Performance Evaluation of Artificial Neural Network-Based Shaping Algorithm for Planetary Pinpoint Guidance

Abstract

Computational intelligence techniques have been used in a wide range of application areas. This paper proposes a new learning algorithm that dynamically shapes the landing trajectories, based on potential function methods, in order to provide
computationally efficient on-board guidance and control. Extreme Learning Machine (ELM) devises a Single Layer Forward Network (SLFN) to learn the relationship between the current spacecraft position and the optimal velocity field. The SLFN design is tested and validated on a set of data comprising data points belonging to the training set on which the network has not been trained. Furthermore, the proposed efficient algorithm is tested in typical simulation scenarios which include
a set of Monte Carlo simulation to evaluate the guidance performances.