Multi-Modal and Adaptive Control of Human-Robot Interaction through Hierarchical Quadratic Programming

Abstract

This paper proposes a novel Hierarchical Quadratic Programming (HQP)-based framework that enables multi-tasking control under multiple Human-Robot Interaction (HRI) scenarios. The objective is to create a unique framework for various types of HRI control modalities, avoiding the necessity of switching between different controllers that are usually tailored to a specific task. To achieve this, we firstly propose a HQP-based hybrid Cartesian/joint space impedance control formulation. This is based on the system's dynamics and provides an adaptive compliance behaviour during HRI, while performing hierarchical motion control. This is validated through a series of experiments that show the accuracy of trajectory tracking and the variable compliance behaviour. We then consider the case in which the human needs to move the robot by acting directly onto it, by proposing a hybrid admittance/impedance hierarchical control, which is then validated through several experiments in which the human moves the robot in the workspace. Next, we formulate a HQP-based force controller for HRI applications in which a specific interaction force must be maintained and lastly, we extend this to simultaneous force and trajectory tracking for applications that need higher position accuracy. Further experiments are conducted, to validate the proposed framework, using a redundant mobile manipulator. Overall, we obtain a single multi-purpose HQP-based control framework, that can switch continuously between interaction modes, enclosing multiple behaviours adjusted online based on the type of interaction.

A Stable Method for Task Priority Adaptation in Quadratic Programming Via Reinforcement Learning

“…Various strategies are employed to equip QP controllers with the capacity to adapt to dynamic environments or altered operational conditions. In [16], the authors propose a comprehensive control framework accommodating diverse parallel control modalities. Each modality addresses a different set of tasks, ordered with strict priorities.…”

“…By incorporating (20) into (16), we obtain the closed-loop system dynamics when the FC task is active:…”