Tesi di laurea magistrale
Navigation methods and planning algorithms with a haptic wearable assistive device for blind people
Corso di studi
INGEGNERIA ROBOTICA E DELL'AUTOMAZIONE
relatore Prof.ssa Pallottino, Lucia
relatore Prof. Bianchi, Matteo
relatore Prof. Bianchi, Matteo
- wearable device
Data inizio appello
Data di rilascio
All over the world 285 million individuals are estimated to be visually impaired: morespecifically 39 million are totally blind, while 246 have low visual capabilities. Accordingto the European Forum Against Blindness (EFAB), 218.513 Italians are blind and 5.967.180suffer from eye diseases. Blindness dramatically limits the quality of life of this people andtheir families, especially in term of autonomous navigation and privacy.In recent years, inaddition to classic methods used as navigation aids to visually impaired people, technologyadvancements have led to the development of portable and wearable navigation systems.The former ones are commonly defined asETAs (Electronic travel aids)which leverageupon sensory substitution, either auditory or tactile, to convey directional cues. The prob-lem of this kind of systems is the lack of wearability, which severely affect their usability.On the contrary wearable approaches propose devices that can be easily integrated withinuser’s body. However, both ETAs and wearable systems do not usually take into accountreal user’s needs except for the final validation phase, this aspect representing one of themain causes that reduce device acceptability and prevent their widespread adoption in realworld scenario. In this thesis, we considered a wearable device that can delivery normal andtangential force cues, which can be used to provide directional cues at the user’s arm. Weproposed to reverse the ”classical paradigm” starting from the investigation of requirementsfrom blind people through interview of visually impaired persons, to identify the most in-tuitive way to deliver navigational cues. Furthermore perceptual capabilities of correctlydetecting such directional cues were first tested with sighted participants. As a second step,we developed an integrated system consists of the wearable device, a RGB-D camera toacquire images from the external environment and a laptop to process images. We thenimplemented two planning algorithms for indoor navigation, a high-level and a low-levelplanning one devised from the state of art techniques. High level planning algorithm pro-vides solution, i.e. the path, the user must follow to move and reach a prefixed point withina structured environment, where everything is known, e.g. fixed obstacles (walls, furniture,doors, etc), starting and goal points. In a real scenario, we usually do not have the map ofthe environment in advance, neither have knowledge about fixed obstacles. For this rea-son low-level planning algorithm is needed since enable to avoid obstacles, both fixed andfloating ones. Relying an analysis on information acquired by the RGB-D camera and thencommanding a suitable stimuli reproduced by the device. Experiments with blindfoldedparticipants demonstrate that our approach, used in conjunction with a classic white cane,improves performance in an indoor navigation task.
1 file non consultabili su richiesta dell'autore.