Unconventional modular library design for reconfigurable manipulators

Abstract

Current trends of escalation in robotic applications to accomplish a variety of tasks has led to the need of customization in the design of robotic manipulators. As the era is rapidly heading towards Industry 4.0, where flexibility is the challenge for mass customization of the products or the maintenance services, the adaptability is required to be induced in the system performing those tasks. Besides, the utilization of robotic systems in maintenance, service and assistance in several application sectors, involves a large variation in layouts, environments and requirements. Therefore, there is a need to handle customization of the robotic systems in an effective and efficient manner. For that, modularity and reconfigurability are the key aspects for the customization of the manipulators, and have been found promising. A modular robotic system has several advantages, most notably the ease of modification, and the ease of repair. Major challenges are associated with the design and development of such modular and reconfigurable systems which can be utilized in a variety of environments and different tasks. How to design general modules but still optimal architecture? How to generate different configurations? How to model them automatically? There are several challenging aspects which have been handled in this work. A three-phase strategy is proposed in this work for the complete design and development of a modular library, which can be assembled together to realize any n−DoF serial robotic configuration. The configuration might be required to accomplish a given set of tasks in a given work-cell. The systematic approaches are proposed for the optimal architectural planning and the design of the modules, configuration realization with unconventional parameters, unified modeling of unconventional configurations, and task-based design synthesis of the modular compositions. A literature-digest is presented for the quick visualization of the state-of-the-art literature related to the research directions of the proposed work, arranged with respect to area and era. A thorough review of various modular designs are presented and are mostly found to be having conventional designs, i.e, adapting to parallel or perpendicular jointed configurations. The designs are majorly based upon the kinematic aspects, for which there is a need to answer about the effect of inertial parameters on the modular manipulator dynamics. Considering this, a generalized approach is formulated to either improve the given modular architecture or to synthesize a new architecture based upon the dynamics of the manipulator. This partially covers the first phase of the work. Architecture Prominent Sectioning−k (APS−k) strategy is proposed in this phase which assumes the architecture as different independent sections, inertial parameters of which can be optimized with respect to minimal joint torques of the given modular configuration. The optimized parameters are remapped into a new optimal architecture of the joint module by re-location of parameters and re-selection of components. The optimal unconventional link architectures are synthesized considering large number of sections (k) and the conceptual design of the curved links are presented. The unconventional and adaptable modular library is proposed which consists of adaptive twist units, link modules, actuators, actuator casings, and a controller. The library is adaptable to unconventional parameters, i.e. able to assemble in non-parallel and non-perpendicular jointed configurations. To validate the reconfigurability and the modularity, various configurations are developed and are categorized into three major types. This classification is based upon the type of the twist parameters available in the configuration. The design and analysis of the modular library such as, assembly feasibilities and rules, worst torque analysis, and comparison of modular library based upon payload-to-weight ratio are the objectives of phase 2. Final phase of the work deals with the unified modeling and integration of the fabricated hardware with the software for motion planning and control of modular configurations. The unified models are required to be computed automatically, as after reconfiguration the kinematics and the dynamics of the configuration changes. A reconfigurable software architecture is proposed which is based upon the Robot Operating System (ROS) platform, with the benefits of abstracting low level machine implementations, enabling the user to focus on high level tasks. The complete approach of task-oriented manipulator configuration design is integrated, demonstrated for standard and unconventional configurations, and implemented for optimal combinations. A non-linear constrained optimization problem is formulated which synthesizes a custom configuration for a given set of tasks, in the given cluttered environment and takes care of the number of Degrees of Freedom (DoF), inverse kinematics of unconventional configurations, joint torques limit based upon the actuator specifications, collision avoidance and motion planning between the task-space locations.The output is the unified model representation of the modular compositions which can be used directly in ROS platform for the execution of the planned trajectories.