3D printed passive end-effector for industrial collaborative robotic arms

The work focuses on the design and prototyping of a novel end-effector for a collaborative robotic arm allowing to grab and drag industrial packages without lifting them. The proposed solution consists of a passive 3D-printed end-effector manufactured using carbon fibre reinforced Onyx material. Thanks to the entirely passive mechanical actuation that exploits the compliance of the main chassis, this end-effector features a simple, scalable, and inexpensive structure. This lightweight end-effector is specifically designed for small and low payload collaborative robotic arms. Specifically, the proposed end-effector includes three main parts. First, a thin blade, with the main function of separating boxes that are close to each other. Second, a rocker arm - rod mechanism, which allows an opposable bracket to be moved in order to grab the correspondent package. This is proportionally and passively actuated by the contact pressure between the package, during its grip, and a paddle (third part), which is composed of a flat leverage and three flexural springs to counterbalance the pushing force. This paddle and the main body of the gripper were designed as a single part exploiting 3D printing manufacturing capabilities. Moreover, we implemented a Simscape dynamic model that predicts the functionality of the end-effector during standard operations. The work shows how to design, develop and validate a new low cost, passive end-effector mainly oriented to collaborative robots. The final prototype demonstrates its entire functionality, and proves fabricability through 3D printing, thus minimizing production costs, weight, and time.


Introduction
Nowadays industrial automation has become utterly important not only for the largest companies, but also for medium and small ones, which have the purpose to be competitive in the market and, thus, lowering costs and lead time. 1 Reducing the manual human activities helps also to avoid hazardous and stressing working conditions.In the majority of cases, repetitive activities such as warehouse packages movements, provide a huge number of benefits when automated. 2 The use of robots or robotic arms increased steadily in the last decade and specially, collaborative robots which will promote a strong revolution in industrial plants thanks to their capability of working close to human workers.In parallel to their development, there is the need for end-effectors designed for each peculiar activity. 3he present research work was developed in the framework of the COORSA project 4 a POR-FESR project supported by Emilia Romagna Region (Italy).The aim of the project was to develop collaborative robotic systems in the field of industrial production and logistics: these systems must be able to perform tasks in unstructured working areas, thanks to the mobile base on which the robot is installed.One of the work case studies was the handling of packages in logistics: the collaborative robot must grip the package from a pallet and transfer it over to a second pallet, using a proper end-effector, collaborating with the operator.
0][11][12] These systems are usually hardly scalable and not suitable for small collaborative manipulators due to the number of components such as sensors and actuators.Moreover, complex systems or uncommon materials lead to expensive solutions, and they require proper actuation systems. 8ew works have been proposed in literature exploiting simple active actuations systems 10,13,14 and also passive actuation systems [15][16][17] which ensure the grip by using adaptive soft elements and rigid ad-hoc geometries with zero mobility.
From the solutions proposed in the literature emerges that soft grip elements are useful for grasping various kinds of elements, but they are not suitable managing juxtaposed packages.On the other hand, rigid geometry ad-hoc end effectors work efficiently managing prescribed objects of the same size, but they usually fail or run into problems whenever the object dimension changes.
In the framework of the COORSA project, a crucial challenge is to propose and design a simple, passive, lightweight, low cost and innovative end-effector for industrial collaborative robotic arms able to perform warehouse packages movements.This solution provides a novel passive actuation system exploiting a kinematics which includes a compliant element to rearm the mechanism.
This artifact, which is appositely designed to grab and drag boxes without lifting them, is especially useful for the weakest or smallest robotic arms such as low payload collaborative arms.The end-effector is equipped with an internal passively actuated mechanism that allows to grab, drag, and release objects by simply engaging the correspondent box.A special thin blade allows to separate boxes that are close to each other as when they are stuffed in a transport unit.
After a first conceptual design step we performed a dynamic simulation that supports the design phase and shows the entire model working during a typical task through Simscape software. 18 full-scale prototype was entirely fabricated by filament fused fabrication (FFF) 3D printing process using Onyx material which provides the desired stiffness, and temperature resistance.Tensile tests were performed to mechanically characterize Onyx material at ambient temperature.According to the technical data sheet provided by the manufacturer, the thermal resistance is ensured by a heat deflection temperature of 145 °C performed under ASTM D648-07 B specification, which is a remarkable result among plastic materials. 19xperimental validation confirmed the functionality of the end-effector, and its peculiar features of a simple, low-cost, scalable, and effective system specifically oriented to collaborative robots.
The present paper reports all the design, development, and validation steps.Firstly, we identified the technical specifications of the project and performed conceptual design of the end-effector.Secondly, we designed all the components of the system focusing on design to assembly criteria and prototyping.Thirdly, the work presents a dynamic simulation oriented to the analyses of the end-effector functionality and to predict the forces acting on the structure in working conditions.Moreover, a static finite element (FE) analysis validates the mechanical resistance of the structure, using Onyx tensile tests data.A following prototyping section illustrates the manufacture with its long carbon fibre reinforced areas.Finally, functionality assessments are reported highlighting the forces acting in different working conditions.

Mechanical design
Table 1 synthesizes the main technical specifications that the end-effector must satisfy, according to COORSA project aims.

Conceptual design
Figure 1 illustrates the conceptual design of the end effector.
The end-effector was thought to be a moving gripper with two ends which close and open themselves respectively to grab and release objects.One of these ends is a thin but large paddle (3 in Figure 1(a)), which has a blade shape (4 in Figure 1(b)) to pass through the gap between two side-by-side boxes, packed on pallet.Its increasing thickness allows to make the gap larger as the blade goes deeper in the slit.Figure 1 shows how this operation will be performed.
Once the blade is completely inserted in the gap it is possible to drag the engaged box in a different position.To enhance the stability of the box during these operations, it is necessary to keep it steadily grabbed and a passive mechanism was designed for this purpose.A proportional compliant paddle (3 in Figure 1(a)) actuates the mechanism when pressed against the box thanks to the downward movement of the arm.The paddle stroke involves a rocker arm (2 in Figure 1(a)) by lifting one of its ends through a rod (1 in Figure 1(a)).On the other side the rocker arm (3 in Figure 1(a)) is fixed to a bracket (2 in Figure 1(a)), which consists of the opposable moving part of the gripper.Figure 1 reports a scheme of the mechanism and how a package is grabbed.
Assuming that the model can be reduced to a 2D system, equation 1 provides the gripper degrees of freedom.
where N is the number of bodies or group of bodies rigidly connected (identified with numbers in Figure 1) and C is the number of hinges (represented by blue arrows in Figure 1).The grab method is sufficiently robust since the engagement between the end-effector and the box is ensured by the geometry of the end effector.From one side, the tapered blade helps to ensure a proper engage of the box, while on the other side, the touching element blocks the box from its sliding.A possible grab failure may only happen if the robotic arm calculates a wrong position or if the blade does not slit correctly between two boxes and hit the upper face of a box.In both conditions, the issue shifts to the robotic controller, which has to reattempt the movement.In particular, in the second case, a grab failure could trigger the controller exploiting a force upper limit to reattempt the grab phase.This should be implemented by introducing a force sensor.
When the drag phase is completed, the box must be released.This action is performed in the same way as the grab phase, but in the opposite verse.While the downward force provided by the robotic arm decreases, the paddle and the rocker arm return back into their neutral position, exploiting the elastic flexural springs response.The tapered end-effector geometry and the opening gripper help releasing the box.

CAD model definition
To comply with different arm robot adapters, the main body of the end effector includes a bayonet connection (Figure 2

(a)).
There flexural springs connect the paddle to the main body of the end effector, thus obtaining a compliant mechanism (Figure 2(b)): the larger the paddle rotation, the higher the reaction force from the flexural springs.By exploiting the additive manufacturing process capabilities, the main body, the flexural springs, and the actuation paddle are merged in a single component: this solution consists of simple, low cost, and efficient structure, which avoids the introduction of conventional mechanical springs and integrates all the functions.
Figure 2(c) shows an additional element which presses against the box when the end effector is actuated, thus enhancing the grip capability at the end of the bracket.This element has a rounded end and is free to rotate around its axis, thanks to its degree of freedom.
The thin blade and other elements are designed as single parts (Figure 2(d)), and they can be produced using standard techniques and commercial materials.

Dynamic simulation
The model implemented using Simscape Multibody 18 has the main purpose to simulate the main functionality of the end-effector involved in an ordinary task.This task consists of a grab and drag phase of the nominal size box specified in the project.Hence, the simulation includes the end-effector, which is fully modelled in a correspondent sub-model, and also external objects or elements that interact with the end-effector: the box and the ground, fixed on the world frame reference.Figure 3(a) describes a schematic of the numerical kineto-dynamic model with the main connections between parts, while Figure 3(b) shows the end-effector sub-model.The model incorporates the CAD geometries of all the parts corresponding to the architecture defined in Figures 1 and 2: all these parts are described as rigid bodies.To correctly describe the kinematic and dynamic response of the system, the simulation also includes a simple motions sequence and multiple body interactions (i.e.contacts), that are considered as hard contacts.
To control the movements of the end-effector, the numerical model includes a simple virtual control unit on the main body of the gripper, in order to simulate the robotic arm connection where the input force comes from.Firstly, the gripper is pushed against the box until the correct height is reached and the box is engaged.Secondly, the gripper slides parallel to the ground for a defined displacement.The ground represents the part fixed to the reference system.
To describe the three flexural springs, the model introduces a fictitious hinge with an elastic torque-rotation characteristic between the paddle and the main body (green dot in Figure 3(b)).The torque-angle relation of the elastic hinge reproduces the experimental response of the compliant paddle.Previous compression tests performed on a paddle prototype in Onyx material with the same geometry and inner carbon fibre reinforcement, not reported here for the sake of brevity, provided the elastic response (i.e. the force-deflection curve), from which we calculated the actual angular stiffness: specifically, we measured a stiffness value equal to 283 N•mm deg .In addition, frictionless hinges (red dots in Figure 3 The numerical model allowed us to investigate the response of different configurations in terms of kinematic and dynamic behaviour and to validate the structure in terms of functionality.A transient motion rendering, first result of the simulation, confirmed the mechanism functionality.In addition, contact forces were calculated during all the motion considering the end-effector only and detecting the ratio between the grabbing force and the actuating force. Figure 4 shows the main dimensions of the kinematics that we chose according to the technical specifications in Table 1, and highlights the stroke of the grabbing element.In particular, we selected the arms lengths in order to either maximize the bracket amplitude of movement and minimize the paddle stroke: a short path of the paddle helps to engage better the package.The final geometry of the end-effector (Figure 4) was chosen after a sensitivity analysis, not reported here for the sake of brevity, which investigated through the Simscape model different arms lengths of the kinematics, according to the technical specifications listed in Table 1: the proposed configuration was aimed at limiting the encumbrance of the end-effector while and avoiding any interference.
Figure 5 shows the results for the paddle and grabbing element obtained from the gripper Simscape simulation sub-model (illustrated in Figure 3(b)).
Neglecting the elastic reaction force of the paddle due to flexural springs, Figure 5(a) reports the force ratio between the grabbing element and the actuation on the paddle as a function of the vertical paddle stroke.As the paddle stroke changes, the arms lengths change varying the final ratio.To ensure no lateral crushes on packages, the actuation force is always higher than the grabbing force.Figure 5(b) compares the vertical stroke of the paddle with the horizontal stroke of the grabbing element.
As shown in Figure 5(a), neglecting the flexural springs elastic force, the grabbing-actuating force ratio slightly depends on the entire vertical paddle stroke.Both horizontal and vertical movements of the touching element and paddle, respectively, are limited by the box size, which defines the minimum grasp angle.This means that different box sizes cause the gripper to grab with different touching force per unitary actuating force on the paddle.From a theoretical point of view, this characteristic may lead to different lateral crushing forces applied on boxes of different sizes.However, the variation in the force ratio is lower than 25% along the entire vertical stroke.It comes with a negligible difference in the lateral crushing force per unitary vertical actuating force.Moreover, the force ratio introduces, in any position, a reduction factor which helps to safeguard the integrity of boxes.
The minimum grasp angle can be affected by box compliance because of small lateral deformations of the box.First of all, this characteristic strongly depends on whether the box is empty or not.Secondly, the box material assumes an important role, in particular in case of empty or partially empty boxes.These two variables have to be taken in account to define the maximum grabbing force.

Mechanical characterization
To investigate the mechanical properties of fibre reinforced 3D printed Onyx we performed experimental tensile tests on 3D printed dog bone specimens.The tensile tests adopted a flat dog bone specimen with a large fillet radius (Figure 6(a)) to avoid failures close to the test machine grips, while the length of the calibrated part was in accordance with UNI 10002 standard (L 0 > 5.65A 0 , where L 0 represents the calibrated length and A 0 is the failure section).Figure 6(b) shows the stress-strain curve for a dog bone specimen with a layup of 10 layers of carbon fibre at 0°.The curve shows a peak tensile stress of about 300 MPa and a Young's modulus of 2620 MPa.

Finite element modelling
Figure 7(a) presents the finite-element (FE) model of the compliant mechanism, which is the most critical part of the structure.The model used linear tetrahedral elements with an average side length of 3.3 mm, for a total of 112581 degrees of freedom.The model implements a linear elastic material constitutive model with the Young's modulus retrieved experimentally and a Poisson's ratio of 0.34.The model applied fixed constraints on the bayonet connection in Figure 2(a) and (a) prescribed vertical displacement equal to 18 mm to the free end of the paddle, to simulate its stroke.Figure 7(b) presents the prediction of von Mises stress contours: as expected, maximum stresses are located in the flexural springs and the predicted maximum value (about 22 MPa) is largely lower than the admissible stress which is about 300 MPa for fibre reinforced Onyx, according to preliminary monotonic tensile tests (Figure 6(a)).
In addition, we investigated the kinematics describing the rod and the bracket connected by a rigid pin (Figure 7(c)).The model included two hinges to constrain both elements: the first one on the pivot of the bracket, the second one on the free end of the rod (Figures 7(c)).To simulate the grabbing force, the model applies a horizontal load equal to 50 N (maximum admissible load according to Table 1) in the external bracket end. Figure 7(c) and (d) show, respectively, the rod-rocker arm mechanism mesh and von Mises stresses, which are again sufficiently  lower than the admissible stress.Since we have a high safety coefficient in static condition and the final application is intended for weak and light robotic arms, we avoided a further check of the fatigue mechanical resistance.

Prototyping
All the main components of the end effector (excluding screws and bolts) were fabricated exploiting Fused Filament Fabrication (FFF) additive manufacturing process through the Markforged II 3D printing machine. 20In particular, we used Onyx, a thermoplastic nylon 66 reinforced with micro carbon fibre. 21This material has optimal mechanical properties such as high elastic modulus and temperature resistance considering that it is based on a common thermoplastic material.Additionally, it has a high tenacity and resilience which ensures significative elongations before failure.
Thanks to the additive manufacturing process, the compliant paddle was produced in a single component including the main body and the three flexural springs.Additionally, we improved strength and fatigue resistance by adding long unidirectional carbon fibre layers in correspondence to flexural springs (Figure 8(a)), creating stiffer parts in the same component during the same manufacturing process.Another important aspect concerns the differential thickness management we used: in bulk components parts which present a very high thickness (more than 4 mm), we defined a filling strategy which avoids to waste material by creating an internal lattice structure that always ensures high stiffnesses, in order to reduce costs Figure 8(b)).This internal lattice structure reduces the overall density around 0.65 g cm 3 increasing the E d (Young modulus over density) ratio value.Figure 8(c) shows the entire prototype.
The prototype is characterized by an overall mass of 380 grams and it costed about 200 € mainly imputable to 3D printing cost.

Experimental validation
To validate the functionality of the prototype in terms of grabbing and releasing of a box, we performed some    simple experimental tests.The operation was done using an electromechanical testing machine and measuring vertical forces applied to the end-effector.The test investigated two boxes, with the dimensions reported in Table 2.
The test consists in two different phases: firstly, to engage and grab the box, the end-effector is pushed against the box at 10mm / min speed for a total stroke of 10 mm, which is a sufficient stroke to involve the mechanism and grab the boxes.Secondly, the end effector goes up to its initial position for the releasing phase, with the same velocity.Figure 9(a) shows the end-effector at the beginning of the test and Figure 9(b) when grabbing of the box was completed.This test has been repeated two times with two different box sizes.
Figure 10 illustrates the compression force on the paddle, as a function of the paddle stroke during the whole test.The red curve corresponds to the smallest box, the blue curve to the large box.
The curves in Figure 10 highlight two different behaviours.The first part, up to about 5 mm simply corresponds to the paddle reaction force, while the bracket is closing towards the package.When the grabbing element starts touching the pack side the force suddenly increases, according to the stiffness of the box wall.
The visible strong hysteresis effect on both curves is given by multiple factors.First of all, friction plays an important role: between the blade and the box and in all the joints of the kinematics it represents a significative cause of the hysteresis-like effect.Specifically, friction provides a force which is positive when the end-effector is moving downwards but negative when the box is being released.In addition, the compliant part of the body certainly introduces a considerable energy loss due to the Onyx internal friction.Moreover, some other involved elements help increasing the hysteresis effect, such as the touching element and the carton boxes.

Conclusions
This work proposed a new passive and compliant actuation end-effector with movable rigid parts.This solution moves towards simplicity and an unexpensive architecture in the contest of package grippers, directly exploiting the additive manufacturing capabilities creating multiple elements in a single flexible component.The simple architecture makes this tool easy, quick to build, and also scalable for bigger or smaller and weak robotic arms.The dynamic Simscape 18 model reproduces in detail the functionality of the end-effector and surrounding environment giving feedback about the whole functionality and other important predictable results.
The reported prototype is almost completely composed of a 3D printed fibre reinforced thermoplastic material.The experimental validation proves that this innovative, passive, and simple solution can be conveniently applied to industrial collaborative manipulators.

Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Figure 1 .
Figure 1.Conceptual model of the end-effector during the engagement (a) and after the grip (b).

Figure 2 .
Figure 2. 3D model parts.(a) bayonet connection, (b) the main body and the compliant paddle in a single component, (c) rod-rocker arm mechanism, (d) end-effector assembly.
(b)) connect the paddle with the rod (3 in Figure 3(b)), the rod with the rocker arm (4 in Figure 3(b)), and the bracket with the touching element (5 in Figure 3(b)).The thin blade is fixed with the main body (1 in Figure 3(b)).

Figure 4 .
Figure 4. Technical drawing of the final configuration of the end effector.

Figure 5 .
Figure 5. Force ratio between grabbing force and actuation force over the vertical paddle stroke neglecting the paddle reaction force (a) and horizontal grabbing element stroke over vertical paddle stroke (b).

Figure 6 .
Figure 6.Technical drawing of the specimen (a), tensile test on 3D printed fibre reinforced Onyx specimen (b).

Figure 7 .
Figure 7. Static analysis of the compliant component: von Mises stress contours.

Figure 8 .
Figure 8. Prototyping: (a) long fibre reinforced layers (b) detail of the internal lattice structure, (c) prototype of the end-effector.

Figure 10 .
Figure 10.Force over vertical displacement of the end effector for package 1 and package 2.