposed method can obtain a real-time comfortable and safe trajectory. Furthermore, various types of roads are designed to test the approach, including a two-lane continuous curve with a single constant-speed or variable-speed obstacle, and a three-lane curve with multiple constant-speed obstacles or variable-speed obstacles. The test results show that the algorithm can effectively improve obstacle avoidance in complex dynamic environments.