List of Beneficiaries
Workplan tables - Detailed Implementation
WP1 : Hardware and software for bi-manual master and slave robots
The objectives of this work package are to : design, build and deliver a dexterous bi-manual slave robot, which offers high performance compliance and sensitivity, but which is suitable for nuclearisation to function in high radiation environments; design, build and deliver a set of state-of-the-art haptic controllers for dexterous bi-manual bi-lateral tele-operation; develop new methods and software for shared control – fusing autonomous trajectory planning with tele-operative control signals; develop algorithms and software to map human hand finger motions (sensed by the hand exoskeleton master) to slave robot hands and grippers which have very different kinematics.
WP2 : Perception and visualization
The objective of this work package is to develop autonomous perception methods which are robust enough to cope with the extremely challenging, variable and unstructured nuclear sort and seg scenes, but which are advanced enough to enable planning and control of autonomous grasps and manipulative actions with dexterous robots. Additionally, a control and visualisation suits will be built to provide infrastructure for bridging between the WP2 perception outputs and the WP3 manipulation outputs, and allowing the human operator to interact with the system.
WP3: Learning and control for autonomous manipulation
The objective of this work package is to improve the autonomy of the robots such that the workload of the human operator is reduced and finally, the throughput of the sort and segregate system is increased. While we think that a fully autonomous robot system for the sort and segregate scenario is not feasible, we are aiming for a semi-autonomous robot system that can be programmed by instructions from the human operator. We will use such system to increase the autonomy of the robots for a multitude of tasks that we will encounter in the sort and segregate scenario, i.e., autonomous grasp planning, autonomous bi-manual trajectory planning, robust reactive grasping and disentangling objects to remove an entangled object from the heap. For each of the considered tasks, we will investigate which types of human instructions are useful to render the programming of these tasks more intuitive. Moreover, we will use active learning to allow the robot to ask for further instructions in the case the robot is uncertain which actions to execute.
Learning to grasp from evaluative feedback, The robot tries out different grasps. The human evaluates each grasp with a score from 1 to 10. The robot uses this score as reward signal to improve its grasping strategy. The figure on the right shows a grasp with high score (9/10) and on the left an unstable grasp with low score (2/10).
The objective of this WP is to design software and apply development method such that is assures: (1) building high quality software modules which implement approaches from WP1-3, (2) integration of these software modules to create demonstration applicants from WP5.
The developed software modules are expected to maximise the following quality measures (see also e.g. [bass2003]): (i) conceptual integrity with respect to a range of algorithms in the scope of this project, (ii) reusability and flexibility to robustly operate in novel conditions, (iii) interoperability to fluently co-operate with other software modules, (iv) maintainability to minimise costs of bug fixes or module re-design due to novel requirements, (v) reliability to remain operational over time without failures, (vi) performance to minimise communication latencies of the modules, (vii) usability to meet the demonstration requirements in terms of functionality and support, (viii) portability to successfully operate with a range of 3rd part software versions, compilers, operating systems, hardware, etc., (ix) deployment readiness to meet a portent end-user requirements.
WP5: Demonstration and performance evaluation
The novel technology developed in this project must be evaluated and demonstrated in an industrial environment. Evaluation will enable the performance and the TRL to be assessed against a benchmark.
Sellafield Ltd. is currently producing a waste sort and segregate facility within the Box Encapsulation Plant (BEP) at Sellafield and an associated supporting test rig at NNL Workington. In the BEP facility, it is proposed that the operators will use manual tele-operation, which in this context is the use of the teach pendant only. There are associated safety, reliability and throughput concerns with this method because the operator is not in close proximity to the robot and gripper and so must rely upon camera views and/or distorted views through lead glass windows.
Any technology developed in this project must deliver and improvement with respect to the manual teleoperation in the BEP facility from the perspective of safety, reliability and throughput. Hence, manual teleoperation, as used in this facility should be used to define the benchmark.
Defining the benchmark requires replicating the BEP facility with a test bed and performing a range of human-subject performance tests. The results of these tests represent the benchmark. As the project progresses, new technology is system integrated into the test bed. The technology can then be evaluated by performing the same human-subject performance tests, which permits improvements to be quantitatively measured against the benchmark. Following the benchmark tests, these subsequent tests will be performed every six months.
Another purpose of the test bed will be to perform demonstration to key end users. This test bed will be constructed at the NNL Workington facility, which is only 31 km from the Sellafield site. The Sellafield site consists of a number of nuclear fuel cycle processing plants. These include fuel storage ponds, fuel handling plants, fuel separation reprocessing plants, waste processing plants and waste storage plants, including departments for asset care, decommissioning and new plant build. All of these plants and departments will benefit from the technology being developed in this project.
The location of the test bed within NNL Workington will be adjacent to the BEP Sort and Segregate test rig that is currently being constructed by Sellafield Ltd. in support of the BEP Sort and Segregate facility. This test rig will consist of three 500 kg payload Kuka robots and the sharing of information and data between the projects, provides a unique opportunity for cross-fertilisation.
NNL is in a unique position, as it understands the complexities of the site and has its Central laboratory and Post-Irradiated Examination facilities on the site, which are both radiologically active facilities. NNL works and collaborates very closely with Sellafield Ltd., which means that it has numerous contacts and networks within Sellafield. As Sellafield Ltd. is a significant customer of the NNL, many Sellafield test rigs are at NNL Workington and many Sellafield persons work at the NNL. NNL propose to use this in order to identify key persons and groups that could be invited to the demonstrations.
Additionally, the UK’s Nuclear Decommissioning Authority (NDA), which oversees the decommissioning of the UK’s legacy nuclear facilities has their main office about 18 km from NNL Workington. This is another key end user and again the NNL has numerous contacts and networks within the NDA, as they are another significant customer of the NSL.
WP6: Dissemination, exploitation and communication
The objective of this WP is to disseminate the work of the project as widely as possible to four different communities for four different purposes. The first is the academic community in robotics and computer vision, in order to allow other scientists to build on our work as effectively as possible. The second is the nuclear end-user community to promote the use of advanced robotics in this market domain. The third is the industrial and commercial robotics industry, to achieve innovation through knowledge transfer, leading to commercialisation of novel robotics hardware and algorithms created during the RoMaNS project. The fourth is the general public, to promote public awareness about the beneficial role for robots in society and also to deliver educational outreach to young people, using robotics as an exciting way to teach science, and promoting childrens’ interest in pursuing science and engineering careers, thus helping ensure the next generation of the robotics workforce in Europe.
The goal of this WP is:
To ensure timely and high quality achievement of project results through appropriate administrative co-ordination;
To ensure the quality control of project results and risk management of the project as a whole;
To provide timely and efficient administrative and financial co-ordination of the project and meet contractual commitments.
UoB (School of Mechanical Engineering) as co-ordinator will have overall responsibility for the project, act as an interface between the consortium and the European Commission, receive the Commission payments, transfer the shares to the individual partners, chair the meetings. UoB administration will serves an info desk concerning all administrative, legal, and financial questions. Precise management issues are presented in more detail in section 3.2 “Management structures and procedures”.