Model Predictive Control for Skydiver Fall-Rate Adjustment

An innovative approach of controlling a system with multiple Degrees-of-Freedom performing complex maneuvers is presented. A system under investigation is a human body maneuvering during a free-fall stage of skydiving, which is a rapidly developing sport. The objective is creating an autonomous system capable of performing skydiving maneuvers in a virtual way, and turning it into a powerful tool for improving instruction methods. The key idea is to build a skydiving skill from a set of basic and advanced maneuvers, replicating human natural skill acquisition stages. For each type of maneuver a control algorithm is designed based on the specific plant dynamics, task constraints, and analysis of body actuation possibilities. The main challenge is combining these control schemes to enable the virtual skydiver perform realistic tasks. In the present work we develop a model predictive controller for the fall-rate adjustment, which is the essential skill for performing cooperative maneuvers. We exploit the differential flatness property of the plant in the vertical dimension in order to combine the fall-rate controller with a previously developed Two- Input- Two-Output controller for horizontal maneuvers. The design is tested in a whole body dynamic simulation of a skydiver model. MPC, differential flatness, QFT, non-linear simulation