Neural-network-based observer for real-time tipover estimation

Tipover issue is an important problem in autonomous mobile manipulators. This issue is becoming more important in new generation of mobile robots where their size and weight are reduced, and they are designed to work on uneven terrains with higher speeds. Estimating the distance from the tipover stability margin would enhance the capability of the mobile manipulators in correcting their motion. For a valid estimation of the tipover margin one must take into account the full dynamic interaction between the vehicle and its manipulator. In mobile manipulators with several degrees of freedom a huge amount of time is needed to solve equations of motion while real-time tipover control needs a fast and on-line access to the data. The proposed approach is a neural-network-based algorithm that enables autonomous mobile manipulator to detect its instable situations. This method greatly reduces the observer calculation time and is fast enough to be used as an observer for real-time tipover control. Accuracy and effectiveness of the proposed method is shown by an example.