Skip to Main content Skip to Navigation

Robotic navigation based on servo task and sensory memory

Abstract : Robots are able to perform a vast group of tasks such as surveillance, mapping and navigation. This thesis focuses on the latter task which can be roughly defined as the ability of reaching a group of poses in the environment. This document presents three contributions that perform robotic navigation, which are introduced below : - The first of them is a system that navigates using a monocular camera; therefore, it belongs to the classification of the Vision-Based Navigation Systems (VBNS). Such system is called Photometric Teach & Repeat (PT&R) since it is based on the teach-repeat paradigm and Photometric Visual Servoing (PVS) to tackle the problem of navigation. The novelty in PT&R is that it autonomously generates its visual memory from a preobtained model of the scene, which is a textured mesh. This memory is used as reference to navigate and it is compared with the onboard measurements, acquired by the onboard camera, to control the robot. Although the experiments on PT&R demonstrated that the system is able to navigate using the visual memory, the generation of the latter is limited to thestraight line and does not explore other paths. - The second contribution is a VBNS inspired by PT&R, called the PhotometricNavigation System (PNS). The main difference between PNS and PT&R is that the former system explores other paths than the straight line using a topological approach. Furthermore, a simulated navigation evaluates the navigability of the path depicted by the visual memory. The success of the experiments’ results of PNS demonstrated its efficacy to autonomously plan a navigable path using the textured mesh and to navigate over the scene. PNS is considered a major contribution of the thesis.- The third contribution, also considered as major contribution, is a generalization of PNS which comprises a wide variety of sensors and models of the scene. This framework is called the Servo Navigation Architecture (SNA) as it tackles the navigation problem by conceiving the latter as a concatenation of successives positioning tasks using a sensory memory as reference
Document type :
Complete list of metadata
Contributor : ABES STAR :  Contact
Submitted on : Thursday, March 31, 2022 - 2:23:10 PM
Last modification on : Friday, August 5, 2022 - 11:22:17 AM
Long-term archiving on: : Friday, July 1, 2022 - 7:10:35 PM


Version validated by the jury (STAR)


  • HAL Id : tel-03626298, version 1



Eder Rodriguez Martinez. Robotic navigation based on servo task and sensory memory. Computers and Society [cs.CY]. Université de Picardie Jules Verne, 2020. English. ⟨NNT : 2020AMIE0065⟩. ⟨tel-03626298⟩



Record views


Files downloads