Longitudinal Trajectory Optimization for Connected and Automated Vehicles by Evolving Cubic Splines with Coevolution

This paper investigates a longitudinal trajectory optimization problem of connected and automated vehicles with an energy-aware non-linear objective. In this paper, we first approximate each vehicle trajectory with a cubic spline function using the proposed solution representation scheme, while the curve shape can be controlled by the knot vectors. After that, we propose a new coevolutionary algorithm that decomposes the initially high-dimensional problem and performs as the optimizer for subproblems. In the local exploitation phase, a problem-specific steepest ascent hill-climbing algorithm is developed to escape from local minimum points and speed up convergences. This proposed approach is compared with several state-of-the-art algorithms in multiple scenarios with different traffic densities and platooning sizes. Simulation results indicate that it can yield near-optimal solutions with reasonable computation times for real-life applications.