Safe parallel robots for human-robot collaboration (Acronym: SAFE)

• Applicant: Professor Dr.-Ing. Thomas Seel
• Subject Area: Automation, Mechatronics, Control Systems, Intelligent Technical Systems, Robotics ; Human Factors, Ergonomics, Human-Machine Systems
• Term: since 2020
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 444769341

Human-robot collaboration (HRC) enables an increase in performance in incompletely automated areas and supports humans ergonomically in their work. Safety requirements specify compliance with threshold values for energy, force, and pressure in the event of physical contact. Low moving masses of serial lightweight robots reduce the occurring forces. However, joint-mounted drives increase the moving masses and inertia of the kinematic chain. An alternative approach is to use parallel robots (PR), where the drives are typically attached to the base and connected to a mobile platform via separate kinematic chains. The kinematic chains have lower moving masses, resulting in an increase in tolerable speed while maintaining the same energy limits. Human safety must be ensured in a collaborative application-whether serial or parallel. Therefore, collisions and clamping must be detected and immediately terminated by a reaction. In the first project period, the feasibility of contact detection and reaction of PRs was demonstrated. In this context, it was shown in various collision and clamping experiments that the contact detection and reaction enable the investigated contacts to comply with the force thresholds specified in the HRC standard (ISO/TS 15066) even at higher speeds. However, research on these methods revealed that safe HRC with PR can only be realized with additional consideration of the inherent limitations of parallel robots, in particular self-collisions, joint angle limits, and singularities in the operational space. For example, a contact reaction in the presence of singular configurations leads to high drive torques of the PR and thus an increase in the risk of injury to humans. Compliance with these operationally critical limitations is crucial for the feasibility and thus the success of the contact reaction. The planned proposal continues the research of PR for HRC by adding a redundancy resolution to the contact removal reaction to maintain these limits. To be able to execute the contact reaction earlier, a faster and more precise contact detection is learned using time series of physically modeled features compared to conservatively pre-set threshold values. For this purpose, the sensors for joint angle measurement are fused with inertial measurement units (IMU) to determine external contact forces using the dynamic model of the PR faster than with methods established in HRC. The novelty of the project is characterized by (i) the contact detection using a single IMU for PR, (ii) considering the time series of physically modeled variables and (iii) the executable reaction related to any contact location of the PR. The new methods for detection and reaction are relevant for unpredictable contacts, which can occur at any location and in any joint angle configuration, and thus enable safe human-robot collaboration with parallel robots for the first time.

• DFG Programme: Research Grants