Cable-Suspended Robotic Collaborative Manipulation

The collaborative grasping and manipulation of objects by multiple agents (humans and robots) is a fundamental problem in robotics, with applications in logistics, manufacturing, and assistive technologies. In most of the existing approaches the robots and humans rigidly and tightly hold the co-transported object, which is usually considered to be rigid. While effective, rigid grasping strictly couples the partners’ motions, limiting flexibility and reducing safety when unexpected events occur. Cable-suspended manipulation offers an alternative where the interaction between the agents and the manipulated object is mediated by lightweight, flexible cables. This non-rigid coupling extends the workspace, improves safety, and reduces the interaction forces. However, the suspended load becomes articulated and the system underactuated. This Thesis addresses these challenges by introducing novel modeling and control strategies for the collaborative manipulation of cable-suspended loads, aiming to establish cable-suspended manipulation as a new paradigm for human-robot, multi-robot, and human-multi-robot collaboration. First, a human-robot collaborative cable-suspended manipulation framework is developed, enabling safe co-transportation with contact detection and distinction based only on force sensing at the robot end-effector. A user study with 10 participants demonstrates its usability and safety, while an extended study with 18 participants shows that introducing wearable haptic feedback enhances operator awareness and reduces task failures. Second, a novel modeling framework is proposed, drawing an analogy with grasping theory with multi-fingered hands and multi-drone cable-suspended manipulation. The formulation defines the transmission of forces through cables and introduces the Cables Cone and Grasp Wrench Space for cable-suspended multi-drone systems. The model is further extended to multi-manipulator scenarios, and a method to optimize object manipulability both in 3D and along desired motion directions, possibly involving a human operator, is presented. These results are validated with experiments in both simulation and real-world systems. Finally, a multi-agent reinforcement learning framework is developed for dual-arm cooperative cable-suspended manipulation, allowing two robotic arms to learn how to handle a cable-suspended platform to accomplish tasks of increasing complexity, including platform manipulation, obstacle avoidance, and object transportation. Overall, this Thesis advances the theoretical foundations and practical implementation of cable-suspended collaborative manipulation, demonstrating its potential for safe and efficient collaboration among humans and robots.