Cardiff University | Prifysgol Caerdydd ORCA
Online Research @ Cardiff 
WelshClear Cookie - decide language by browser settings

Human perception capabilities for socially intelligent domestic service robots

Noyvirt, Alexandre 2016. Human perception capabilities for socially intelligent domestic service robots. PhD Thesis, Cardiff University.
Item availability restricted.

[thumbnail of 2016NoyvirtAPhD.pdf]
Preview
PDF - Accepted Post-Print Version
Download (12MB) | Preview
[thumbnail of NoyvirtA.pdf] PDF - Supplemental Material
Restricted to Repository staff only

Download (156kB)

Abstract

The daily living activities for an increasing number of frail elderly people represent a continuous struggle both for them as well as for their extended families. These people have difficulties coping at home alone but are still sufficiently fit not to need the round-the-clock care provided by a nursing home. Their struggle can be alleviated by the deployment of a mechanical helper in their home, i.e. a service robot that can execute a range of simple object manipulation tasks. Such a robotic application promises to extend the period of independent home living for elderly people, while providing them with a better quality of life. However, despite the recent technological advances in robotics, there are still some remaining challenges, mainly related to the human factors. Arguably, the lack of consistently dependable human detection, localisation, position and pose tracking information and insufficiently refined processing of sensor information makes the close range physical interaction between a robot and a human a high-risk task. The work described in this thesis addresses the deficiencies in the processing of the human information of today’s service robots. This is achieved through proposing a new paradigm for the robot’s situational awareness in regard to people as well as a collection of methods and techniques, operating at the lower levels of the paradigm, i.e. perception of new human information. The collection includes methods for obtaining and processing of information about the presence, location and body pose of the people. In addition to the availability of reliable human perception information, the integration between the separate levels of paradigm is considered to be a critically important factor for achieving the human-aware control of the robot. Improving the cognition, judgment and decision making action links between the paradigm’s layers leads to enhanced capability of the robot to engage in a natural and more meaningful interaction with people and, therefore, to a more enjoyable user experience. Therefore, the proposed paradigm and methodology are envisioned to contribute to making the prolonged assisted living of elderly people at home a more feasible and realistic task. In particular, this thesis proposes a set of methods for human presence detection, localisation and body pose tracking that are operating on the perception level of the paradigm. Also, the problem of having only limited visibility of a person from the on-board sensors of the robot is addressed by the proposed classifier fusion method that combines information from several types of sensors. A method for improved real-time human body pose tracking is also investigated. Additionally, a method for estimation of the multiple human tracks from noisy detections, as well as analysis of the computed human tracks for cognition about the social interactions within the social group, operating at the comprehension level of the robot’s situational awareness paradigm, is proposed. Finally, at the human-aware planning layer, a method that utilises the human related information, generated by the perception and comprehension layers to compute a minimally intrusive navigation path to a target person within a human group, is proposed. This method demonstrates how the improved human perception capabilities of the robot, through its judgement activity, ii ABSTRACT can be utilised by the highest level of the paradigm, i.e. the decision making layer, to achieve user friendly human-robot interactions. Overall, the research presented in this work, drawing on recent innovation in statistical learning, data fusion and optimisation methods, improves the overall situational awareness of the robot in regard to people with the main focus placed on human sensing capabilities of service robots. The improved overall situational awareness of the robot regarding people, as defined by the proposed paradigm, enables more meaningful human-robot interactions.

Item Type: Thesis (PhD)
Date Type: Completion
Status: Unpublished
Schools: Engineering
Subjects: T Technology > TA Engineering (General). Civil engineering (General)
Uncontrolled Keywords: Human sensing; Service Robotics; Situational Awareness; Navigation Planning
Date of First Compliant Deposit: 5 May 2016
Last Modified: 10 Aug 2021 13:22
URI: https://orca.cardiff.ac.uk/id/eprint/90508

Actions (repository staff only)

Edit Item Edit Item

Downloads

Downloads per month over past year

View more statistics