Improving Real-Time Energy-Efficient Trajectory Planning Via Machine Learning

This research demonstrates benefits from leveraging simple machine learning algorithms to generate a computationally light, suitably tailored heuristic function to enable mobile robots to make faster, more accurate decisions in navigation tasks. This can be achieved by collecting data from simulations and training top-performing algorithms like Light GBM, MLP, Ridge Regressor, etc., into a heuristic learning model. Our approach is anticipated to replace existing naive heuristic functions by reducing computation time to solution and targeting exploration.