Sie sind auf Seite 1von 9

INTRODUCTION

y Problem of mobile target trying to evade detection by one or more mobile sensors in closed n/w area. y Game b/w sensors(cat) &intruders(mouse) & ends ,mouse is in the sensing range of cat(s). y Cat tries to determine strategy to minimize detection time of the mouse & mouse tries to maximize. y A coordination protocol for the cats to catch the mouse by: * Forming a cohort to limit mouses degree of freedom in escaping detection &*Minimizing overlap in spatial coverage of cohorts members. y When RmRc, develop a dynamic prgming sol for mouses strategy, assuming info about the cats movt. y When Rm>Rc ,how the mouse determine its movt based on local observations of the cats movements. y Differences b/w SBC & CBC. 2

Let Us Start With The Topic


y Let country C bordered with another country C. For national safety, y

y y

y y

imp for C to monitor presence of intruder from C. Zone-vast impractical for C to ensure - all sensors working. With advances in robotics & aerial vehicles-feasible for zone to be monitored by balance of mobile robots etc. Imp to control coordinate routes to achieve effective area monitoring. Use group of surveillance sensors, to secure a network area sensors may be mobile. Ex: carried by robots. sensor have rangedetect an intruder. although they dont have sufficient density to ensure complete coverage. 3 cases: Blind mouse, seeing mouse with independent cats, seeing mouse and coordinated cats, all considered for both 3 strategies used-DP(uniform scan),Bouncing,RWP.
3

THE BLIND MOUSE:


y Rc Rm. In this mouse detected before it can see cat; so mouse considered a blind mouse. y Strategy of the mouse :blind mouse cant sense cats actual movt, to avoid detection, the mouse assumed to have info on the cats movt. i.e. mouse knows cats movt & sensing range. probability of finding at least one cat for any position in n/w & best design obtained

DP ALGORITHM

Fig2-Greedy strategy not optimal

EXPERIMENTAL RESULTS(1)
y Blind Mouse

*Benefits of DP sol-mouse: Refer table, performance of diff strategies-comparedone cat. Significant higher avg detection time-DP *Uniform presence matrix benefit cat: From table-uniform strategy reduced compared to non uniform RWP.-determined nature of scan, allow mouse predict cats movt. *Effects of Rc:1 cat & mouse in 500*500m area. Latter play DP,cat RWP/bouncing in 2 diff runs-fig 9.Rc increases avg detection time decreases *No of cats: Above case with cat playing either of 2,Rc=25m-fig 10,Nc increases, avg detection time decreases *Effects of Vc &Vm on DP sol:1 cat & mouse-fig11 Vc is high, avg detection time reduces & increases with Vm is high. Shows fast cat force mouse to be fast to avoid detection. *fig 12(a),12(b) gives idea of safe movt of mouse.

EXPERIMENTAL RESULTS(2)
y The Seeing Mouse, Independent Cats:

*Cats dont coordinate-plays bouncing strategy with Rc=5m&Rm=10m,simulation runs in100*100m area *Movt Direction for mouse: Vc=Vm=5m/s, 6 cats & a mouse-fig 13 shows min distance of mouse from each cat, as function of mouses movt direction -thick lines shows min distance of mouse from any cat. *Effects of Rc & Nc:10 cats fig 14-Result shows that feasible sol decreases exponentially as Rc increases(as cats sensing capability increases degree of freedom in choosing sol is restricted)-fig 15,16(a) Nc increases, both optimal min distance from any cats & detection time decreases. *Effects of Rm & Vm: fig 16(b)-Rm increases, mouse can detect cat earlier & avoid cats. Detection time increases first, beyond 9m,no significant changes. Far away cat info is useless.-fig 16(c)-shows detection time increases as mouses speed increases-faster mouse avoids cat more quickly. *Comparison of diff strategies: For Cats-best is Bouncing with uniform presence matrix, for Mouse depends on Cats: cat-RWP, then mouse-Bouncing, cat Bouncing then mouse-Centric

EXPERIMENTAL RESULTS(3)

Das könnte Ihnen auch gefallen