Design of Robots for Unstructured Environments - A Customisable Agile Quadruped, a Sensorised High-Force Gripper, and an Intelligent Drilling End-Effector

Robots for unstructured environments explore and physically interact in variable, uncertain, and unpredictable settings, as the contexts in which they are deployed are not adapted to facilitate their operation. Their technology is not yet mature for industrial use, but interest is rising for simple industrial inspection tasks. The work presented in this thesis aims to facilitate the industrial deployment of robots for unstructured environments by providing robotic design case studies. Three novel devices are introduced, with the purpose of advancing agile quadrupedal locomotion, high-force grasping, and robotic power drilling. The first device is Leon, an agile quadruped designed from commercial actuators, eliminating the need for actuator development while retaining full access to the platform’s hardware and software. Leon demonstrates a simple yet effective design approach, which simplifies agile robot design by using COTS components. The second device is DAGANA, a robust, modular, high-force-capable 1-DoF gripper that enables high-payload operations with everyday objects in unstructured contexts. DAGANA features a multimodal sensing system that mitigates sensor occlusion during grasping, and its modular design allows extension of its dexterity and sensing capabilities. The last device is Smart Drill, a sensorised robotic power-drilling end-effector that enables low-payload robotic arms to perform drilling operations. Smart Drill integrates intelligence into its design, allowing the endeffector to directly control the task execution in a robotic-platform-agnostic manner, facilitating seamless installation across different robotic arms. While Leon’s first prototype is still to be manufactured, its design is evaluated and validated through simulations and experiments on its actuator. Conversely, physical prototypes of DAGANA and Smart Drill have been fabricated and used in extensive laboratory experiments. The experiments not only assess the design’s effectiveness but also showcase the applications of software pipelines appropriately designed to execute tasks relevant to real-world applications