Controllable Cluster of FlatSats Use Case Dr. Paul Darby, Assistant Prof. CAPE CubeSat Faculty Advisor, Electrical and Computer Engineering, University of Louisiana at Lafayette Lafayette, Louisiana 70504 – Research Area: Mobile Grid Computing Informa(onal Value : Highly Valued Informa(on (HVI) is that which is innova(ve, informa(ve, and provided under “just in (me” arrival, i.e. to the right loca(on(s) at just the right (me to be “ac(onable” and useful. The network design under QID focuses on the intelligent dissemina(on of informa(on in the background in support of the cyber-physical processes of the portable and/or mobile device cluster or swarm, termed as the “Mobile Grid.” Cyber-physical Collec(ve Ac(on : Not all informa(on is created equally. Its value is dependent upon the extent to which it can be used facilitate ac(onable decisions and the ensuing cyber-physical processes the decisions effect. These cyber- physical processes may include decisions at each device, or collec(vely at a number of devices (e.g. satellite, robots, mobile ground sta(ons) to sense, process, communicate, receive, actuate and/or to move. Ac(ons may be taken singularly or collabora(vely. Increasing Informa(onal Value : The value of informa(on, whether in packe(zed form or not, is affected by its informa(on content, size, accuracy, arrival (me (with respect to other events going on), its arrival loca(on(s), and whether or not it is informa(ve and ac(onable at the receiving device(s). The efficient dissemina(on of informa(on in a network has the poten(al to provide increased value from the perspec(ve of the mission process at hand. Demonstra(on Via Control of Disseminated Cyber-physical Checkpoints : Checkpoin(ng is more crucial in Mobile Grid (MoG) cyber-physical systems than in conven(onal grid compu(ng networks due to host mobility, dynamicity, less reliable wireless links, and the resultant frequent wireless disconnec(ons. In QID, the MoG takes on a new paradigm in that cyber-physical checkpoints are done as opposed to conven(onal checkpoints, allowing devices taking the place of the failed devices to intelligently con(nue the interrupted physical ac(on (i.e. sensing, movement, etc.) either singularly or collec(vely. For the checkpoint recovery process to work prac(cally, checkpointed data from each cyber-physical sub- process running on a given Mobile Host (MH), should be transmiTed to and replicated on other strategic MHs in the MoG, for safe storage and use if needed. So, effec(ve and highly robust dissemina(on of checkpointed data is crucial. Exploi(ng the wireless broadcast medium, Random Linear Network Coding (RLNC), has demonstrated significant gains over tradi(onal data rou(ng, but it alone may not be adequate for the MoG environment. QID, being Informa(on Value Aware, controls the behavior of the RLNC store and forward func(ons and adapts them based upon feedback from exchanged network metrics processed via machine learning techniques. RLNC’s LIstening-speaking Informa(on value flow Profile (LISP) can be modulated under QID to effect op(mal informa(on value flow when and where in the network it is needed in support Mission ac(ons. Tes(ng is underway via simula(on, and demonstra(on is intended for a FlatSat cluster working in conjunc(on with Dr. Darby’s patent pending ESG-Grid or virtual ground sta(on network. Ejec%on of 10 FlatSats Methodology: QID – Quiescent Intelligent Dissemination of Information in Mobile Grids (MoGs) Photo: CAPE 2 CubeSat Experiment Case: ESG-Grid u%lizes Computa%onally Augmented Random Linear Network Coding (RLNC) Fountain Code, under Reinforced Machine Learning Methodology, QID To Maximize Informa%on Value Flow 1 U Cube 10 1X10X10cm FlatSats ejection vector Rotation Progression CubeSat Ground Stations CA-RLNC Packet Dissemination in the MoG Fn F3 F2/ D S F1 D Fn-1 Ps Ps Ps Ps Ps Ps Batch Completion Probability. 0 0.2 0.4 0.6 0.8 1 1.2 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 P b Composite Probability, P C Probability of Complete Batch at one or more Forwarding Hosts n=2 n=4 n=8 Motivating Example Pb = l n " # $ $ % & ' ' i h " # $ $ % & ' ' Pc i (1 − Pc ) h−i i=m h ∑ * + , , - . / / l=k n ∑ l j h " # $ $ % & ' ' Pc j (1 − Pc ) h− j j =0 m−1 ∑ * + , , , - . / / / n−l ρ j = γ ji p i i =0 m−1 ∑ p 0 , p 1 ,, p i ,, p m−1 GF(2 8 ) h ≥ r ≥ m P c = P s × P w × P i Functional underpinnings! composite, success, listening window, innovative Best case analysis, yet we can still learn a great deal. RLNC Batch consists of m packets. The Essence of QID QID System and Control (partitioned) State Spaces. QID’s RL QI-Predictor Controller Middleware. P(A ⇐ A j ) = C(m, j) C(m, l) l ∑ , for all l Economy for Broad Appeal ESG Cloud CubeSat AXSEM Radio Board & PIC 24 Raspberry PI computer Smartphone & Free App Less than $ 200 ESG-Grid Computationally Augmented RLNC Rotation Progression CubeSat Ground Stations P 0 **** P 1 P k **** **** **** P m TS0 TS1 TS2 *** TSk CA: RLNC SDP ESG Cloud Communications Coordination Engine