Dynamic ranking algorithm for landing site selection

Abstract Landing on planets is always a challenging task due to the distance, communication delays, and hostile environments found. In this paper we present a dynamic ranking algorithm, based on a hybrid aggregation operator, for selecting the best site for landing. The proposed algorithm uses feedback historical information from previous iterations to ensure a reinforcement behavior in the decision process. Keywords: aggregation operators, dynamic ranking algorithm, hazard maps. 1 Introduction Past planetary lander missions tended to focus on pre-qualified landing sites, which implied a smooth terrain with little risk and few geologic features [4]. Landing on planets is always a challenging task due to the distance and hostile environments found. Improving the landing site selection process implies greater onboard autonomy, due to communication time delays and data volume involved. ASTRIUM Space Transportation has been consistently improving the hazard avoidance techniques for on-board piloting autonomy [2, 4] (denoted piloting function). Hazard avoidance includes three separate critical functions [2, 4]: hazard mapping that estimates ground features based on an imaging sensor data (camera or Lidar); site selection that chooses a suitable landing site based on available hazard maps, mission, propulsion and guidance constraints; and a robust guidance to reach the selected target. In this work our inputs are hazard maps of dimensions 512x512 pixels that provide assessments of terrain features and trajectory constraints. From these maps we have to select the best site (pixel x, y in the aggregated map). Since the selection of a suitable landing site is a critical task for any planetary mission success, the motivation for this work is to build a fuzzy multiple attribute decision-making process to select the best site. Specifically, in this paper we focus on the second step of a fuzzy multiple attribute (or criteria) decision-making process, the ranking of alternatives with respect to the global aggregated degree of satisfaction [6].For this purpose we present a dynamic algorithm for landing site ranking, taking into account past historical data from previous iterations, during the piloting function. The proposed dynamic ranking algorithm uses a hybrid aggregation operator, based on ideas from the uninorm aggregation operator [3, 9], for combining past and present data because it ensures a full reinforcement behavior [1, 10]. If we encounter a collection of high values we want that the resulting aggregation value to be more positive than any of the individual values. On the other hand, if we encounter a collection of low values we want that resulting aggregation value to be more discriminative than any individual values. The first concept is called upward reinforcement and the second concept is called downward reinforcement. Most aggregation/ranking methods are only either