Integrated approach to active materials, process and control for dexterous manipulation with soft grippers – MANIMAT

The project methodology develops an integrated approach combining soft and active materials, robotic design, and advanced manufacturing and control, which are the partners' areas of expertise. Fluidic elastomers, shape memory alloys and electroactive polymers are considered in the project, with their partial or complete combination providing benefits in terms of expected performance.

The work is structured around three complementary axes. These represent the three main work packages of the project. They bring together teams from the Institut Pascal, ICube and Pprime laboratories:

• design and modelling, including the choice of geometries and materials adapted to the functional requirements of the fingers (WP1),

• manufacturing, the mastery of which is crucial to guaranteeing the quality, reproducibility and controllability of the parts produced (WP2),

• control, initially individual for each finger, then coordinated for all the fingers making up the gripper (WP3).

Work package WP1 impacts the fields of robotics design, smart materials, modelling and artificial intelligence. The methodology is based on real-time finite element modelling and deep learning in order to obtain rapid behaviour calculation models. Such models are used in the design and control of the gripper fingers. They must be integrated into the control architecture of the partners' multi-axis system for dexterous manipulation. The CAD design of a finger and the gripper is included in the work package.

Work package WP2 concerns the field of multi-material manufacturing for robotics. The scientific challenge here is to obtain, through an appropriate choice of process, a manufactured device with properties as close as possible to those specified. An innovative process based on experimental characterisations of representative composite samples is defined. The technologies studied are multi-material additive manufacturing and polymer prototyping processes, such as vacuum casting.

From a control perspective (Work package WP3), the proposed approach builds on the consortium's previous work and its real-time multi-axis control framework, based on industry standards for communication, programming and movement. The aim is to ensure a high level of coordination between joints and fingers in order to produce perfectly synchronised movement.

These packages are added to a fourth set of experiments (WP4) and an additional coordination lot (WP0). Tests are carried out to manufacture representative samples of active materials - flexible envelopes in order to establish rules for multi-material design and manufacturing. The self-sensing property of active materials is also studied and implemented. Objects of standard size, varying shapes and weights, are intended to be used to test the movements and forces involved.

WP1 Design and modelling

The initial work identified the possibilities for mobility through deformation and modelling of different types of active materials. A thesis developed a generalised method for analysing grippers and dexterous manipulation indicators, thus providing a basis for comparative analysis. A simulation-based control strategy for pneumatically actuated soft fingers was then developed using the SOFA platform's real-time finite elements. Two inversion methods were explored: quadratic optimisation and a neural network-based approach. Both methods were evaluated on several trajectories. Four internships enabled intermediate analyses of the use of self-sensing and of hybridisation between a fluidic elastomer and shape memory alloy for variable stiffness. These last two approaches opened up post-project prospects and secured new funding.

WP2 Multi-material manufacturing

An internship and a thesis developed several solutions:

• An integrated design of fingers and tools that anticipates the constraints of the manufacturing process from the design phase onwards.

• An original use of a low-pressure injection device inspired by the medical field, resulting in high moulding quality, characterised by the absence of “incomplete” defects.

• Consideration of the complex behaviour of silicones: hyperelasticity, viscoelasticity and sensitivity to degradation by the Mullins effect. These difficulties were addressed through a material characterisation phase and the implementation of a finger-training procedure to stabilise behaviour and limit the influence of the Mullins effect. A master's degree internship also evaluated the potential for integrating granular materials.

WP3-4 Control and experiments

• A test bench for evaluating prototype silicone fingers has been set up by a post-doctoral researcher. This bench is based on a modular industrial real-time control system using PLC, high-precision pressure regulators, and a motion capture system dedicated to detailed analysis of the deformation of the fingers being tested. This original test bench can test several fingers and supports hybrid actuation combining pneumatic and electric actuators.

• Around thirty soft silicone fingers, available in several versions, have been manufactured and tested. Four examples of a reinforced multi-material version have been developed for multi-finger grasping tests.

• A two-finger gripper was developed as an intermediate proof of concept, including a hybrid multi-active material version.

• A soft hybrid gripper (pneumatic bending/electric abduction-adduction) with three fingers and its controller have been assembled and are currently being tested.

Mastering the design, modelling, manufacture and control of soft grippers is an important scientific lever for handling fragile objects of various shapes. It opens up economic prospects in sectors such as agri-food, health and logistics, while promoting the integration of robotics into human contexts, with social and cultural benefits in terms of safety, ergonomics and acceptability. Variations in dimensional scale are possible if material damage limits and activation energy levels are taken into account. Improved human-machine interactions will enable decision-makers and employees across economic sectors, as well as the general public, to become more familiar with soft robotics, thereby contributing to the acceptance of its new technologies.

The project has resulted in benefits for any economic sector that uses object grasping, where traditional operations need to evolve towards more flexible methods, and where dexterous manipulation eliminates the need to reconfigure the robotic arm for adjustments of a few centimetres and tens of degrees, which can be potentially risky during human-robot interaction. The benefits can be applied to a wide range of situations involving humans, enabling in particular:

• better adaptation to repetitive tasks that are still performed manually due to the need for versatility,

• facilitation of unstructured tasks,

• additional level of safety thanks to softness, and thus:

• easier reorganisation of tasks in the shared workspace.

These situations will facilitate the achievement of Sustainable Development Goals by (i) increasing production efficiency and improving working conditions; (ii) promoting innovation and developing resilient infrastructure. The scientific fields benefiting from the project's spin-offs are robotic design, modelling of flexible and active materials for rapid computation, applications of artificial intelligence to mechanical architecture, multi-material manufacturing and control of soft robots for grasping. The scientific approach adopted remains transferable to other types of soft robotic devices.

Several avenues for development are being considered to extend and deepen the results obtained, including the integration of sensors. Two spin-off projects have already begun: an I-Site CAP20-25 project and a NAQ (Nouvelle-Aquitaine) project.

International peer-reviewed journals
[1] A. Pagoli, F. Chapelle, J.-A. Corrales-Ramon, Y. Mezouar, and Y. Lapusta, “Review of soft fluidic actuators: classification and materials modeling analysis,” Smart Materials and Structures, vol. 31, no 1, p. 013001, 2021, doi: 10.1088/1361-665X/ac383a.
[2] A. Pagoli, F. Chapelle, J. A. Corrales, Y. Mezouar, and Y. Lapusta, “A soft robotic gripper with an active palm and reconfigurable fingers for fully dexterous in-hand manipulation,” IEEE Robotics and Automation Letters, vol. 6, no 4, p. 7706- 7713, 2021, doi: 10.1109/LRA.2021.3098803.
[3] V. Loboda, A. Sheveleva, F. Chapelle, and Y. Lapusta, “Impact of an interface electrode charge and materials polarization to a conductive interface crack,” Mechanics Research Communications, p. 103923, 2022, doi: 10.1016/j.mechrescom.2022.103923.
[4] A. Pagoli, F. Chapelle, J.-A. Corrales-Ramon, Y. Mezouar, and Y. Lapusta, “Large-Area and Low-Cost Force/Tactile Capacitive Sensor for Soft Robotic Applications,” Sensors, vol. 22, no 11, p. 4083, 2022, doi: 10.3390/s22114083.

International communication
Invited presentation
[1] O. Piccin, “Manufacturing of soft robots with polymers and elastomers,” 10th International Conference on Biomimetic and Biohybrid Systems, Workshop on “Perspective for soft robotics: the field’s past and future,” Jul. 2021.

National communications
Invited presentation
[1] O. Piccin, “Design and Manufacturing: Basics and Importance of Material-Process-Shape Relationships,” Summer school on the topic of deformation in robotics, Lille, France, Jul. 2022.
Poster presentations
[2] M. Otti, A. Pagoli, F. Chapelle, B-C. Bouzgarrou, Y. Lapusta, “A soft dexterous manipulator integrating smart materials,” Summer school on the topic of deformation in robotics, Lille, France, Jul. 2022.
[3] B. Kraehn, L. Meylheuc, O. Piccin, “Design and manufacturing of a flexible gripper for dexterous manipulation,” Summer school on the topic of deformation in robotics, Lille, France, Jul. 2022.

The objective of the MANIMAT project (dexterous MANIpulation with active and soft MATerials) is to provide a solution to the scientific and technological challenge of obtaining a soft gripper adapted simultaneously to the versatility of object grasping with passive shape adaptation and to the modification of the object configuration from inside the gripper.

Indeed, with the evolution of service and industrial robotics, the requirements related to grasping are increasingly complex and go beyond simple clamping, which is sufficient in contexts where the environment and elements handled are mastered. With soft grippers, the scientific community is looking for maximum object grasping versatility. However, this type of gripper is currently not well suited for dexterous manipulation, i.e. the ability to control the position and orientation of objects during handling.

Our solution provides an answer to this challenge based on the following two hypotheses: (i) possibility of synthesizing a soft finger having a type of controllable bending adapted to dexterous manipulation and exerting sufficient contact forces, (ii) development from this finger of a multi-digital gripper having the necessary mobilities between the fingers and the palm. The methodology develops an integrated approach combining active and soft materials, robot design, advanced manufacturing and control, which are the partners' areas of expertise. Fluidic elastomers, electroactive polymers, and shape memory materials are considered in this project, their partial or complete combination enabling to bring advantages for the expected performances. During the project, we build innovative fast computational models that account for the behaviour of soft structures incorporating active materials. These models are necessary to design and manufacture, following a “multi-material” approach, the external shape and the internal arrangement of the finger and gripper. In addition, they must be integrated in the dexterous manipulation-based control architecture for articulated multi-axes systems already developed by the consortium. Manufacturing tests of representative samples of active materials - flexible envelopes combinations are also carried out to establish rules for multi-material design and manufacturing. The self-sensing property of active materials is also investigated and implemented. The project is discretized into four main work packages: multi-material design, multi-material fabrication, modular design, and experimental evaluation. It brings together teams from the Institut Pascal, ICube and Pprime laboratories.

The areas that can benefit from the project's outcomes are the robotic design using a multi-material approach, the modelling of smart materials for fast calculation, the applications of Artificial Intelligence on the modelling and optimisation of materials and robotic structures, the multi-material manufacturing, and the control of soft robots for grasping. The scientific approach adopted is transferable to other types of robotic devices.