In this project, a feedback controller is developed for planar pushing manipulations. Realtime control is achived by using nonlienar optimization in a Model Predictive Control setting.
Straight line tracking without feedback control: Straight line tracking with feedback control: Controller recovers the trajectory from external perturbation: Circular tracking: Circular tracking with external perturbations:
In this project, we proposed a navigation and manipulation algorithm based on the concept of affordance and trajectory optimization method, that will enable a mobile manipulator to remove unexpected obstacles on the way to destination and have the capability to decide whether to pick or push items.
Turtlebot SLAM with object recognition and localization. Specified objects (office chair) are detected and their locations are marked on the fianl map. RTAB-Map and YOLO Object Detection are used in this project.