Design and deployment of human-robotic applications using advanced digital technologies

One of the key objectives of this research work is to support the design of robotic cell configurations, including the cell layout and equipment design. Exploring different cell configuration alternatives, their feasibility as well as their performance in terms of a set of Key Performance Indicators is often cumbersome and takes up a significant amount of time. In addition, a new cell configuration may require physical validation, necessitating the use of a real robotic cell, leading to potential production disruptions. This poses a direct challenge for the industry. Such challenges may be partially addressed by employing robotic process simulation platforms. In this thesis, two different studies have been conducted in robotic cell and equipment configuration: simulation-based design of a) gripper fingers and b) baggage handling systems with aid of advanced digital manufacturing platforms. The simulation-based design of robotic gripper fingers was implemented by integrating a Computer-Aided Design platform and a 3D process simulation platform with a physics engine. The CAD together with the Physics-based simulation framework were deployed to work in tandem to automatically redesign and validate an initial gripper finger. One of the key advantages of the proposed approach is that it is capable of testing several gripper design configurations for handling multiple workpieces. Further, the physics-based process simulation allows the realistic exploration of different configurations and what-if scenarios. The study pertaining to airport baggage handling systems was conducted in collaboration with Dublin Airport Authority and led to the analysis of a set of baggage handling layouts, digitally constructed and evaluated with the aim to investigate their performance. Both studies proved that significant time and cost savings may be achieved by employing digital manufacturing and process simulation tools in complex manufacturing and logistics applications.
The second key objective of this thesis is to address the problem of accurately estimating the pose of a human operator on a shop floor taking advantage of vision and Inertial Measurement Sensors. Initially, data from two spatial computing developer kits together with IMU sensors were fused to estimate and track the pose of the human operator and the status in the shop floor. One drawback of the approach was a drop in the frequency of the fused output data. To address this limitation, a forward kinematic-based approach was employed, fusing the information from the spatial computing sensor and the IMU using Kalman filter. This led to significantly improved performance and accuracy. The next study employed a series of multiple different vision systems towards exploring the idea of whether combining the capabilities of different hardware systems and software frameworks may lead to better performance and accuracy in detecting the human operator pose. The study involved 3 different spatial computation kits, i.e., the Azure Kinect, Intel D455 and ZED2. The results indicated that at a distance less than 3 m, Azure Kinect demonstrated better tracking performance, followed by Intel D455 and ZED2, while at ranges higher than 3 m, ZED2 had excellent tracking performance.
Lastly, two studies were performed to address the challenges of integrating and deploying human-robot collaborative applications, taking advantage of mobile robotic platforms. The first study focused on a novel straightforward approach, with an easy-to-implement control strategy involving haptic-force and compliance control feedback from the robot arm was developed to guide the base of the mobile robot to the location. Furthermore, this was developed for applications where the co-manipulation of long parts is required. The second project employed Edge-AI devices and a mobile robotic platform using ROS for demonstrating how Edge computing principles may be employed in demanding collaborative robotic applications.