• stima
• robot
• navigazione
• Kalman Filter
• IMU
• umanoide

The main objective of this work is to design and implement a state estimator on the humanoid robot COMAN. First we tested a state estimator based on data recorded from a walking gait offline in Matlab environment. The data used came from IMU and leg kinematics. Subsequently, we compared the results with the results coming from the motion capture and implemented it in the online controller developed in C++. Most of the challenges were found on the hardware: indeed, the presence of backlash in the joints founds is reflected in difficulties in precision and need for a careful tuning of the Kalman Filter used for the estimation. In conclusion, the whole work is tested on the robot to see how it affects the controller.