@article{GENERAL,
      recid = {7547},
      author = {Figueiredo, H and Cook, D and Biggs, W},
      title = {The Test and Evaluation of Intelligent Autonomous Systems  (IAS)},
      journal = {Conference Proceedings of EAAW},
      address = {2019-07-02},
      number = {GENERAL},
      abstract = {Intelligent autonomous systems (IAS) are set to become a  feature of future defence programmes, and their  introduction will pose challenges to traditional systems  engineering and acquisition practice. Whilst the need for  operational and technical assurance will endure, the need  to manage programmes that deliver iteratively and which are  continually evolving requires fresh thinking. New  increments may well be undergoing acceptance testing  against a backdrop of continual developments to higher  level concepts of operation – there may be no stable  baseline. Furthermore, the unbounded and potentially  non-deterministic nature of IAS means testing alone is  unlikely to provide satisfactory assurance, especially for  systems that are able to learn from previous mission data.  Lastly, at the very core of what is referred to as  human-autonomy teaming is a notion of trust by the operator  of the IAS, a trust which builds through development,  integration, training and deployment, implying a blurring  of boundaries between that which is technical assurance and  that which is operational. In response to these challenges,  QinetiQ are investing in UK test and evaluation  infrastructure and developing, with partners, approaches  that will mitigate the risks posed by these new  technologies. As outlined in this paper, these include  distributed live, virtual and constructive facilities,  brokered policy enforcement by software and developing a  body of trusted software components. The paper argues how  these developments address the identified challenges,  highlighting remaining gaps and drawing on evidence from  ongoing UK MOD research and development; it concludes with  the approaching investment and programmatic choices that  will need to be made to ensure best-for-enterprise  outcomes.},
      url = {http://library.imarest.org/record/7547},
}