Assessing the applicability of modern robotics to nuclear arms control verification (2012-14)
US State Department BAA Key Verification Assets Fund Program
Project key dates
- 01/10/2012: Funds agreed. Project start date;
- Nov 2012: consultation meeting with customer, Washington DC;
- Dec 2012 – Feb. 2013: Development of technical questionnaire and interviews with inspection experts;
- Jan-Apr 2013: Visits to robotics centres of excellence;
- Sep 2013: End of year report to customer. Submission of request for continuation funds, for six months beginning 1 October 2013.
- Aug 2014: Project archived.
Background and project description
Whoever is involved in in verifying the dismantlement of nuclear weapons by a nuclear-weapon state in the future, the potential for violations of the NPT exists—whether unintentional or deliberate—in the form of information leakage. Data that might be revealed about nuclear warhead or component design would be highly proliferative if leaked.
To prevent this, could robotics be used to help inspectors verify nuclear warhead dismantlement in a future arms control agreement? The proposed project will explore a new approach to deploying technology in restricted access scenarios that are directly applicable to nuclear arms control verification and transparency. This project will investigate whether it would it be possible to design a robotic system to replace the human inspector, so that any classified information it collected and processed, never emerged from the facility.
If human inspectors, tasked with making a yes/no decision, were allowed to make verification measurements it is likely that they would want to base verification decisions on attributes such as mass thresholds or isotopic ratios of the fissile object(s) removed from the physics package, employing Non-Destructive Assay (NDA) technologies. On the face of it, since the NDA measurement devices, the analytical software and the hardware already exist, it might be possible to integrate them all into a semi-autonomous robot inspector. The possibility of using neural networks to train the system may also be explored.
This proposed piece of work addresses the overarching need; expressed in the State Department ‘VTRD Needs’ paper, under the heading of ‘Rethinking Verification in the Information Age’, that existing verification technologies are by themselves insufficient to address all the challenges that are likely to be posed by arms control and transparency initiatives of the future. This project will require information-gathering, expert interviews and analysis, on the subject of robotics for nuclear arms control verification and a systematic analysis of key needs and requirements—and existing and future possible solutions to those needs and requirements.