Skip to Main Content
In this paper, a novel control architecture and strategies, involving noninvasive brain-computer interface (BCI), are presented, and a virtual prototype of quadruped walking robot, named QWR-I, is developed for simulation of navigation behaviors of the BCI-based robot. The BCI can provide limited patterns of user's intention based on EEG signals, following user's motor imagination. The proposed control architecture, dealing with both robot's autonomous planning and user's decision acquired from BCI, is elaborated. To achieve efficient navigation in uneven terrain environment where the robot is located, three control modes are proposed, taking trade-off between robot's autonomy and user's flexibility. Movement efficiency is priority in relatively even terrain by strolling mode, while safety is considered more in uneven terrain by terrain mode. The coding protocols, which semantically represent the same set of BCI signals with different meanings according to the situated control mode of the robot, and switching strategies between different modes are elucidated. To satisfy the user's intention of desired path to destination to greatest extent possible on the premise of walking stability, cooperative control strategy based on centroid stability margin while the robot walks is presented.