Robot Assisted Real World Implementations of Sensor Deployment Algorithms

Abstract

Existing researches in the field of sensor deployment technologies are purely based on simulations & emulations and lacks the real characteristics of the Wireless Sensor Robot Network (WSRN) like heterogeneity of environment, dynamic obstacle distribution, non doubt on the correctness and reliability of the existing deployment techniques. To overcome these crucial challenges, we designed five novel algorithms, three for single robot scenarios namely Basic-SCAN, Opportunistic-SCAN (Opp-SCAN) and Focused Coverage SCAN (F SCAN) and two for multi-robot scenarios namely Cooperative Deployment Mode (CDM) and Divide & Conquer (D&C). In the first part of the thesis we present Basic SCAN & Opportunistic SCAN algorithms which are used to perform the initial deployment task by a robot. The robot enters the ROI from a known starting point in an unknown environment and executes either Basic SCAN or Opportunistic SCAN to deploy the sensors. The major difference between the two algorithms is the way in which the deployment takes place. In the section part of thesis, we propose an F - Coverage version of both Basic and Opportunistic SCAN algorithms which guarantees the added performance and coverage on top of deployment algorithms. Finally, in the third part of the thesis, we propose two novel methods of multi robot team based sensor deployment termed as Divide & Conquer (D&C) mode and Cooperative Deployment (CDM). The D&C mode is much simpler with less complexity when compared to CDM which is much intelligent and require more cooperation among the robots for carrying out the task. This work has been carried at three different levels. In the first phase, we implemented our algorithms on MATLAB simulation environment and prove that proposed model outperform existing deployment techniques backed by solid mathematical models. In the second phase, we extended our work by implementing them on WEBOTS (a near to real world robotics tool) and achieved consistent and outperforming results. Finally, we proved our claims by cross-compiling our algorithms and deploying them on real E-puck robots and executing it on the real test bed scenario. In order to evaluate the performance, we considered key factors like coverage, robot distance, completion time and message overhead and prove that SCAN based deployment outperforms the back tracking deployment in both single and multi-robot scenarios. In terms of coverage, the BASIC SCAN algorithm performs better than all other algorithms, but at the cost of higher total distance travelled. When it comes to distance and message overhead, the Opportunistic SCAN algorithm performs better than all others. In order to ensure the consistency of the proposed model, we have rigorously tested the algorithms in numerous scenarios with dynamically changing obstacle distributions with an optimum confidence interval. The major differentiating aspect and contribution of this work was to propose novel and standard sensor deployment framework, especially in the areas where there is no GPS and where human reachability is not possible.