Summaries - Office of Research & Innovation
Back Fundamental Issues for Observability of Adversarial Swarm Strategies
|Division||Graduate School of Engineering & Applied Science|
|Department||Mechanical & Aerospace Engineering|
Kaminer, Isaac I.
|Sponsor||Office of Naval Research (Navy)|
Successful defense against a large scale swarm attack on any Naval vessel has been identified by the CNO as critical to maintaining US naval superiority. In this proposal we directly respond to the CNOs challenge by addressing the problem of determining the internal co-operating strategies of an attacking swarm and by estimating parameters that define this strategy. This proposal addresses challenges of determining internal cooperating strategies for dynamic, large scale swarms; the computational tools we propose to continue developing to address these challenges; and the work we propose to establish impact assessments via the scalable implementation of large scale swarm strategies. This research is a continuation of the multi-year project we proposed to ONR in FY2017, to develop fundamental quantitative tools currently lacking for defense against autonomous swarms and thus to enable the use of informed real-time strategies against large-scale autonomous swarms, for defense and operations. In the last few years, ONR funded research has led to the development of multiple models for simulating swarm behavior. These projects consider challenges related to autonomous swarms from what could be termed an insider's perspective that is, from the perspective of creating, controlling, and driving one's own autonomous or semi-autonomous swarm.
Our research approaches the challenges of swarm defense from the opposite direction from the outsider's perspective of modeling the effects of and guiding the behavior of an enemy swarm. Such interactions are characterized by limited knowledge of the underlying swarm mechanisms, and also limited access to these mechanisms in the form of applying physical force or taking advantage of reactive behaviors. For real-time applications, computational demands and resource limitations often make the pursuit of globally optimal tactical strategies unrealistic. These restrictions necessitate instead the design of balanced strategies: robust, resource-efficient strategies, which exceed required performance metrics while retaining computational speed. The construction of such strategies requires the quantification of myriad fundamental features needed for cost/benefit analysis, including but not limited to: attacking swarm capabilities in terms of both fire and movement, uncertainty propagation due to limited information, efficacy of defensive force application, risk assessment, and intel gathering and observation.
Our previous work on observability has provided theoretical and computational tools, as well as strategies in the form of intruder trajectories, for the problem of detecting the internal cooperating strategies of an adversarial swarm by estimating a set of parameters that define a particular swarm cooperating strategy. Our work on herding has provided tools for generating optimal defender trajectories given parameter estimations for adversarial swarm strategies. Our objective this year is to utilize our developed estimation tools to fill a remaining strategic gap. Ultimately, robust defense strategies need to act without certain knowledge of attacker behavior model. This uncertainty is generated by uncertain model parameters, as tackled last year, but it is also found in the model itself. What is needed is the identification and estimation of features which transcend specific swarm models, features found across all models whose estimation yields decisive strategic insight. What we plan to do is to use the estimation tools developed last year to test the observability and utility of multiple proposed 'invariant' metrics, such as reactivity, centrality, and fragility. These will be tested on the Swarm Database we propose to develop this year - a database of a variety of different implemented swarming strategies.
|Keywords||Autonomous systems Observability swarms|
|Publications||Publications, theses (not shown) and data repositories will be added to the portal record when information is available in FAIRS and brought back to the portal|
|Data||Publications, theses (not shown) and data repositories will be added to the portal record when information is available in FAIRS and brought back to the portal|