AGRA: AI-augmented geographic routing approach for IoT-based incident-supporting applications

Energy-Aware Mobile Edge Computing for Low-Latency Visual Data Processing

“…Note that SPIDER has no strong assumptions on MANETs' topology such as unitdisk graphs or symmetric links. We remark that our SPIDER improves the baseline performance of the geographic routing and builds upon our previous work on Attractive, Repulsive and Pressure Greedy Forwarding (ARPGF) [34]. Similarly to ARPGF, our SPIDER alternates Attraction, Repulsion and Pressure forwarding modes.…”

“…RELATED WORKS Routing (SPIDER) approach that benefits from the physical obstacle knowledge derrived from the satellite imagery by using deep learning-based detectors [32,33] available at the edge. SPIDER builds upon our previous work [34] that theoretically guarantees avoidance of a local minimum as well as the shortest path approximation. Hence, this approach shows practiacal performance improvements over the recent stateless geographic routing algorithm which uses a notion of pressure forwarding for a local minimum recovery [35].…”

“…where ϕ(n) is the potential function of node n that allows us to have theoretical guarantees on packets delivery with O(3.291) approximation of the shortest path [34]; E(n)…”

“…geographical locations [34]. We discuss how nodes can get such additional obstacles' geo-information of their radius and center coordinates in the next section.…”

“…MEC FRAMEWORK ARCHITECTURE in the disaster-incident scene. Node n in MANET stores information about only those obstacles that locate in two obstacle's radius proximity from n to compute ϕ(n) (see Equation 4.2) later [34].…”