Modelling of Quadrotor Ground Effect Forces via Simple Visual Feedback and Support Vector Regression

摘要

A new method using visual feedback for modelling ground effect forces experienced by a quadrotor is described and experimentally demonstrated. The quadrotor is flown over a surface with streamers attached and the onboard camera images are processed to calculate an aggregate energy value proportional to the total amount of change in the current image compared to previous images. Since this image energy will be related to the downwash coming from the quadrotor, it can be combined with the current throttle value to make an e-Support Vector Regression predictive model of the ground effect force. This model is experimentally demonstrated to be robust to additional data outside the training set as well as accurate enough to improve control performance.