Impedance controlled human–robot collaborative tooling for edge chamfering and polishing applications

Highlights

• A curve tracing approach, where an impedance-controlled robot undergoes a constrained motion along a parametric-curve in response to an external force (from human–robot physical interaction for collaborative operation).

• Application of the proposed framework to collaboratively carry out standard industrial tooling tasks such as edge-chamfering and edge-polishing to achieve improved surface quality.

• The tool trajectory generation with minimal end-user programming/re-programming.

Abstract

Surface finishing, as the final stage in the manufacturing pipeline, is a key process in determining the quality and life span of a product. Such a task is characterized by low contact forces and minimal material removal from the object surface. Despite the advancements in machine learning and artificial intelligence, human workforce is still irreplaceable in performing such tasks due to superior dexterity and adaptability, but this is often prone to risks such as hand-arm vibration syndrome due to hand-held tools. Therefore, we propose a collaborative approach to assist the human in carrying out such tasks with the help of two case studies: Human–Robot-Collaborative edge chamfering and polishing tasks, based on an impedance controlled collaborative curve tracing technique.

We propose a collaborative framework, where the robot assists an operator to guide the end-effector/tool along a pre-defined parametric curve. The algorithm is demonstrated in two scenarios. In the first case, we address a collaborative chamfering task whereas the second case focuses on a polishing application (for straight edges). For these kinds of tasks, the curve to be traced assumes the shape of a straight line along the edge. We make use of the compliant feature of a cobot, which allows the user to physically guide the robot in the task space, to generate a mathematical model for the tool path. From the end-user perspective, this is more intuitive than the classical programming-based path planning approaches. In the process of machining, to enhance the path tracking accuracy and to ensure constant tool-surface contact, we implement guidance virtual fixtures through impedance control. As a result, the machining error is reduced.

Introduction

Machining processes often leave residual materials on object surfaces, which adversely affects not only the aesthetics but also the performance and life span of the component. Hence, the finishing operations are necessary as the final stage in the manufacturing pipeline where the desired surface profile is achieved through a minimal amount of material removal. Further, key finishing processes such as polishing are considered to be so influential in determining the quality of the final product [1], [2] that they can consume up to 50% of total manufacturing time [3], [4], [5]. Despite the growth of robotics industry as a whole, the finishing tasks are carried out mainly by skilled labour [6], [7] as robots cannot match human performance at contact tooling tasks which involves significant interaction with the dynamic environment [8].

Ideally, in surface tooling, an operator should be effortlessly manoeuvring the tool to modulate process parameters such as the cutting angle and feed rate. Nevertheless, such adjustments are quite challenging in practice, due to the tool weight, inertia and the reaction forces due to cutting. Consequently, the operator is unable to sufficiently control the material removal resulting in variabilities in the manufactured parts. In addition, for typical finishing tasks such as polishing, failing to maintain a constant applied force leads to various surface defects such as form deviations and local imperfections [9]. At the same time, from the operators perspective, maintaining the tool position with a constant force against the chattering and vibrations for prolonged duration leads to ergonomic issues [10].

In the recent past, a new category of robots known as collaborative robots (cobots) has emerged that can share the workplace with humans and interact physically without compromising workplace safety [11], [12]. The Human–Robot Collaboration (HRC) can increase productivity as it combines human decision making and adaptability with the precision and repeatability of the robots. Besides, unlike the conventional industrial robots that require expensive work cells, cobots can share the floor with humans leading to lower costs. Such HRC systems are an active research area for a wide range of applications such as surgeries [13], [14], drilling and sanding [15], welding [16], and assembly [17], [18]. Manufacturing companies, in particular, are increasingly deploying the commercially available cobots such as KUKA iiwa, UR and Franka Panda due to their affordability, safety and intuitive programming capability [8].

Cobots, in comparison to the conventional rigid robots, possess superior design (e.g. backdriveable actuators equipped with force/torque sensors) and control schemes (such as impedance control). According to Hogan [19], a robot is said to be impedance controlled if the force is regulated in response to changes in trajectory, while in admittance control, the robot kinematics is regulated in response to a force. Thus, by regulating the dynamic relationship between the position and force, a robot can deal with unknown environment effectively. For example, Gaz et al. [20] proposed a control algorithm for Human–Robot Collaboration targeting manual polishing process where the workpiece is held by the robot and the tool is carried by the operator. The proposed admittance controller can distinguish between the external force applied by the tool and the force applied due to the human interaction with the robot. As a consequence, the robot can secure workpiece while simultaneously allowing the operator to kinesthetically reorient the robot pose. A similar study on polishing task is presented in [21], demonstrating how the robot can be made to adapt its configuration to account for human ergonomics. The compliance feature of robots is also widely exploited by Learning from Demonstration (LfD) paradigms as outlined in [22], where the robot learns from the human operator by carrying out a collaborative polishing task.

In this paper, we focus on two prominent finishing tasks: polishing and chamfering. Collaborative surface finishing being an area of active research, several studies have been conducted to bring about the coexistence of humans and robots realizing a shared workspace. For example, the work presented in [23] implements a Human–Robot collaborative platform for surface finishing, particularly targeting polishing tasks. Yet, the human–robot interaction is merely through graphical interfaces and the user is not allowed to enter the workspace while the tool is active. The study discussed in [20] presents a collaborative polishing task, where, the role of the robot is to hold the workpiece while the human carries out the surface tooling. Though the user is free to interact physically with the robot to adjust the pose of the workpiece, the robot does not assist the operator in balancing or manoeuvring of the tool. Peternel et al. [24] proposed a hybrid force/impedance controller, where a collaborative surface polishing task is carried out with an assistive robot. While polishing, the robot ensures proper contact through force control (perpendicular to the surface to be polished) simultaneously allowing a compliant motion in parallel to the surface through impedance control. Furthermore, the robot performs gravity compensation for effortless tool guidance. A similar robot–surface interaction scheme for aircraft canopy polishing is presented in [25]. However, the human does not interact physically with the robot, instead controls it remotely through a joystick to accomplish the task.

Application of an uniform pressure is crucial in obtaining a defect-free polished surface. The work in [26] proposes a collaborative approach which allows the operator to physically guide the robot, simultaneously maintaining the contact between the tool and the surface (thus ensuring a uniform pressure, but normal to the surface).

Collaborative chamfering, however, is a marginally researched area. To the authors’ knowledge, no study has been done on the implementation of a human–robot collaborative chamfering task involving user–robot physical interaction. Considering the significance of a quality chamfering for increased product life span and performance, we believe more research needs to be done in this area.

Though Human–Robot Collaborative operations are proven to be very promising to carry out contact tasks, some schemes are necessary to be implemented within an HRC framework to ensure high surface quality and a safe working environment. One of such necessary schemes is the implementation of path/motion constraints to enhance human performance using virtual fixtures [27]. The purpose of virtual fixtures is quite similar to a ruler that is used to draw a straight line, except that the ruler is a tangible fixture. Virtual fixtures are software generated forces used to create the illusion of an actual fixture. They assist the human operators by keeping the tool along the desired path or region and prevent any possible movements towards specific forbidden axis or plane. In collaborative tasks, the virtual fixtures have proven to enhance operator performance by up to 70% [28].

One major advantage of virtual fixtures is that they are not restricted to a specific task and are portable to a wide range of applications as validated by Henry et al. [29]. In fine-manipulation tasks, where the accuracy of task execution is quite crucial and even the smallest of the deviation in the path can drastically affect the completion of a task such as surgery or polishing, such a scheme is indispensable. The majority of the research focusing on the virtual fixtures are developed for teleoperations. Rosenberg studied the implementation and testing of such virtual fixtures for telerobotic manipulation in the peg-hole assembly process with the aid of vision systems [30]. A similar experiment has been performed by Bettini [31], where the target was to achieve two types of motions: manipulator motion (a) towards a point and (b) along a path. As in the previous study, the task is assisted by a vision which delivers the manipulator an idea of the reference path.

In this paper, we extend our previous work on collaborative curve tracing [32] by demonstrating its applicability with the help of actual industrial case studies. Here, we implement virtual fixtures through impedance control to enhance the accuracy of the tooling task, while the proposed curve tracing mechanism generates the tool trajectory in response to the human–robot interaction force. We primarily target tooling tasks such as edge polishing and chamfering (for straight edges), where, one can formulate an analytical model for the tool path. With the proposed approach, an operator physically interacts with a cobot to perform surface machining to achieve the desired profile.

In contrast to the classical programming based path planning approaches, we make use of the geometrical properties of surfaces to identify the tool path along an edge. A cobot with impedance modulation capability can be operated in a compliant (low impedance) mode, where it undergoes a motion in response to external (interaction) forces. Owing to this feature, the operator physically guides the robot in compliant mode to sample triplets of non-collinear points from two intersecting surfaces, to identify the edge to be chamfered/polished. During the execution, to ensure improved path tracking accuracy against lack of hand-arm steadiness and chatter, we implement virtual fixtures through impedance control. Consequently, the chances of tool slipping are averted while operating on unstable regions such as sharp edges, resulting in a safer work environment. Moreover, with the proposed curve tracing technique, the operator is granted fine control over the tool position along the edge (while the robot maintains the user-defined orientation), feed rate and the number of cycles.

In this work, we perform two different trials for each of the tooling tasks, with the robot acting as (a) a gravity compensator and (b) as an assistant, making use of the proposed impedance controlled collaborative framework.

The rest of the paper is organized as follows. In Section 2, we present a theoretical framework for Human–Robot collaborative curve tracing tasks. Sections 3 Experimental case study 1: Edge chamfering, 4 Experimental case study 2: Edge polishing demonstrate the applicability of the proposed framework in performing edge chamfering and polishing respectively, followed by the discussion and conclusion in Section 5.