Sensor Networks
NeTS-NOSS: RFID-Based Sensor Networks: Exploiting Diversity and Redundancy
PI: Samir R. Das
RFID (Radio Frequency Identification) devices or tags, in their basic form, are useful as identification and proximity sensors. When augmented with other environmental sensors, RFIDs can also be turned into other wireless sensors. This project addresses ubiquitous RFID-based systems enabled by dense deployment of RFID tags/sensors, and RFID readers, where the readers are themselves inter-connected using a wireless ad hoc network. The goal of the readers is to access the RFID tag/sensor data and relay the necessary information to an application running on a central computer system. The project considers design and evaluation of protocols and algorithms to exploit redundancy and diversity in the form of tag multiplicity, reader multiplicity, antenna diversity, and multiplicity of operating modes for the tags. Communication protocols for tag-to-reader and reader-to-reader communication, as well as mechanisms for efficiently answering higher-level queries from applications are o f particular interest. Expected results from the project include efficient protocols and algorithms to exploit redundancy and diversity in RFID systems, and demonstration of their efficacy by experimental or simulation-based methods. The project has significant potential for impact on practical utility of RFIDs, by improving the accuracy and efficiency of RFID access and queries. Potential applications include object tracking on factory floors, and sensing applications in home and office environments. This project is in collaboration with Nitin Vaidya, University of Illinois, Urbana-Champaign. (NSF)
MAC Protocol for Large and Dense Wireless Sensor Networks
PI: Wendy Tang
The recent development of small and affordable microsensors that can communicate with each other via radio transceivers have resulted in the rapid growth of wireless sensor networks.
We target large and dense wireless sensor networks that have large processing power and need simultaneous peer-to-peer communications.
We propose the use of a novel MAC (Multiple Access Control) Protocol, the Cayley Pseudo-Random (CPR) Protocol for large and dense wireless sensor networks. It uses a novel channel assignment scheme based on the pseudo-random connection of a dense Cayley graph. It explores the potential benefits of utilizing the entire communication bandwidth in a dense sensor network. Today’s existing MAC protocols mostly consider a single channel or a small number of multiple channels.
By utilizing all or most of the available frequency channels, the CPR protocol can support many, simultaneous peer to peer communications. Other features of the protocol include minimal collisions and a decentralized routing algorithm that avoids global time synchronization. These features all help to minimize communication overheads associated with collisions and time synchronization and therefore making the protocol energy efficient.
The design of this protocol is being integrated with hardware to achieve an optimal physical implementation. For application, we have developed a prototype of a wearable device, the Health Tracker, for wireless health monitoring. Currently, the device is capable of measuring body temperature and blood oxygen level. These measurements are then wirelessly communicated to a computer for monitoring. Our goal is to integrate our CPR protocol into the system. The implementation of the CPR protocol will support many simultaneous monitoring of hundreds or thousands of users. This is especially useful in the scenario of catastrophe or emergency situations in which large number of users needed to be monitored simultaneously. (NSF)
A Battery-Aware Approach for Energy Efficiency in Wireless Sensor Networks
PI: Yuanyuan Yang
Wireless Sensor networks are envisioned to emerge as a key tool for a wide spectrum of applications in various facets of human endeavor, ranging from outer-space exploration, medical treatment, emergency response, battlefield monitoring to habitat and seismic activity monitoring. These applications place an increasing demand on routing, communication and sensing functions of sensor networks. However, with battery technology lagging behind, the lifetime of sensors has not been improved as fast as processing speed of microprocessors. Therefore, carefully scheduling and budgeting energy in sensor networks has become an urgent and critical issue in sensor network design. Previously proposed routing protocols did not take fully consideration of some special energy-related characteristics of sensor networks, such as battery behavior, sensor node density, memory access and transmission power consumption, thus cannot achieve high energy efficiency in sensor networks.
This project systematically investigates the energy efficiency issues in sensor networks based on a new energy flow model. We are developing a battery-aware and density-aware cross-layer mechanism with dynamic power management that provides a comprehensive solution to achieve high energy efficiency in sensor networks. Specially, we focus on the following tasks:
1) develop battery-aware routing protocols to maximize the battery lifetime of sensor nodes by fully considering both the mathematical battery models and the collaborative nature of sensor nodes;
2) investigate several promising, realistic approaches for accurate traffic prediction on sensor nodes to reduce power consumption of sensor nodes while enabling them to function fully in sensing and routing activities;
3) design density-aware routing algorithms for sensor networks in a dynamic power management approach to reduce traffic redundancy and minimize power consumption of packet routing in the network;
4) integrate the above approaches into a cross-layer mechanism that globally optimizes various power saving solutions to achieve high energy efficiency and low packet loss rate network-wide. (NSF and ARO)
Research in Intelligent Fault Detection and Diagnosis
PI: Imin Kao
This research integrates smart sensors and methodology in the interpretation and collaboration of sensory information for the diagnosis of mechatronic systems by employing sensors and sensor networks. The objectives of this research include the following: (i) design and fabricate miniaturized, multi-function, and telemetric sensors, integrated to engineering systems, and (ii) interpret sensor data in decision-making process for intelligent fault detection and diagnosis (FDD) with both model-based and feature-based approaches. Using new smart multi-function, telemetric, and integrated sensors in systems will provide necessary sensory information (e.g., pressure, flow, and temperature) for the next-generation diagnosis. In addition, MEMS micro transducers will be integrated to the mechatronic arrangement for smart sensor-actuator systems. (NSF, Festo Corp.)
Continuous Medical Monitoring with Wireless Sensor Networks
PI: Ki Chon, Co-PIs: Samir Das, Himanshu Gupta & CR Ramakrishnan
In this project, faculty from Biomedical Engineering and Computer Science collaborate on continuous medical monitoring of patients using wireless-equipped micro-sensors. The goal is to develop platforms so that such monitoring can be done in an unobtrusive fashion in any setting – in hospitals, nursing homes, elderly care centers, medical triages, or even at homes. Our work will be particularly useful for people suffering from chronic conditions where continuous, low-cost monitoring is desired without affecting daily routines significantly. Regardless of the setting, the goal is to generate medically meaningful alerts for impending medical emergencies without requiring continuous attention of trained medical professionals.
Our medical monitoring system consists of four main components:
1) sensors
2) wireless motes with a small processor that can gather samples, perform any preliminary processing and communicate the samples
3) a real-time streaming database with long-term storage capabilities
4) a signal processing system with inference engine
In the basic form, the sensors sample physiological data (e.g., EKG, EEG, pulse oximetry, respiration, etc) from a subject and send it to attached motes with wireless communication ability. Other than physiological sensors, we will also use sensors that detect movements, pose, etc anticipating that they can provide useful information about the subject’s functions that can correlate with certain medical conditions. The motes relay the data to the central database to be used by the signal processing system and inference engine. The data to be relayed could be the raw samples or semi-processed or compressed data depending on the characteristics of the signal and available bandwidth in the motes.
The sensors and a set of motes will be worn by the subject. Due to battery power limitations, the motes have short radio ranges. Thus, multihop relaying in an “ad hoc network” will be necessary to route the data to the database/inference engine. Such motes will be pre-deployed to cover the area the subject may move around. Wide-area wireless links (e.g., using a cellular telecom system or even a wi-fi-based system) will be used possible when such coverage may not be available.
The following 4 research directions will be pursued in an inter-disciplinary context.
1) Wireless communication/networking
2) Collaborative sampling
3) Signal Processing/Inference Engine
4) Streaming Database
(NIH)
Docking of Autonomous Vehicles/Robotic End-Effectors via Wireless LOS sensors
PI: Goldie Nejat
In a typical two-stage autonomous-vehicle/robotic end-effector motion execution, the first stage yields an initial approximate movement toward a desired goal, whereas the second movement (“docking”) is a corrective fine-motion action based on high-precision feedback. The use of external task-space sensors, via passive or active sensing techniques, has been often advocated during the docking stage in order to reduce the detrimental impact of systematic errors on vehicle-motion accuracy. Frequently, however, even in the presence of task-space sensors, a vehicle’s pose cannot be determined accurately due to the inability of the proximity sensors to measure orientation as precisely as position. In order to address this conundrum, we have developed a generic multi-Line-of-Sight (LOS) electro-optical sensing-based short-range guidance system that solely utilizes indirect proximity measurements to determine vehicle-motion commands. A multi-LOS system can be configured using several LOS sensing modules to provide sufficient and accurate sensory data for guidance-based motion planning of vehicles/robots translating and/or rotating freely in multi-dimensional space. The minimum necessary number and the types of LOS in configuring such a system would depend on (i) the mobility requirement for the localization problem at hand and (ii) the motion range of the vehicle/end-effector. Each LOS sensing module is to consist of a RF transceiver used to send sensory data to an unmanned base station for processing, sensor fusion and motion-command generation. Motion commands are then generated by a guidance algorithm and sent (wirelessly) to the vehicle/robot for implementation. In addition, new sensor pose information is sent to each sensing module to update its location. Such a scenario, allows for optimal flexibility i.e., the elimination of communication cables to allow for easier access, and to avoid obstructions and limitation in mobility of the vehicle within the docking workspace. Furthermore, the required computing power of the vehicle/robot is minimized. Applications include autonomous docking of ships, unmanned ground vehicles, and manufacturing robots. The objective is to eventually have, via the internet, multiple users monitoring the real-time pose data for coordination of multiple vehicles/robots. (NSERC /International)
Embedded Sensors for Detection of Friction in Materials
PI: Chad S. Korach
The process of material wear occurs predominantly at the surface of materials in contact. Typical models of wear processes rely on the frictional forces that are generated between two contacting surfaces, be it for abrasive, adhesive, or oxidational mechanisms. As current applications dictate, developing wear models at the asperity or micro and nano-scales is necessary to fully capture the wear regimes in surfaces and structures on the order of lengths scales of the surface roughness. To this end, measuring the frictional forces at the very near surface will enable asperity-level mechanistic wear models to be developed. The models will aid in failure prediction of moving components in automation, robotics, and machining. This project utilizes embedded sub-surface detection methods that detect changes in stress intensity at and near the asperity contacts. A network of such sensors may communicate wirelessly to increase the speed of data retrieval and of analysis and processing. Utilizing characteristics of the mating surfaces, such as surface roughness and mechanical properties, mechanics-based physical modeling is used to develop relationships between the detected stress levels and the local friction coefficients. The goal of the research is to improve the predictive capabilities of component lifetimes and to increase the energy efficiency of component systems by reduced friction leading to substantial cost savings. (SBU)