Aiming at the real-time trajectory planning of autonomous vehicles with velocity keeping under highway scenarios, an optimal trajectory planning algorithm based on the Frenet coordinate system is proposed. Firstly, the vehicle motion is decoupled using the Frenet coordinate system to construct lateral/longitudinal independent integrator systems. Secondly, according to the initial configuration and the target configuration which can be embedded into the behavior layer, a limited set of quartic and quintic polynomial candidate trajectories are generated by sampling. Thirdly, the loss function is build using the safety and comfort evaluation index with Gaussian convolution and acceleration deriative as the core to evaluate the trajectory cost. From the trajectories checked by curvature and acceleration, the one that can minimize the loss is selected as the optimal solution. The results show that the proposed algorithm can meet the requirements of road-type scene, the motion trajectory is smooth, comfortable and safer.