Periodic Reporting for period 2 - RoMaNS (Robotic Manipulation for Nuclear Sort and Segregation)
Reporting period: 2016-11-01 to 2018-10-31
Cleaning up legacy nuclear waste, in the UK alone, represents the largest environmental remediation project in Europe. Decommissioning UK nuclear sites will take > 100 years, with costs >€200billion. Most EU countries share similar challenges, with worldwide decommissioning costs of order €1trillion. At least 20% of this work must be done by robots.
In particular, robotic arms and hands must be used to perform complex grasping and manipulation tasks. At one UK site 69,600 cubic metres of old nuclear waste must be examined and repackaged into 179,000 new storage containers. Robots must cut open very old containers with uncertain contents, sort through the contents, cut and compress waste where needed, then manipulate waste materials into modern storage containers for safe disposal. Waste must be “sorted and segregated”, so that very expensive high-level waste containers are not unnecessarily filled with less hazardous low-level waste.
Currently these tasks are performed by teams of human operators, painstakingly controlling the robot very slowly using joysticks. In contrast, RoMaNS demonstrated how advanced robotics methods can be used to solve these problems more quickly and efficiently.
Firstly, we developed new robot arms and hands, capable of highly dexterous and sophisticated behaviours, while still being resilient in high radiation fields.
Secondly, we developed “tele-presence” systems which enable human operators to control the remote “slave” robot while feeling the forces experienced by the robot when it contacts objects.
Additionally, we developed “autonomous” robot control methods. Now the operator can simply mouse-click on an object displayed on the video monitor. The robot’s vision system will detect the object and automatically guide the robot to grasp it.
Finally, we developed a variety of novel approaches to “shared control”, where both human and AI collaborate to control the robot, with the human controlling part of the motion, while the AI takes care of other aspects.
We developed very robust new algorithms for automatic planning and control of robot motions, to achieve autonomous grasping, cutting and manipulation, guided by robotic vision systems. We carried out the first ever deployment in history worldwide, of an autonomous vision-guided robot arm inside a real radioactive environment.
Finally, we implemented an advanced shared control system, combining both our advanced autonomy work and the novel teleoperation work onto a single demonstration. This was implemented on the RoMaNS industrial robot testbed, which we built inside the UK's Workington nuclear industry site, and it was tested by nuclear industry workers under full nuclear safety and security regulation.
Delivering such cutting edge technology into high-security sites of an extraordinarily conservative industry is very difficult. The RoMaNS project, due to close industry-academia collaboration, has had large success with overcoming these barriers, and transforming the industry's appetite for advanced robotics, especially autonomy.
Our academic research contribution has also been very large, with 60 international peer-reviewed publications spanning numerous disciplines from dynamics and control, through advanced vision and machine learning, to human factors and HRI.
Safe decommissioning of legacy nuclear sites is one of the major challenges faced by any society which has a history of engaging in nuclear activities. Improving our ability to automate such tasks, directly removes humans from performing hazardous work, but also accelerates the remediation of the overall societal hazard.
As well as using robots to help the nuclear industry, RoMaNS is also showing how nuclear applications can help the robotics community. EU robotics research expertise needs to find applications for advanced robotics which are: 1) needed in the near term; 2) economically large; 3) societally important; 4) can only be solved with robots. Nuclear clean-up offers an ideal opportunity to apply EU’s world-leading robotics and AI capabilities.
The RoMaNS project has made significant advances beyond the state-of-the-art, in diverse fields of study, ranging from mechanical engineering, to novel computer algorithms, to adapting psychology and human-factors methodologies to investigate human-robot interaction.
The project generated 60 peer reviewed scientific publications. RoMaNS researchers have given over 50 invited talks. We have also organised and/or hosted a large number of workshops on robotic manipulation, and applications for robotics to hazardous environments at major international conferences such as ERF, ICRA and IROS. We have collaborated with other EU experts to create a new euRobotics Topic Group on roboticas for nuclear and other extreme environments.
Industrial and commercial impact:
1) RoMaNS project partner CEA developed a novel robot arm, and created a new spin-out company, Sybot, which will market the robot commercially.
2) CEA also developed a new haptic exoskeleton hand/glove, which is being transferred to CEA spinout company Haption for commercial exploitation.
3) CEA also developed a novel dexterous 3-finger slave hand, which may also be commercialised.
4) Robot vision and autonomous grasping work of UoB has been transferred to industry through UoB’s knowledge transfer partnership with KUKA-UK.
5) Vision-guided autonomous robotic grasping methods, are being adapted for a variety of applications in manufacturing industry.
6) Partners UoB and NNL have extended the vision-guided robot arm control methods, to control a robot arm to deploy a powerful laser cutter, for cutting and dismantling highly contaminated legacy nuclear waste objects.
7) Most importantly, the (highly conservative) nuclear industry has become convinced that advanced robotics, including autonomous robot control, can be introduced safely and effectively in the near-term.
Educational outreach and public communication of science:
RoMaNS has engaged with the public in a variety of ways, ranging from “Pint of Science” lectures (where a scientist gives a lecture to the public in a pub setting), to being interviewed on the BBC Radio4 “Today” program.
We have also engaged in a large amount of educational outreach work. PI Stolkin ran one week robotics summer schools in 2015, 2016, 2017 and 2018 at the historic Royal Institution of Great Britain, where approx. 100 school children designed and built their own robots, motivated by the issues of cleaning up nuclear waste. We also worked with Royal Institution to run numerous robotics workshops with schools across the UK. We also collaborated with educational researchers in the US to create innovative curricula in which school children build and deploy their own environmental monitoring sensors.