Force-based control for human-robot cooperative object manipulation

Doctoral thesis, 2021

In Physical Human-Robot Interaction (PHRI), humans and robots share the workspace and physically interact and collaborate to perform a common task. However, robots do not have human levels of intelligence or the capacity to adapt in performing collaborative tasks. Moreover, the presence of humans in the vicinity of the robot requires ensuring their safety, both in terms of software and hardware. One of the aspects related to safety is the stability of the human-robot control system, which can be placed in jeopardy due to several factors such as internal time delays. Another aspect is the mutual understanding between humans and robots to prevent conflicts in performing a task. The kinesthetic transmission of the human intention is, in general, ambiguous when an object is involved, and the robot cannot distinguish the human intention to rotate from the intention to translate (the translation/rotation problem).

This thesis examines the aforementioned issues related to PHRI. First, the instability arising due to a time delay is addressed. For this purpose, the time delay in the system is modeled with the exponential function, and the effect of system parameters on the stability of the interaction is examined analytically. The proposed method is compared with the state-of-the-art criteria used to study the stability of PHRI systems with similar setups and high human stiffness. Second, the unknown human grasp position is estimated by exploiting the interaction forces measured by a force/torque sensor at the robot end effector. To address cases where the human interaction torque is non-zero, the unknown parameter vector is augmented to include the human-applied torque. The proposed method is also compared via experimental studies with the conventional method, which assumes a contact point (i.e., that human torque is equal to zero). Finally, the translation/rotation problem in shared object manipulation is tackled by proposing and developing a new control scheme based on the identification of the ongoing task and the adaptation of the robot's role, i.e., whether it is a passive follower or an active assistant. This scheme allows the human to transport the object independently in all degrees of freedom and also reduces human effort, which is an important factor in PHRI, especially for repetitive tasks. Simulation and experimental results clearly demonstrate that the force required to be applied by the human is significantly reduced once the task is identified.