Kalman Filter for Contact Detection and Localization during shin collisions in a dynamic legged robot

Nowadays Dynamic Legged robots are one of the major field of research in robotics thanks to their capability of movement in different, various and challenging scenarios. They are used in outdoor environments to help human and collaborate with them, performing tasks in situations that are critical and hazardous even for human safety. The demanding work for researchers is to develop techniques to maintain always a robust motion in such difficult terrains, allowing the robot to accomplish the desired tasks. The robot's motion is performed using innovative methods based on trajectory planning optimization, control algorithm and state estimation. Although several studies in these fields and optimal results achieved, shin collision can occur during the locomotion, causing the getting stuck of the robot and preventing, as a consequence, the achievement of predetermined goals. A solution to deal with it, is to evaluate and estimate the contact point along the shin when the robot enters in collision with undesired objects,in order to use this information as a feedback to stabilize the trunk controller and help the robot to overcome the obstacles. This thesis extends the work developed in a 2D plane, building a novel model in a 3D environment and considering both velocity and acceleration to detect and localize the contact point in a more accurate way. The estimation of this point is performed with the contribution of the Kalman Filter, one of the most common technique of filtering, in order to avoid noise corruption due to the usage of sensors and to determine the optimal output, following the trial and error procedure based mainly on the tuning of the parameters. The validation of the obtained results is executed both in simulation and on the Hydraulically actuated Quadruped robot (HyQ), including experiments with non flat terrains. This thesis is the outcome of a one year project performed at the Dynamic Legged Systems Lab (DLS) at IIT in Genova.