Robot obstacle accommodation: Mechanics, control, and applications.

This thesis introduces a new methodology of robot motion control in cluttered environments called obstacle accommodation. This methodology does not attempt to avoid physical contact between a robot and obstacles. Instead, it controls the contact to prevent damage to the robot. This methodology has a substantial advantage over traditional obstacle-avoidance methods in environments in which totally avoiding contact is impractical such as natural environments. Obstacle accommodation represents a new robotic paradigm in which obstacles can contact the robot at unspecified positions, i.e., a contact point can be at any point on any link. In this thesis, basic studies, including kinematics, motion planning, dynamics, and control have been conducted under this new paradigm. This thesis begins by providing a general formulation of the motion constraints due to contact with obstacles. Next, a new inverse kinematics is presented that provides joint motion for robots under contact constraints. Third, a new motion planning algorithm for robot motion in a cluttered environment is provided. Fourth, a force analysis and a dynamic analysis are conducted. The force analysis and dynamic analysis give insight into the contact force joint torque relations and into the dynamic performance of a robot in contact with obstacles. Fifth, two types of controllers are developed so that the robot is not damaged despite contact with obstacles. Finally, this thesis presents the design and motion control of a mechanical snake robot as an example of a robot that needs obstacle accommodation control because it is required to work in natural environments in which obstacle avoidance is not practical.