Presentation on theme: "A stochastic process algebraic abstraction of detection evidence fusion in tactical sensor networks Dave Thornley International Technology Alliance"— Presentation transcript:
A stochastic process algebraic abstraction of detection evidence fusion in tactical sensor networks Dave Thornley International Technology Alliance Imperial College London
The paper Example stochastic process algebra model that abstracts a decision maker and information sources Sequence taken from INCIDER example slides Appropriate detail to measure… –Situational awareness evolution over time –Fratricide risk as a function of timing of decision Full PEPA code and excerpts of pre-processing and Matlab scripts.
Analytic stochastic modeling Constructive simulation –produces example traces of evolutions of the system –Repeated Monte Carlo simulation enables estimation of distributions Analytic times stochastic modeling –Directly calculate distributions as equilibria, or transient functions of time
Scenario - simplified decision sequence Decision maker tasked with screen a la INCIDER –Hypothesize Blue and Red, refine opinion as necessary Available information –Briefing High confidence in Red Low confidence in Blue –Electro optical detect/classify with ranges Single detection event after Far range reached Single classification event after Mid range reached –Call HQ for update Imposes minimal confidence on Blue if none in region Increases confidence in Blue if local activity –Send scout to eyeball at Mid range Returns opinion to raise confidence in Red or Blue Entities are either Red or Blue, entering Far, proceeding to Mid, through Near to Decide We focus on fratricide risk – erroneous CID of true Blue entity
Modeling Scheme - MARS Mission Abstraction, Requirements and Structure (MARS) Federated Analytic Traffic drives the system (FAT) Passive SA components driven by intelligence sources Intelligence sources query physics Decision maker –interrogates SA –Selects actions that in turn drive some of the physics
Scenario components FAT ( (Sensors SA) DM ) Policy includes orders and tests Signals include EO interpretation, TID comms, Scout vision and HQ picture Evidence raises, lowers or sets confidence in Red and Blue hypotheses Model that can generate Red and Blue traffic, and the SA maintenance and decision making sequences for each has 1597 states with a 5 phase Erlang FAT transition process
Starting states for transient analysis (DM_Idle|SA_CBL|SA_CRM|HQ_Idle|HQ_History_Zero|Scout_Near| Scout_History_Zero|TID_Listening|EO_Idle|FAT_Blue_Far|FAT_TC_0) (DM_Gather|SA_CBL|SA_CRM|HQ_Idle|HQ_History_Zero|Scout_Near| Scout_History_Zero|TID_Listening|EO_Detected|FAT_Blue_Far|FAT_TC_0) From the instant of arrival of a Blue entity in range Far: 1(DM_Idle|SA_CBL|SA_CRM|HQ_Idle|HQ_History_Zero|Scout_Near|Scout_History_Zero|TID_Listening|EO_Idle|FAT_Idle|FAT_TC_0) 2(DM_Idle|SA_CBL|SA_CRM|HQ_Idle|HQ_History_Zero|Scout_Near|Scout_History_Zero|TID_Listening|EO_Idle|FAT_Red_Far|FAT_TC_0) 3(DM_Idle|SA_CBL|SA_CRM|HQ_Idle|HQ_History_Zero|Scout_Near|Scout_History_Zero|TID_Listening|EO_Idle|FAT_Blue_Far|FAT_TC_0) 4(DM_Idle|SA_CBL|SA_CRM|HQ_Idle|HQ_History_Zero|Scout_Near|Scout_History_Zero|TID_Listening|EO_Idle|FAT_Red_Far|FAT_TC_1) 5(DM_Idle|SA_CBL|SA_CRM|HQ_Idle|HQ_History_Zero|Scout_Near|Scout_History_Zero|TID_Listening|EO_Detectable|FAT_Red_Far|FAT_TC_0) 6(DM_Idle|SA_CBL|SA_CRM|HQ_Idle|HQ_History_Zero|Scout_Near|Scout_History_Zero|TID_Listening|EO_Idle|FAT_Red_Far|FAT_TC_2) 7(DM_Idle|SA_CBL|SA_CRM|HQ_Idle|HQ_History_Zero|Scout_Near|Scout_History_Zero|TID_Listening|EO_Detectable|FAT_Red_Far|FAT_TC_1) 8(DM_Idle|SA_CBL|SA_CRM|HQ_Idle|HQ_History_Zero|Scout_Near|Scout_History_Zero|TID_Listening|EO_Idle|FAT_Red_Far|FAT_TC_3) 9(DM_Idle|SA_CBL|SA_CRM|HQ_Idle|HQ_History_Zero|Scout_Near|Scout_History_Zero|TID_Listening|EO_Detectable|FAT_Red_Far|FAT_TC_2) 10……
FAT traffic progress
Evolution of SA and decisions
Future work Multiple entities per scenario –Collapse symmetries to control state space size Automated construction of fusion abstraction –e.g. CART-like analysis to select boundaries between states of approximately constant implications Automate model generation from mission descriptions –Situational awareness pattern triggers –Stochastic probe formulation
Thank you – questions… Tell me more about PEPA. Why are you not using Moebius/DNAmaca/xyz? What are you going to do about accuracy of timing? This is all very smeary. Surely the state space of any mission of realistic size will be unmanageable? Why did you keep rattling on about QoI? When will I get this as a tool?