A Sensor-Based System for Motion Analysis of Parkinson’ Disease Patients. Recently, I have been participating in the design, implementation and testing of a wearable sensor system employed for inertial motion analysis in Parkinson’ disease (PD) patients. Our proposed system is part of an ongoing investigation whose final ojective is aimed at efficiently timing the self-administration of prescription drugs in people suffering from PD. To this end, we employ wireless sensors for capturing distinctive motion patterns that can indicate the onset of dyskinesia lapses as the prescription drug wears off. Our prototype incorporates three pairs of accelerometer/gyroscope sensors, each connected to a wireless node equipped with an IEEE 802.15.4 radio. Sensor data are transmitted to a computer that is employed for visualization. We have successfuly implemented fast-Fourier transform and principal component analysis algorithms in TinyOS 2.1.1 aimed at preprocessing data before being transmitted to a network coordinator in order to reduce the number of wireless packet transmissions. The next steps of this project will include incorporating supervised and unsupervised machine learning schemes that help in the state assessment process of PD patients, followed by clinical trials.
Handoff of Sensor Devices in Wireless Healthcare Environments. I am currently implementing and testing the performance of a cellular-like system of sensor devices using IEEE 802.15.4 radios for possible use future healthcare environments. The motivation of this research stems from the necessity have reliable solutions that enable patients in non-critical condition to be monitored at home, instead of at a hospital. There is an increasing interest in the applicability of pervasive computing technologies for healthcare, given the ever growing gap between the aging population worldwide, and the number of available medical doctors, nurses, and the corresponding facilities. The main objective of this research is to gauge whether networked wireless sensor devices around a human body can forward digitized vital signs data reliably inside a house or a nursing home. A number of challenges arise here due to the fact that these sensor devices are severely constrained in hardware resources. In particular, I am interested in understanding the impact of fast-fading radio signals over the reliability of such a system. To this end, I am currently developing and testing simplified handoff schemes that allow the coordinator of a Wireless Body Area Sensor Network to establish reliable data links with access points within a range of 10 metres or less.
Video over Wireless Sensor Networks. In this project, we pushed the limits of the available bandwidth available in IEEE 802.15.4 radios to forward raw Motion JPEG video. The main constraint here is that the maximum packet rate observed in multi-hop WSNs is seldom realized due to various factors. One limitation is that commercially-available hardware provides only a single, half-duplex radio interface. Additionally, most data forwarding schemes found in the WSNs literature employ a single radio frequency channel for their transmissions. Therefore, packet transmissions must be spaced out to avoid collisions in the wireless medium, decreasing the packet rate that can be attained. Therefore, we introduced an approach that multiplexes the transmission of individual data packets through two separate paths, each of which employs a different subset of frequency channels, thus enabling packet rates closer to the maximum value. We demonstrated the practicality of our approach over commercial WSN devices to forward constrained video feeds, and conducted performance evaluations to gauge its efficiency.
Programmable Middleware for Wireless Sensor Networks Applications Using Mobile Agents. For this project, we designed, implemented and tested the performance a middleware platform developed for interpreting mobile agent scripts in Wireless Sensor Networks (WSN). We coined our system "Wiseman", which was inspired by an earlier agent system originally devised for the coordination of distributed process in wired networks. To this end, we developed a simplified interpreter that can be embedded in resource-limited wireless sensor devices for processing text-based codes that realize diverse WSNs tasks. The Wiseman interpreter occupies 19 KBytes of embedded code and 3 KBytes of SRAM in the Crossbow Micaz. We studied the effects of employing the different agent migration methodologies supported by the interpreter and observed that codes’ migration delay, which can take as little as 235mS per hop. The main feature of this work is that it enables dynamic programmability of a WSN using human-readeable, text-based codes. A copy of my mobile agent system for WSN (WISEMAN) can be found here.
Service Directory Placement Algorithm (SDPA), and the Service Directory Placement Protocol (SDPP). SDPP was introduced as an alternative solution that enables directory-based service discovery in wireless ad-hoc networks. SDPP enables a service directory to be moved from one host to another in response to changes in the dynamics of a service discovery overlay system and the network topology, as opposed to settling for the common assumption of a service directory being confined to a single host. The objective is to incur in less bandwidth overhead by conveniently placing the directory in a network neighbourhood wherein a significant amount of service discovery queries are likely to be generated. To accomplish this objective, the directory placement problem is formulated as a Semi-Markov Decision Process. The challenge here is to find an efficient policy for directory placement by means of the Reinforcement Learning technique within the context of a low-mobility wireless ad-hoc network. Computer simulations using OMNeT++ have revealed the effectiveness of our approach.
Formation of Bluetooth Scatternets Using Mobile Agents. We investigated how the information obtained during the service discovery process can be employed to dynamically reconfigure the topology of Bluetooth multihop networks, known as scatternets. This work was based on the premise that efficiently coupling service discovery methodology with a scatternet formation protocol could help to improve the overall quality of service and quality of experience for the end user. Therefore, we designed a novel protocol for this purpose by relying on mobile programs that produce near-optimal results according to a predefined topology formation policy. This approach offers improved flexibility when compared to existing protocols that employ the message-passing communications model for topology management in wireless ad-hoc networks. The effectiveness of the proposed scheme has been demonstrated through computer simulations.
QoS-Routing for MPLS Networks Employing Mobile Agents. In the early 2000's, the introduction of the Multi-Protocol Label Switching and Differentiated Services schemes was poised to introduce powerful features to the Internet backbone, making a significant contribution to the overall end-to-end provision of Quality of Services. However, we deemed that to achieve such an improvement, these technologies required not only effective support from current routing algorithms, but also enhanced capabilities from them that were being explored. Thus, our contribution was targeted at supporting QoS-routing through the use of mobile software agents by realizing multipoint-to-point routing trees. We found through computer simulations that our proposed approach was feasible, and that actual networks could benefit from it. We also found that when simultaneous routing sessions take place, the number of agents in the network may increase abruptly, and so appropriate hardware design steps are needed to avoid queuing congestion at the swtiches/routers.