Periodic Reporting for period 1 - REMODEL (Robotic tEchnologies for the Manipulation of cOmplex DeformablE Linear objects)
Reporting period: 2019-11-01 to 2021-04-30
The REMODEL project aims to develop low-cost and affordable hardware and software tools as well as innovative methodologies enabling the implementation of manufacturing activities involving the manipulation of complex deformable objects in industrial automated production lines. REMODEL main target is the robotic handling of Deformable Linear Objects (DLOs) and the routing of multiple DLOs to create wiring harnesses, by deeply understanding their complex behaviour, including elasticity and plasticity, and embedding this knowledge, together with proper manipulation skills and perception, based on vision and tactile sensing, into a bimanual robotic system. In WIRES, where three of the REMODEL consortium partners were involved, the basic tools to enable the routing of electric wires and their fitting into standard electromechanical components by means of a conventional industrial manipulator with fixed base have been investigated and developed. REMODEL will extend the applicability of these robotic technologies to other domains, by introducing bimanual coordinated dual-arm manipulators and making them able to manipulate long cables as well as complex wiring harness made of multiple cables and with several branches and connectors at each end. Moreover, the presence of obstacles, multiple contact points along the cable path will be taken into account, allowing the usage of the developed technologies in the field of aerospace and automotive manufacturing. Aiming at showing the applicability of the REMODEL outcomes to other fields, the handling of silicon hoses for medical usage and its inspection for quality check will be also considered.
REMODEL aims to promote EU 2020 targets of a smart and inclusive economy, fostering socially sustainable, safe and attractive workplaces by creating high-tech companies involved in innovative manufacturing. This will increase EU industrial competitiveness and sustainability in a global world through R&I activities for the timely development of new knowledge-based production, technologies and systems, competitive and sustainable production plants, industrial automation, machinery and robotics.
- Modeling, identification, prediction and tracking of DLO shape and behaviour have been developed, and this models and methods will be improved and extended to the wiring harness case in the next period;
- Manipulation devices and robotic platforms specialized for the manipulation of DLOs have been developed for all the use cases, and preliminary tests have been executed for the evaluation of specific capabilities;
- 3D environment reconstruction, part identification, localization and deformation tracking is under development, and these parts of the system will be tested in the second reporting period;
- Control algorithms based on perception data and DLO models exploiting both data-driven and model-based methods, to join task planning and feedback control have been developed for the single-cable case and are now under development for the wiring harness case;
- DLOs grasp and manipulation control by vision and tactile data have been developed and tested, the fusion with proximity data will be considered in the next period;
- Automated task planning to generate task sequence for the robots automatically from digital product design have been developed and tested;
- Teaching by demonstration and tutored learning of robot skills for new assembly references and tasks are under development by means of different approaches and techniques.
The REMODEL robotic system is conceived to acquire information about the product directly from the CAD platform, without any human intervention, and to preprocess the product itself before the human intervention needed to complete the manufacturing. In this way, many of the most repetitive and stressful operations can be carried out by the robot, eventually also during night shifts, while the workers will be devoted to more qualified and decision-making activities, in which the added value coming from their expertise can be fully exploited. This includes also the teaching of the robot itself for the execution of the manufacturing process. The task learnt by the robot will be stored in a semantic way, then adapted online to the specific operative conditions and initial configuration of the deformable objects to be manipulated, thanks to the feedback provided by the sensory equipment. Moreover, specific tools, control algorithms, modeling and online identification techniques will be developed to manipulate DLOs, considering also the case of complex connection networks such as wiring harnesses composed by multiple branches. Due to the complexity of the DLOs’ behavior and of the manipulation tasks, machine learning techniques will be also investigated.