Cloud DEVS-based computation of UAVs trajectories for search and rescue missions

Abstract

This paper presents a new Cloud-deployable DEVS-based framework for optimizing UAV trajectories and sensor strategies in target-search missions. Its DEVS-support provides the framework with a well-established, flexible and verifiable modeling strategy to include different types of models for the UAV, sensor and target dynamics; for the target and sensor uncertainty, and for the optimizing process. Its Cloud deployability allows speeding up the hundreds of evaluations/simulations required to optimize this NP-hard problem which involves computationally-heavy models when solving real-world missions. The framework, carefully designed to handle different types of target-search missions, currently optimizes, using a multi-objective Genetic Algorithm, free-shape trajectories of multiple UAVs, equipable with several and various types of static/movable sensors, to detect a target within a given search area. The framework is implemented in xDEVS and automatically deploys models over a set of containers in the Google Cloud Platform using the Google Kubernetes Engine. The results of the paper show that our deployment policy allows us to speed up the computation up to 3.35 times and lets the operator simultaneously optimize several search strategies to determine the most appropriate for the given scenario.

Discrete Event Systems Theory for Fast Stochastic Simulation via Tree Expansion

“…Also, DEVS has been employed as a basis for the simulation of Markov Decision Process (MDP) models employing its modular and hierarchical aspects to improve the explainability of the models with application to optimization processes such as financial, industrial, etc. [46][47][48][49][50][51][52]. Capocchi, Santucci, and Zeigler [53] introduced a DEVS-based framework to construct and aggregate Markov chains using a relaxed form of lumpability to enhance the understanding of complex Markov search spaces.…”

Discrete Event Systems Theory for Fast Stochastic Simulation via Tree Expansion

“…Also, DEVS has been employed as a basis for the simulation of Markov Decision Process (MDP) models employing its modular and hierarchical aspects to improve the explainability of the models with application to optimization processes such as financial, industrial, etc. [46][47][48][49][50][51][52]. Capocchi, Santucci, and Zeigler [53] introduced a DEVS-based framework to construct and aggregate Markov chains using a relaxed form of lumpability to enhance the understanding of complex Markov search spaces.…”