| Hyper-redundant manipulators are long and slender robotic arms capable of searching, exploring, and operating in confined environments. The very large number of actuated degrees of freedom requires the complex activation of simultaneous actuators to perform point-to-point motions between two poses. There exists two broad classes of motion planning algorithms in high dimensional spaces, effector-based and joint-based methods. While effector-based resolution methods can provide a quick solution to the end-effector path, they require an inverse kinematics solver to fit the joint space to the desired trajectory. Meanwhile, joint-based planners solve the path planning problem directly at the actuator level, but the statistical nature of these processes is sensitive to the dimensionality of the problem and falters in higher order systems. The goal of this work is to improve the joint based path planning of hyper-redundant manipulators by exploiting the behavioural strategies observed in octopuses. The research focuses on two search algorithms for the resolution of the inverse kinematics and the path planning problem of discretely-actuated manipulators. The first part of this thesis presents an inverse kinematics method inspired by the dynamic quasi-articulated structure of octopus limbs. A novel probabilistic roadmap path planner, based on the antagonistic muscle contraction waves responsible for the movement of the tentacle, is then proposed along with a method for focusing the search area during the roadmap construction phase.;Keywords: hyper-redundant manipulators, inverse kinematics, obstacle avoidance, path planning. |
