Document not found! Please try again

Wireless sensor networks: Enabling technology for ... - Semantic Scholar

9 downloads 9401 Views 417KB Size Report
Sep 18, 2006 - diffusion of wireless sensor network technology. This paper .... Examples of parasitic devices are bio-metric watches and body-tracking inertial ...
ARTICLE IN PRESS

Microelectronics Journal 37 (2006) 1639–1649 www.elsevier.com/locate/mejo

Wireless sensor networks: Enabling technology for ambient intelligence L. Benini, E. Farella, C. Guiducci Dipartimento di Elettronica Informatica e Sistemistica, Universita` di Bologna, Viale Risorgimento 2, 40136 Bologna, Italy Received 22 August 2005; received in revised form 22 February 2006; accepted 19 April 2006 Available online 18 September 2006

Abstract Wireless sensor networks are one of the most rapidly evolving research and development fields for microelectronics. Their applications are countless, and the market potentials are huge. However, many technical hurdles have to be overcome to achieve a widespread diffusion of wireless sensor network technology. This paper summarizes the trends of evolution in wireless sensor network nodes, focusing on hardware architectures and fabrication technology. We describe four generations of sensor networks (obtrusive, parasitic, symbiotic and bio-inspired), moving from the recent past to the future. We outline the key research challenges and the common themes in the field. r 2006 Elsevier Ltd. All rights reserved. Keywords: Wireless sensor networks; Embedded systems; Energy efficiency; Bio-sensors

1. Introduction Embedded electronic systems are pervasive: from alarm clocks to PDAs, from mobile phones to cars, almost all the devices we use are controlled by embedded electronics. Over 99% of the microprocessors produced today are used in embedded systems, and recently the number of embedded systems in use has become larger than the number of humans on the planet. Ideally, we should be able to interact with embedded systems through user interfaces that are an extension of our natural interaction paradigms (spoken language, gestures, facial expression, etc.). Embedded devices should ‘‘disappear into the environment’’ and behave ‘‘intelligently’’: intelligent systems respond to and even anticipate our needs by understanding what is implied by the sound of our voice, movements, expressions. Ambient intelligence (AmI) is the vision that technology will become not simply embedded, but invisible, fully hidden in our natural surroundings, but present whenever we need it, enabled by simple and effortless interactions [1–3]. The term AmI has been defined by the ISTAG (Advisory Group to the EU Information Society TechnolCorresponding author.

E-mail address: [email protected] (L. Benini). 0026-2692/$ - see front matter r 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.mejo.2006.04.021

ogy Program) as ‘‘the convergence of three major key technologies: ubiquitous computing, ubiquitous communication, and interfaces adapting to the user’’. Wireless sensor networks [4,5] (WSNs) are commonly recognized as one of the technological cornerstones of AmI. Agile, low-cost, ultra-low power networks of sensors can collect a huge amount of critical information from the environment. Using a biological analogy, a sensor network can be seen as the sensory system of the intelligent environment ‘‘organism’’. Sensor networks are irregular aggregations of communicating sensor-nodes, which collect and process information coming from on-board sensors, and they exchange part of this information with neighboring nodes or with nearby collection stations. Sensor network applications in AmI are countless, ranging from monitoring of ecosystems and industrial processes, to asset and people tracking, to maintenance of buildings, etc. [6]: each application domain has distinct requirements and constraint which will drive the development of sensor node architectures. In this survey we choose to focus our attention on one specific application driver, namely monitoring of humans and animals (in short, biomonitoring). There are several reasons for this choice. First, human monitoring is an extremely important field, both for ethical and commercial reasons. Second, its field of application poses extremely challenging requisites on

ARTICLE IN PRESS 1640

L. Benini et al. / Microelectronics Journal 37 (2006) 1639–1649

unobtrusiveness, security and safety. Finally, this is one of the AmI application fields where research and developments are most active. Design, implementation, and deployment of a WSN involves a wide range of disciplines and considerations for numerous application-specific constraints. In the last five years, significant progress has been made in the development of WSNs, and some WSN-based commercial products have already appeared on the market. The purpose of this paper is to give a snapshot on current development and future direction of evolution in this field. Our survey will emphasize node architectures and hardware technology, while only passingly mentioning software and network architecture issues. This choice of focus is dictated by space limitations. We refer readers interested in the very significant research challenges in software and network architectures to one of the many excellent surveys that have recently been published on these topics [7,8].



2. Trends in sensor networks In bio-monitoring, the ‘‘physical world’’ that has to be monitored is a living body. Hence, sensor nodes should be placed in close proximity of the subject’s body, and they constitute a body-area network. The evolution of sensor nodes for bio-monitoring is driven by the ‘‘disappearance’’ requirement, and it leverages all technology options available, from the ever-shrinking standard microelectronic technology, to the emerging microfabrication processes allowing the integration of heterogeneous devices onto a small physical volume. As a result, we can clearly see a direction of fast evolution. We summarize this trend by introducing an evolutionary sequence of four generations of sensor nodes, characterized by a progressively decreasing level of obtrusiveness:





Obtrusive. These devices are constantly perceived by the target subject because their size and weight is large enough to be a source of nuisance. Nevertheless, they are portable and they can communicate with other nodes or with data-collection gateways. Many current commercial devices are obtrusive: examples include halter electro-cardiographs and body tracking systems based on wearable cameras and markers (which limit movements to the area where markers are present). Obtrusiveness is dictated mainly because of two key issues: high power dissipation (a significant fraction of 1 W), which implies large batteries and/or short time between recharges, and a cumbersome sensory interface (e.g. electrodes glued on the skin, needles, fixed infrastructure required in the environment, etc.). Parasitic. These nodes are perceived by the subject as physical objects, but their size, weight and structure does not pose serious limitations to normal behavior. Examples of parasitic devices are bio-metric watches and body-tracking inertial sensors. The physical volume



of these nodes should not exceed a few cubic centimeters, and their weight should be in the order of the tens of grams. Considering the volumetric energy density of current battery technology, the power consumption of these nodes must not be larger than a few milliwatts. Several parasitic nodes have been recently commercialized, and these devices represent the current state-of-the-art in WSN. Symbiotic. Moving beyond the state of the art, the research community is pushing toward more aggressively scaled, cubic millimeter-sized devices (called ‘‘smart dust’’ [9]) which may enable a number of new in-body bio-monitoring applications. The technical challenges to be solved are, first, the implementation of autonomously-powered nodes, able to scavenge power from the body (temperature gradients, movement, in-body chemical reactions, etc.). Battery free operation limits average power consumption to a few tens of microwatts. Second, their size limitation imposes demanding requirements on process integration and microfabrication: wireless communication, electronic processing, chemical processing, microfluidic capabilities must be packed within a few cubic millimeters. Finally, even bigger challenges are posed by safety requirements: the nodes should be short and long term bio-compatible. We call these nodes symbiotic because they have a true mutual advantage relationship with the target organism. Bio-hybrid. As an end point of our evolution trend, we envision bio-inspired nodes, both from the architecture and from the technology viewpoint. The physical scale of these devices approaches a few cubic microns (or less), and the interface between the sensor target and the sensor itself disappears. Molecular engineering and nanotechnology may make these device a reality in the near future. Several exploratory research efforts have demonstrated that some of the key functionality required in a sensor node can be implemented by molecular-scale devices that are often engineered using bio-molecules. These devices operate autonomously, powered by chemical reactions inspired to biological systems. The construction process and the architecture of these devices will also resemble natural processes in biology: bottom-up self-assembly, self-replication and self-repair will be required in addition to safety and biocompatibility.

The trend and the fundamental characteristics of the four generations of sensor nodes are summarized in Table 1. In the following sections, we discuss in details the architectures and the main features of existing obtrusive and parasitic sensor nodes, drawing examples from commercial products as well as from innovative research prototypes. We then move to symbiotic and bio-hybrid devices, for which only incomplete, but very promising research solutions have been investigated.

ARTICLE IN PRESS L. Benini et al. / Microelectronics Journal 37 (2006) 1639–1649 Table 1 Summary of the four sensor node generations Node

Maturity

Power (W)

Size (m3 )

Obtrusive Parasitic Symbiotic Bio-hybrid

Commercial Prototype/commercial Research/prototype Concept/research

1–101 102 –103 105 –106 o107

103 106 109 1015

1641

are developing new chips that will combine several of the now separate functions, for example radio and processing capabilities, analog to digital conversion and memory functions. Clearly, the integration of functions on a single chip is advantageous, for both power and size. Research prototypes of system-on-chip (SoC) nodes are the first step toward the goal of symbiotic systems. Examples of SoC hardware include Smart Dust [16], the BWRC picoradio node [17], and the PASTA node [18]. 3.1. Sensor node architecture

3. The state of the art: from obtrusive to parasitic sensor nodes The design space of WSN is multidimensional: each practical implementation is the result of a complex tradeoff among design choices [6]. Characteristics such as resources, energy, form-factor and consequently costs, will guide our survey along the evolution of sensor nodes from shoebox-size devices to dust-like. Reducing physical size has always been one of the key design issues. Therefore, the goal is to provide powerful processor, memory, radio and other components while keeping a reasonably small size, dictated by a specific application. Shoe box: Commercial personal digital assistants (PDAs) contain significant computing power in a small package (about 200 cm3 ). They can be considered general-purpose sensor nodes for their multiple sensing (e.g. from video/ audio sensors to temperature and accelerometers) and transmission capabilities (Bluetooth, WiFi, infrared interfaces). They are usually based on RISC microprocessors and run off-the-shelf operating systems such as Windows CE, Linux, or real-time operating systems. These systems have maximum flexibility and programmability to adapt to various context and applications. Sensor nodes based on PDA-class hardware and software are power hungry and expensive in terms of manufacturing cost per unit. Usually this kind of nodes consists of a stack of base circuits comprising the processor, radio and power supply, which are coupled with the desired sensors, packaged in a ruggedized box for protection from hostile environments. Examples of this class of sensors nodes are WINS [10], iBadge from UCLA [11] and mAmps [12]. Dice: Dedicated embedded sensor nodes of the dimension of a matchbox or a dice are usually based on lowpower microcontrollers and small components. These platforms still use commercial off-the-shelf components, but they focus on small form factor devices, low-power processing and communication, and simple sensor interfaces. Examples includes the Berkeley Mote family [13], the UCLA Medusa family [14], Ember nodes [15]. Because of the simple microcontroller used, these platform typically support a very lightweight programming environment. Dust: At present, almost all prototype and commercial wireless sensor nodes are made with commercial off-theshelf components. However, some of the WSN companies

A generic sensor network node hardware consists of several subsystems [19]: a microprocessor, data storage, sensors, actuators, a data transceiver, and an energy source (see Fig. 1). In the next subsections, we investigate and analyze the main characteristic of processing unit, power block, sensor and communication unit. Subsequently, the final subsection introduce a case study of a wireless sensor system for motion capture with accelerometers, called WiMoCA. 3.1.1. Processing unit and memory A sensor node is a multi-functional unit performing many different tasks, from managing acquisition to handling communication protocol schedule and preparing data packets for transmission, after filtering, synchronizing and signal processing on data gathered from sensors. Thus, each sensor node requires processing and storage capabilities. The choice of the processing unit not only decides the intrinsic ‘‘intelligence’’ of the node but also influences its size and power consumption. A first and most common choice is to use generalpurpose microprocessors and microcontrollers of different power and resources. Intel StrongARM microprocessor, Texas Instruments MSP 430 and Atmel AVR microcontroller are examples of commonly used processors. Commercial PDAs, that we can consider as high-end sensor nodes, use the StrongARM microprocessor. These systems have maximum flexibility and programmability to adapt to various context and applications. However, a

Fig. 1. Sensor node functional components.

ARTICLE IN PRESS 1642

L. Benini et al. / Microelectronics Journal 37 (2006) 1639–1649

processor executing software is far less efficient (in terms of energy consumption, manufacturing cost per unit, and performance) than a fixed-logic application specific integrated circuit (ASIC). Moreover, the size of their batteries makes these devices obtrusive. Nowadays, microcontroller are essentially general-purpose computers on a single chip including not only memory and processor, but non-volatile memory and interfaces such as ADCs (analog-to-digital converters), UART, SPI, counters and timers. Thus, such devices can be considered both highly capable and quite inexpensive. They usually can switch between various operating modes, such as active, idle, and sleep modes, each characterized by a different amount of energy consumed. Recently, very interesting solutions in terms of power consumption have been designed, obtaining an ultra-low energy microcontroller consuming even less than 12 pJ/instruction in certain conditions [20]. However, microcontrollers do not have high clock speeds as microprocessors for PDAs and generally cannot run complex and resource-hungry operating systems. They can be programmed in assembly languages that are usually specific to the particular devices or family of devices. The main advantage is that assembly code take relatively less memory and run fast. Standard processing can be augmented or substituted with DSP (a dedicated digital signal processing unit), PLD (programmable logic devices) or ASIC or reconfigurable hardware that implements computational demanding or performance constrained tasks. DSP can be viewed as efficient specialized micro-controllers for performing the mathematics involved in manipulation of analog information, data compression and general digital signal processing for sound, images and video. PLD are, as the name implies, devices that can be programmed to fulfill user-specified tasks. Most popular PLD devices are the field-programmable gate array (FPGA). FPGAs are generally used to design special purpose functional units that may be very efficient for a certain limited task, thus being useful when computationally intensive signal or image processing are required. Use of FPGA in sensor nodes is an area of current exploration [19]. The main drawbacks with this technology are cost and power consumption, compared to microcontrollers. Although some low-power FPGAs are in the market, their consumption is not as low as a sensor node should be, and their cost is also excessive for most WSN applications. The development of ultra low-power, low-cost FPGs is a necessary condition for the diffusion of these devices in sensor nodes. ASIC implementation is usually characterized minimum by size and power consumption for the task targeted, as well as maximum speed. On the other hand, an ASIC does not have flexibility for node-level adaptation. A particular case of sensing nodes where ASICs have an important role is represented by video sensors, needing much more powerful processing capabilities w.r.t. a microcontroller. Vision sensors are effective for medium to long-range

sensing, because vision data is rich in information that is very easily accessible to human operators. However, they produce a large amount of data that must be processed to reduce the information content before being transmitted over low-bandwidth radio interfaces. On the other end, video compression itself can dominate the energy cost due to its high computational requirements. For this reason, exploiting hybrid architectures with programmable microcontrollers and ASIC accelerators helps decreasing power consumption [21]. Memory elements should be carefully designed as they affect in several ways node processing capabilities. They are used to store not only control software (such as an embedded operating system) but also application code and data. As such, they must have the capability of buffering data coming from sensor to allow their processing before transmission. However, low-power microprocessors have limited storage, typically less than 10 KB of RAM for data and less than 100 KB of ROM for program storage. This limited amount of memory consumes most of the chip area and much of the power budget. Designers typically incorporate larger amounts of flash storage on a separate chip [4]. Flash memory chips have become both very dense and relatively inexpensive due to their use in electronic cameras and voice recorders. Hence, they offer an attractive means for data storage on a sensor node. The ultimate limit in energy-efficient computing and storage can be achieved by ultra-low power circuits with supply voltage much lower than 1 V [22,23], with CMOS devices in weak inversion. In fact, energy analysis of CMOS circuits shows that the optimal power supply typically occurs when the devices are operated at very lowvoltage supply. Clearly, this approach requires memories working reliably in the same conditions. Recently, a 65 nm SRAM energy per operation for logic has been presented [24] which minimizes the operating voltage to 180 mV. ULP circuits have not yet become mainstream for sensor network applications mainly because of their strong sensitivity to temperature and their low-noise margins. 3.1.2. Power supply The power supply block usually consists of a battery and a DC–DC converter. Although batteries in the last decade have become smaller and less expensive, battery energy density does not scale exponentially as other technologies (disk capacity, CPU speed, available RAM, etc.). Nevertheless, for short lifetimes and small-size sensor networks, batteries are still a reasonable solution. Alternative power sources must be explored [25]. Fuel cells are a possible chemical alternative to lithium batteries. Research efforts is addressed at miniaturizing this energy source with the result to extend a node lifetime up to several times compared to usual batteries, through microfuel cells on a chip (output power 12 mW=cm2 [26]). However, many open research issues still remain to be addressed before microfuel cells can be used to power up sensor nodes [27]. Other alternative power sources are

ARTICLE IN PRESS L. Benini et al. / Microelectronics Journal 37 (2006) 1639–1649

nuclear ones, microheat engines and microturbines. In general, all these sources are in early stages of development and not yet available in a variety of configurations at low cost, are noisy and must be improved in terms of robustness. Moreover they pose safety issues. There are many option for harvesting energy from the environment instead of using energy stored locally on the node [28]. The most common example is the use of solar cells for outdoor systems [29]. They offer 15 mW=cm2 in direct sun, but their power density decreases in cloudy days and drastically reduces in indoor environment [30]. Vibration can be used as a source of scavenged energy, obtained through piezoelectric or electrostatic conversion. Vibration can have different order of magnitude whether the excitation is coming from industrial machines, vehicles or floors and walls and human-body movements. Studies [31,32] showed that we can expect a range of power density varying from 4 mW=cm3 provided by a vibrational microgenerators of 1 cm3 in volume and 800 mW=cm3 from machine-induced stimuli (2 nm motion at 2.5 kHz). Electronic systems harvesting energy from ambient-radiation sources are another possibility, but they need to be close to the radiating source or benefit of a large collection area. In any case they collect extremely limited power (less than 1 mW=cm2 ). Finally power can be harvested from humanbody motion, temperature, explicit interaction such as squeezing, shaking, pushing and pumping objects. One major challenge in this field is design of non-invasive power harvesting devices. A notable work in this area is the one of Paradiso et al. [32,33]. From this short overview, it is quite evident that no single energy source will fit all environments and applications, thus designers must choose one or a combination of power-sources depending on application requirements. Whether the power source is stored or harvested, power management at hardware–software level must be done both to optimize power consumption and monitoring available energy for power-awareness. Dynamic power management is mainly based on exploiting node components inactivity [34]. Unused devices or components can be shut down and activated when required. This can be done both at node level and at network level. Moreover, the communication protocol among nodes can drastically influence power consumption of the overall network [7], being the transceiver generally the major contributor to energy consumption. Minimizing the duration and the range of communication consistently with network parameters, together with algorithms for energy use by the overall network can be used for example to distribute transmission tasks and delay death of nodes with coordinator role. 3.2. Communication The shift from wired to wireless links for communication in sensor networks is the first revolutionary design element facilitating an unobtrusive introduction in the environment

1643

of sensory and smart embedded systems enabling ambient intelligence. The networking capability of WSNs is built up in layers. The link corresponds to a physical level. Radiofrequency (RF), acoustic, optical and infrared links are possible. Each has advantages and limitations. Optical systems can be based on laser light, LEDs or IrDA interfaces. Generally, they are not energy hungry but require a free line of sight between the transmitter and the receiver. Both LEDs and IrDA [35] enables short-range communication, while laser light can cover longer distances. Furthermore, it enables transmission at rates up to around 1 GHz without the need of an antenna. Another possible physical link can be based on ultrasonic carriers, but communication ranges are too short for WSN applications. However, acoustic communication of sensor data are employed heavily underwater where typical limitations and challenges are: acoustic frequencies limit the bit rates that can be transmitted; carrier frequencies from a few kilohertz to a few tens of kilohertz are commonly employed underwater; transmission rates are less than 10 kbits per second; multi-path propagation of signals is frequently verified. RF is the most common channel used in sensor network systems although it uses the often-limited energy in a sensor node at a relatively high rate. Therefore, a consistent effort is made on designing ultra-low power transceivers [36]. Several aspects affect the power consumption of a radio, including the type of modulation scheme used, data rate and transmit power. Dynamic power management techniques are possible switching among operation modes: transmit, receive, idle and sleep. Many prototypes and commercial solutions adopted very simple radio transceivers (such as the RFM TR1001 [37] or the Chipcon CC2420 [38]) with ad hoc and low-power network protocols for medium access control (MAC) and routing when necessary. Communication is not only about the physical link, but it regards also the protocol level. Fig. 2 shows different technologies representing the evolution of wireless communication in terms of bit rate, communication range and application mobility. General purpose systems such as ultra-portable devices (PDAs, cell phones) use radio communication based on WiFi, GSM, GPRS and Bluetooth standards. WiFi is particularly power hungry, thus being inadequate to WSNs. GSM and GPRS are controlled by telephone companies. Bluetooth is designed mainly for computer cable replacement, so it is not the best choice for a sensor network. Nevertheless, the relatively high data rate of Bluetooth suggests to use it for gateways between sensor nodes and infrastructure networks. Zigbee and IEEE802.15.4 [39,40] are standards developed for wireless sensor networking. IEEE802.15.4 is a low data rate solution from 20 to 250 kbps, depending on the frequency band used—compared to a nominal 1 Mbps for Bluetooth and 54 Mbps for Wi-Fi’s 802:11g technology. For sending sensor readings, which are typically a few tens of bytes, high bandwidth is not necessary, and ZigBee’s low

ARTICLE IN PRESS 1644

L. Benini et al. / Microelectronics Journal 37 (2006) 1639–1649

Fig. 2. Wireless communication standards and their characteristics [41].

bandwidth helps it fulfill its goals of low power, low cost, and robustness, thus augmenting sensors nodes lifetime to months and years. It is intended to operate in an unlicensed, international frequency band (868 MHz Europe, 915 MHz America, 2.4 GHz worldwide). IEEE802.15.4 defines the physical layer and the MAC. For these optimized short-range wireless solutions, the other key element above the physical and MAC layer is the network/security layers for sensor and control integration. 3.3. Sensors and interface electronics The trend for sensors is evolving from remote sensors, far from the actual phenomenon, to local microsensors. Under the same conditions of accuracy, sensitivity is enhanced in one case by augmenting the sensitive area to distinguish the targets from environmental noise (e.g. using a matrix of sensors or augmenting their size), in the other through the increased vicinity to the physical phenomenon allowed by the small form-factor. Microelectro-mechanical systems (MEMS) integrate mechanical elements, sensors, actuators, and electronics on a common silicon substrate through microfabrication technology. Putting together silicon-based microelectronics with micromachining, MEMS technology opens the way to complete systems-on-a-chip, integrating a complete sensor node with transmission, computational, storage and sensing capabilities onto the same die. Actual MEMS sensors already integrate some processing capabilities. In many cases output is already converted in a digital form and some signal conditioning is available. Inertial sensors from ST microelectronics for example integrate features such as I2C/SPI digital output interfaces, can be programmed to provide interrupt at certain threshold, embed self-testing capabilities. This is very attractive because the final sensor node can spare on many additional interface components, with evident benefits in terms of size, power consumption and reliability (although some effort in designing low-power conditioning circuits such as ADC has been recently done [42]). In conclusion, MEMS sensors and actuators are at present the best choice in terms of size, power consump-

Fig. 3. WIMOCA sensor node.

tions and cost. They augment the decision-making capability of microelectronic integrated circuits with ‘‘eyes’’ and ‘‘arms’’, to allow microsystems to sense and control the environment. 3.4. The WiMoCA system As a case study of WSN node design, we describe the WIMOCA dice-size ð20  20  18 mmÞ node, shown in Fig. 3, designed at the University of Bologna [43]. It is deployed in a wireless body area sensor network for a wireless/wearable distributed gesture recognition system for HCI and AmI applications, with nodes mounted on many parts of the human body. WiMoCA (wireless motion capture with accelerometers) in its basic setup is equipped with a tri-axial MEMS accelerometer, but many other sensors have been added (e.g. gyroscopes, bend sensors). It has a modular architecture to ease fast replacement and update of each component. It is composed by three sections namely microcontroller/sensors, RF and power supply. In the present prototype, the microcontroller is an ATMEL ATMEGA8 and the transceiver an RFM TR1001. We also designed a special node that takes care of interfacing the BAN with an external processing unit, e.g. a PC or workstation. This node, namely a gateway, uses the same microcontroller and RF transceiver but it is not equipped with a sensing unit. The gateway interfaces with the external processing unit (PU) through a Bluetooth wireless

ARTICLE IN PRESS L. Benini et al. / Microelectronics Journal 37 (2006) 1639–1649

link. Network support has been developed. Our system has been designed for real-time interactive applications with low-power requirements and for this reason we focused on minimizing software overhead by implementing our own component drivers and communication layer. The goal of WIMOCA is enabling natural interaction and context-aware services to inhabitants of a smart environment. WIMOCA nodes are worn to form a body area network (BAN), which captures and processes behavioral information about the user, enabling the environment to react consistently and contextually to the user’s personal needs and activities. This is achieved transparently from the user and proactively if possible, or through natural interaction if an explicit request is issued by the user. For this purpose, WIMOCA BAN is mainly based on inertial sensors for user activity detection and gesture recognition.

1645

The conceivable evolution of sensor systems could be depicted as developed along two phases. The first one is the application of conventional (micro) and non-conventional (nano) technology to the bio-physical world and, in particular, by patterning and disposing biological matter and by handling fluids in a microscopic, high-controlled way to separate and process small volumes of sample or selectively interact with single cells (see Section 4.1). The second phase will be dominated by the exploitation of bioscience for the creation of micro and nanoscale systems and materials with unique properties and functions. These systems will be characterized by properties of selfassembly, reversibility, adaptability, self-replication, possibility of interaction at the atomic scale [53] (see Section 4.2). Examples of existing systems and promising elements for the deployment in WSNs will be given for both phases focusing on the cited bio-monitoring applications (Sections 4.1.1 and 4.2.1).

4. Looking forward: symbiotic and bio-inspired sensor nodes 4.1. Symbiotic sensor nodes Some of the most exciting and challenging fields of application of future, wireless sensor networks are deeply related to areas as bioscience, biotechnology and nanoscience. We can mention for instance: (i) point-of-care portable or in-body, easy-to-use, stand-alone systems to perform medical analysis out of clinical laboratories [44,45]; (ii) in vivo controlled drug release systems, which can be ingested or injected into a human body and which must act to deliver appropriate quantity of drugs or other caring means in a pre-determined, self-regulating or realtime-controllable way [46,47]. In order to be adapted to these applications, systems should be autonomous in terms of avoiding as much as possible the direct contribution of a human operator to their functions—which can be achieved by integrating different devices and circuitry for data processing—and in terms of power supply. In addition, a key factor for the development will be the possibility to connect them to obtain information and to control them from remote. The hardware design of nodes for medical and biological applications has to address bio-compatibility issues and the high specific interaction properties at the molecular and atomic level. This can be achieved through development of several technologies, as: (i) advanced synthetic—or not— molecular receptors [48]; (ii) innovative three-dimensional membranes for highly-controlled molecular release [49]; (iii) bio-hybrid systems as bio-coated nanoparticles to vehicle drugs or transfection agents inside cells [50]; (iv) bio-inspired systems, to mimic peculiar characteristics of bio-systems as adaptability and self-regulation [51]; (v) smart organic surfaces to transduce electronically specific mechanical or optical stimuli [52]. From this incomplete list, it is clear that the high bio-nanoscience content involved in the design of new sensor systems will fuel innovation in many fields as molecular sensing, new transduction techniques of physical signals, new smart, multi-functional materials.

The first phase consists in the development of lab-on-achip, BioMEMS, micro total analysis systems ðmTASÞ. These systems are characterized by deep integration and miniaturization. Integration will exploit microtechnology to create structures using both standard microelectronics materials and non-conventional materials. Silicon, glass and metals will be employed to integrate data and signal processing, semiconductor sensors, microchannels, electro-mechanical microcomponents, thermal actuators. New-concept materials of polymeric nature (e.g., polydimethylsiloxane) will be preferred for their bio-compatibility, and their suitability to be patterned to create channels for handling and processing of biological samples. Aggressive integration will lead to (i) self-contained, easy-to-use, reliable analysis systems, by integrating on a single chip microfluidics and heating-control elements for the pretreatment of the samples (processing, separation); (ii) high-throughput and flexible systems, by integrating sensors of different nature on the same device; (iii) improved speed and reliability by allowing the same sample to be tested at the same time with different probes patterned on microsites into a unique substrate. Integration of batteries and communication modules (as outlined in the previous section) will lead to stand-alone, self-contained controllable-programmable onfield systems. Miniaturization is a key factor for the development of inbody devices as it reduces dramatically obtrusiveness. At the same time, it leads to some key advantages. Miniaturization of sensing element, sample pretreatment, actuators and delivery chambers reduce the amounts of reagents needed and, thus, costs necessary to conduct a chemical process: miniaturized handling volumes are in the nanoliter to picoliter range rather than the microliter range or larger in conventional experiments; moreover, small volumes lead to higher effective concentrations [54], thus reducing time

ARTICLE IN PRESS 1646

L. Benini et al. / Microelectronics Journal 37 (2006) 1639–1649

to result and improving performance. It should be noted that the ability to handle molecular receptors and therapeutic agents which are at the scale of the target molecules is needed to provide the necessary sensitivity and selectivity. 4.1.1. Examples and applications Point-of-care devices for the implementation of outpatient clinics to enable high level of care outside laboratories and hospitals is extremely desirable for reasons related to cost, comfort and efficiency. Microfabricated systems for in-body molecular detection and bio-parameters monitoring would be a breakthrough in this domain [55]. Many attempts have been done in the direction of implementing on-chip common analytical techniques. The most successful are gas chromatography [56], capillary electrophoresis [57] and polymerase chain reaction [58]. The latter is one of the most interesting approaches of microfabricated implementation which relies on the good properties of thermal conductivity of silicon and the possibility to easily integrate thermal resistances. Existing devices integrate only a few of the basic steps involved in molecular detection which is the most promising, challenging and complex analysis demand. The steps can be subdivided according to their function: (i) presensing steps: extraction, separation, amplification; (ii) sensing steps: sensing, transduction for the generation of electrical signals; (iii) read-out steps: signal conditioning and data processing. Handling of microfluidic samples for the integration of pre-sensing steps can be achieved by assembling basic fluidic components like electrokinetic and chemical separation channels, valves, mixing structures and chemical reactors. Even nanoliter-scale structures have been used for continuous flow, stopped flow, and thermal cycling reactions [59,60]. The integration and miniaturization of the sensing steps involves patterning and immobilization of molecular receptors on surfaces in an array format [61], the implementation of matrices of sensors, transducers and actuators on active substrates which can contain circuits for signal processing. Some examples report the use of array of microfabricated electrodes used as substrates of molecular-receptor sites for active receptor immobilization [62], target manipulation [63], electronic transduction of target recognition [64]. The latter demonstrates the electrical detection of chemical-labeled-DNA molecules on a silicon chip developed with standard 0:5 mm CMOS process. A section of the die which exhibits surface postprocessing to expose sensing gold sites is shown in Fig. 4. Recently, the feasibility of a label-free fully-electrical DNA detection technique easily integrable on chip has been also demonstrated [65]. Some very interesting applications of integrated semiconductor sensors for DNA and protein detection include the use of field-effect transistors where the gate has been substituted with an electrolyte conductive solution [66–68].

Fig. 4. Infineon technologies has designed a silicon chip for chemicalmediated electronic detection of DNA sequences. Gold electrodes, which can be modified with DNA sensing elements have been exposed on the chip surface and connected to the internal circuitry of the chip. Courtesy of Infineon Technologies.

The use of surface and bulk acoustic wave sensors [69], nanowire potentiometric sensors based on field effect [70] and microcantilever employed as surface-stress and gravimetric transducers [71] have also been recently investigated. The possibility to release drugs from in-body controlled microsystems could have a tremendous impact on medical procedures, eliminating the injection pain and infections due to frequent injections and offering a mean to reduce side effects and drug volumes by allowing a more precise and efficient delivery. The use of conventional micromachined devices, equipped with wells, microfluidics and circuitry has been tested. Santini et al. presented a multiwell silicon chip. The release of drugs from the compartments is controlled by an electrochemical stimulus which selectively dissolve covering gold membranes. This signal could be pre-programmed or controlled real-time by sensors coupled with the device [72]. Microfabricated devices are usually built on 0.5 mmthick wafers. This size allow the delivery of the device into the human body by ingestion ð 1 mmÞ but it is too large for injection ðo200 mmÞ, inhalation ðo100 mmÞ and releasing into circulation ðo10 mmÞ. An off-wafer technique has been proposed which release a micromachined part of the wafer ðo100 mmÞ with a small well containing drugs ð100 mmÞ [73] (Fig. 5). In vivo programmable chemical synthesisers which could provide drug synthesis on demand are very promising tools [74]. Coupling of microdelivery systems with sensing devices and processing will be crucial for pre-programmed or remote-stimulated controlled therapeutic treatment. Sensing functions based on wireless energy scavenging from remote sources have been demonstrated. RF [75,76] and ultrasounds have been employed [77] to design pressure sensors and bio-sensors.

ARTICLE IN PRESS L. Benini et al. / Microelectronics Journal 37 (2006) 1639–1649

Fig. 5. Required dimensions of in vivo drug-delivery devices are indicated with respect to the different delivery approaches. Standard and innovative microfabrication techniques which can provide the adapted level of miniaturization are reported. Off-wafer technique is based on surface and bulk micromachining of silicon and polymeric materials to obtain micrometric structures (an example has been sketched [73]). Bottom-up techniques refer to the employment of self-assembly and the specific biomodification of nanoparticles.

4.2. Bio-inspired sensor nodes The possibilities of sensing and actuation functions of bio-molecular-inspired devices relate on the use of biomaterials and on their unique properties. The core of these devices is made by bio-materials, bio-hybrids and two or three-dimensional assembled structures which are characterized by high bio-compatibility, adaptability, unique specificity toward targets and capability to change their state (often reversibly) by a switching stimulus (remote or in situ). Moreover, nanoscale systems are often needed to perform specific functions at an atomic and molecular level. Bio-materials which have been recently employed in innovative sensing or delivery tools are supramolecular systems, photo-chromic and thermo-chromic molecules, cell cultures, oligonucleotides (strands of DNA molecules) [52]. They are employed in association with inorganic molecules, deposited on solid substrates and entrapped in membranes.

4.2.1. Examples and applications Recent advances in in vivo therapeutics concerned the employment of isolated microchambers containing colonies of genetically engineered cells. They are able to produce and release in vivo substances as insulin [78] and stimulate locally anti-cancer mechanisms [79]. Cells have also been employed on conductive substrate to sense their electrical reactions to toxins or viruses [54]. Metal or semiconductor nanoparticles coupled with bio-materials are extremely powerful tools. They have been widely employed in sensing devices, drug delivery and remote control of molecular reactions. A silicon bio-chip for DNA-sequence detection has been developed by employing bio-hybrids gold nanoparticles-oligonucleotide probes (the

1647

Fig. 6. Bio-hybrid complex, gold nanoparticle/oligonucleotide for electronic detection of DNA sequences on a microchip [80]. The following steps: (i) target molecule detection (left); (ii) bio-hybrid binding and subsequent silver deposition (right) determine a 106 increase of the conductance between the two electrodes. Molecule detection (left); (ii) biohybrid binding and subsequent silver deposition (right) determine a 106 increase of the conductance between the two electrodes.

technique is shown in Fig. 6) [80]. Bio-hybrids bind to the target molecules which have been previously detected on an insulating substrate. Successively silver clusters can be grown on gold nanoparticles and provide a highly conductive electrical path between two near-deposited electrodes. Nanoparticles can incorporate drugs forming selfassembled structures [81]; they are among innovative synthetic methods of drug delivery which can overcome drawback related to the use of viral vectors in terms of costs, manufacturability and safety. The most interesting property concerns the fact that derivatives of functionalized nanoparticles are not detected by the immune systems, but they can penetrate the cell-membrane and targeting the nucleus. Once inside the cell, under lowpH conditions they dissolve and release their genetic content [82]. Nanoparticles can control molecular reactions of biomaterials immobilized on their surface by proving a local heating. Remote control of the hybridization of a DNA molecules by RF has been demonstrated. A 38-mer hairpin loop DNA molecule have been covalently linked to a 1.4 nm gold nanoparticle. The strand forms a duplex at room temperature, but experience a separation if heated. The reaction has been leaded by inductive coupling of a radio-frequency magnetic field to the gold nanoparticles and is completely reversible [83]. Electrically-controllable organic surfaces that can change reversibly their physical and chemical properties can be employed to release drugs with varying rate. A switchable surface obtained by an organic self-assembled

ARTICLE IN PRESS 1648

L. Benini et al. / Microelectronics Journal 37 (2006) 1639–1649

monolayer of mercaptohexadecanoic acid on gold changes its hydrophobicity as a consequence of small alterations of the substrate potential [84].

5. Conclusion In this paper, we surveyed the evolution of wireless sensor networks, starting from early shoebox-sized obtrusive devices and moving to forward-looking bio-hybrid devices based on nanoscale molecular engineering. A few common themes have emerged: the quest for a flexible and inexpensive micro-fabrication technology to reach the ultimate limits of integration and bio-compatibility, the strong push toward power reduction, from watts to microwatts (or less), the requirement for a holistic design approach, where all system components are jointly optimized. The opportunities are immense. WSNs can open huge markets in consumer applications, as well as in security and environmental monitoring. Looking forward, the evolution of sensor networks toward symbiotic and bio-inspired architectures could drastically improve the health conditions and lifetime expectation of a large number of people.

References [1] S. Marzano, E. Aarts, The new everyday view on ambient intelligence, Uitgeverij 010 (2003). [2] T. Basten, M. Geilen, H. De Groot (Eds.), Ambient Intelligence: Impact on Embedded System Design, Springer, Berlin, 2003. [3] W. Weber, J.M. Rabaey, E. Aarts (Eds.), Ambient Intelligence, Springer, Berlin, 2005. [4] D. Culler, D. Estrin, M. Srivastava, Overview of sensor networks, IEEE Comput. 37 (8) (2004) 41. [5] F. Zhao, L. Guibas, Wireless Sensor Networks: An Information Processing Approach, Morgan Kaufmann, Los Altos, CA, 2004. [6] K. Romer, F. Mattern, The design space of wireless sensor networks, IEEE Wireless Commun. 11 (6) (2004) 54. [7] J.A. Stankovic, T.E. Abdelzaher, Chenyang Lu, Lui Sha, J.C. Hou, Real-time communication and coordination in embedded sensor networks, Proc. IEEE 91 (7) (2003) 1002. [8] I.F. Akyildiz, W. Su, Y. Sankarasubramaniam, E. Cayirci, Wireless sensor networks: a survey, Comput. Networks 38 (4) (2002) 393. [9] J.M. Kahn, R.H. Katz, K.S.J. Pister, Emerging challenges: mobile networking for smart dust, J. Commun. Networks 2 (3) (2000) 188. [10] hhttp://wins.rockwellscientific.com/WST_Content.htmli. [11] hhttp://nesl.ee.ucla.edu/projects/ibadge/default.htmi. [12] S. Park, et al., Proc. Int. Symp. Wearable Comput. (2002) 0231. [13] J. Hill, D. Culler, A wireless embedded sensor architecture for system-level optimization, Technical Report, Computer Science Department, University of California at Berkeley, 2002. [14] hhttp://nesl.ee.ucla.edu/projects/ahlosi. [15] hhttp://www.ember.comi. [16] J.M. Kahn, R.H. Katz, K. Poster, Real-time communication and coordination in embedded sensor networks, Proceedings of the Fifth International Conference on Mobile Computing and Networking, Mobicom 1999, 1999, p. 271. [17] J. Rabaey, J. Ammer, J. da Silva, D. Patel, S. Roundy, Picoradio supports ad-hoc ultra low-power wireless networking, IEEE Comput. Mag. (2002) 42. [18] hhttp://pasta.east.isi.edui.

[19] J. Feng, F. Koushanfar, M. Potkonjak, System-architectures for sensor networks issues, alternatives, and directions, IEEE Int. Conf. Comput. Design (ICCD) (2002) 226. [20] B.A. Warneke, K.S.J. Pister, An ultra-low energy microcontroller for Smart Dust wireless sensor networks, Solid-State Circuits Conference, 2004, Digest of Technical Papers, ISSCC, 2004 IEEE International, vol. 1, 2004, p. 316. [21] J.M. Rabaey, M.J. Hammer, J.L. Da Silva, PicoRadio supports ad-hoc ultra low-power wireless networking, IEEE Comput. 33 (7) (2000) 42. [22] A. Wang, A.P. Chandrakasan, A 180 mV subthreshold FFT processor using a minimum energy design methodology, IEEE J. Solid-State Circuits 40 (1) (2005) 310. [23] B.H. Calhoun, D.C. Daly, N. Verma, D.F. Finchelstein, D.D. Wentzloff, A. Wang, Seong-Hwan Cho, A.P. Chandrakasan, Design considerations for ultra-low energy wireless microsensor nodes, IEEE Trans. Comput. 54 (6) (2005) 727. [24] B.H. Calhoun, A. Chandrakasan, A 256 kb sub-threshold SRAM in 65 nm CMOS, IEEE ISSCC06, 2006, p. 628. [25] S. Roundy, P.K. Wright, J. Rabaey, A study of low level vibrations as a power source for wireless sensor nodes, Comput. Commun. 26 (2003) 1131. [26] Z. Xiao, G. Yan, C. Feng, P.C.H. Chan, I.-M. Hsing, Integrated fuel cell micro power system by microfabrication technique, Transducers ’05, vol. 2, 2005, p. 1856. [27] C. Dyer, Fuel cells and portable electronics, Symp. VLSI Circuits (2004) 124. [28] M. Laibowitz, J.A. Paradiso, Parasitic mobility for pervasive sensor networks, Pervasive 2005, 2005, p. 255. [29] R. Amirtharajah, J. Collier, J. Siebert, B. Zhou, A. Chandrakasan, DSPs for energy harvesting sensors: applications and architectures, IEEE Pervasive Comput. 4 (3) (2005) 72. [30] V. Raghunathan, A. Kansal, J. Hsu, J. Friedman, M. Srivastava, Design considerations for solar energy harvesting wireless embedded systems, IPSN (2005) 457. [31] P.D. Mitcheson, et al., Architectures for vibration-driven micropower generators, J. Microelectromechanical Syst. 13 (2004) 429. [32] J.A. Paradiso, T. Starner, IEEE Pervasive Comput. 04 (1) (2005) 18. [33] N.S. Shenck, J.A. Paradiso, Energy scavenging for mobile and wireless electronics, IEEE Micro 21 (3) (2001) 30. [34] L. Benini, G. de Micheli, Dynamic Power Management, Springer, Berlin, 1998. [35] hwww.irda.orgi. [36] B.W.Cook, A.D. Berny, A. Molnar, S. Lanzisera, K.S.J. Pister, An ultra-low power 2.4 GHz RF transceiver for wireless sensor networks in 130 nm CMOS with 400 mV supply and an integrated passive RX front-end, IEEE ISSCC06, 2006, p. 370. [37] hhttp://www.rfm.comi. [38] hhttp://www.chipcon.comi. [39] hhttp://www.zigbee.orgi. [40] S.C. Ergen, ZigBee/IEEE 802.15.4 Summary, 2004. [41] hhttp://www.ictc.it/decina/i. [42] N. Verma, A.P. Chandrakasan, A 25 mW 100 kS/s 12b ADC for wireless micro-sensor applications, IEEE ISSCC06, 2006, p. 222. [43] E. Farella, A. Pieracci, D. Brunelli, A. Acquaviva, L. Benini, B. Ricco,´ Design and implementation of WiMoCA node for a body area wireless sensor network, IEEE International Conference on Sensor Networks, SENET05, 2005, p. 342. [44] hhttp://www.devicelink.com/ivdt/archive/02/04/002.htmli. [45] R. Langer, N.A. Peppas, Advances in biomaterials, drug delivery, and bionanotechnology, Bioengineering, Food, and Natural Products 49 (12) (2003) 2990. [46] D.A. LaVan, D.M. Lynn, R. Langer, Moving smaller in drug discovery and delivery, Nat. Rev. 1 (2001) 77. [47] D.A. LaVan, T. McGuire, R. Langer, Small-scale systems for in vivo drug delivery, Nature Biotechnol. 21 (10) (2003) 1184. [48] A.B. Sanghvi, K.P.-H. Miller, A.M. Belcher, C.E. Schmidt, Biomaterials functionalization using a novel peptide that selectively binds to a conducting polymer, Nat. Mater. 4 (2005) 496.

ARTICLE IN PRESS L. Benini et al. / Microelectronics Journal 37 (2006) 1639–1649 [49] N.A. Peppas, Y. Huang, M. Torres-Lugo, J.H. Ward, J. Zhang, Physicochemical Foundations and Structural Design of Hydrogels in Medicine and Biology, Annu. Rev. Biomed. Eng. 2 (2000) 9. [50] L. Brannon-Peppas, J.O. Blanchette, Nanoparticle and targeted systems for cancer therapy, Adv. Drug Delivery Rev. 56 (11) (2004) 1649. [51] R.S. Tu, M. Tirrel, Bottom-up design of biomimetic assemblies, Adv. Drug Delivery Rev. 56 (11) (2004) 1537. [52] A.N. Shipway, I. Willner, Electronically transduced molecular mechanical and information functions on surfaces, Acc. Chem. Res. 34 (2001) 421. [53] J.Z. Hilt, Nanotechnology and biomimetic methods in therapeutics: molecular scale control with some help from nature, Adv. Drug Delivery Rev. 56 (11) (2004) 1533. [54] R. Bashir, BioMEMS: State-of-the-art in detection, opportunities and prospects, Adv. Drug Delivery Rev. 56 (2004) 1565. [55] L.J. Kricka, Microchip, microarrays, biochips and nanochips: personal laboratories for the 21st century, Clin. Chim. Acta 307 (2001) 219. [56] R.H. Lambert, J.A. Owens, Utilization of a portable microchip gas chromatograph to identify and reduce fugitive emissions at a pharmaceutical manufacturing plant, Field Anal. Chem. Tech. 1 (6) (1997) 367. [57] V. Dolnik, S. Liu, S. Jovanovich, Capillary electrophoresis on microchip, Electrophoresis 21 (2000) 41. [58] I. Rodriguez, M. Lesaicherre, Y. Tie, Q. Zou, C. Yu, J. Singh, L.T. Meng, S. Uppili, S.F.Y. Li, P. Gopalakrishnakone, Z.E. Selvanayagam, Practical integration of polymerase chain reaction amplification and electrophoretic analysis in microfluidic devices for genetic analysis, Electrophoresis 24 (2003) 172. [59] A.T. Wooley, et al., High-speed DNA genotyping using microfabricated capillary array electrophoresis chips, Anal. Chem. 69 (11) (1997) 2181. [60] Y. Shi, et al., Radial capillary array electrophoresis microplate and scanner for high-performance nucleic acid analysis, Anal. Chem. 71 (23) (1999) 5354. [61] M.C. Pirrung, How to make a DNA chip, Ang. Chem. Int. Ed. 41 (2002) 1276. [62] P. Caillat, D. David, M. Belleville, F. Clerc, C. Massit, F. RevolCavalier, P. Peltie, T. Livache, G. Bidan, A. Roget, E. Crapez, Biochips on CMOS: an active matrix address array for DNA analysis, Sensors Actuators B 61 (1999) 154–162. [63] M.J. Heller, A.H. Forster, E. Tu, Active microelectronic chip devices which utilize controlled electrophoretic fields for multiplex DNA hybridization and other genomic applications, Electrophoresis 21 (2000) 157. [64] F. Hofmann, A. Frey, B. Holzapfl, M. Schienle, C. Paulus, P. Schindler-Bauer, D. Kuhlmeier, J. Krause, R. Hintsche, E. Nebling, J. Albers, W. Gumbrecht, K. Plehnert, G. Eckstein, R. Thewes, Fully electronic DNA detection on a CMOS chip: device and process issues, Proceedings of IEEE International Electron Devices Meeting, San Francisco, USA, 2002, p. 488. [65] C. Guiducci, C. Stagni, G. Zuccheri, A. Bogliolo, L. Benini, B. Samorı´ , B. Ricco´, DNA detection by integrable electronics, Biosensors Bioelectron. 19 (8) (2004) 781.

1649

[66] J. Fritz, E.B. Cooper, S. Gaudet, Peter K. Sorger, Scott R. Manalis, Electronic detection of DNA by its intrinsic molecular charge, Proc. Natl. Acad. Sci. 99 (8) (2002) 14142. [67] F. Uslu, et al., Labelfree fully electronic nucleic acid detection system based on a field-effect transistor device, Biosensors Bioelectron. 19 (12) (2004) 1723. [68] F. Pouthas, et al., DNA detection on transistor arrays following mutation specific enzymatic amplification, Appl. Phys. Lett. 84 (2004) 1594. [69] I. Willner, et al., Amplified detection of single-base mismatches in DNA using microgravimetric quartz-crystal-microbalance transduction, Talanta 56 (5) (2002) 847. [70] Z. Li, Y. Chen, X. Li, T.I. Kamins, K. Nauka, R.S. Williams, Sequence-specific label-free DNA sensors based on silicon nanowires, Nano Lett. 4 (2) (2004) 245. [71] J. Fritz, M.K. Baller, H.P. Lang, H. Rothuizen, P. Vettiger, E. Meyer, H.J. Guntherodt, Ch. Gerber, J.K. Gimzewski, Translating biomolecular recognition into nanomechanics, Science 288 (2000) 316. [72] J.T. Santini Jr., M.J. Cima, R. Langer, A controlled release microchip, Nature 397 (1999) 335. [73] S.L. Tao, M.W. Lubeley, T.A. Desay, Bioadesive poly(methyl methacrylate) microdevices for controlled drug delivery, J. Controlled Release 88 (2003) 215. [74] C.L. Hansen, E. Skordalakes, J.M. Berger, S.R. Quake, A robust and scalable microfluidic metering method that allows protein crystal growth by free interface diffusion, Proc. Natl. Acad. Sci. 99 (2002) 16531. [75] A.D. DeHennis, K.D. Wise, A wireless microsystem for the remote sensing of pressure, temperature, humidity, J. Microelectromechanical Syst. 14 (1) (2005) 12. [76] M.A. Fonseca, J.M. English, M. von Arx, M.G. Allen, Wireless micromachined ceramic pressure sensor fro high-temperature applications, J. Microelectromechanical Syst. 11 (4) (2002) 337. [77] Y. Porat, A. Penner, E. Doron, Implantable acoustic bio-sensing system and method, US patent 6432050, 2002. [78] T.A. Desai, et al., Microfabricated biocapsules provide short-term immunoisolation of insulinoma xenografts, Biomed. Microdevices 1 (1999) 131. [79] J.H. Brauker, R.L. Geller, W.D. Johnston, S.A. Levon, D.A. Maryanov, Implanted tumor cells for the prevention and treatment of cancer, US Patent 6156305, 2000. [80] S.J. Park, T.A. Taton, C.A. Mirkin, Array-based electrical detection of DNA with nanoparticle probes, Science 295 (2002) 1503. [81] H. Cohen, et al., Sustained delivery and expression of DNA encapsulated in polymeric nanoparticles, Gene Ther. 7 (2000) 1896. [82] D.G. Anderson, D.M. Lynn, R. Langer, Semi-automated synthesis and screening of a large library of degradable cationic polymers for gene delivery, Angew. Chem. Int. Ed. 42 (2003) 3153. [83] K. Hamad-Schifferli, J.J. Schwartz, A.T. Santos, S. Zhang, J.M. Jacobson, Remote electronic control of DNA hybridization through inductive coupling to an attached metal nanocrystal antenna, Nature 415 (2002) 152. [84] J. Lahann, S. Mitragotri, T.-N. Tran, H. kaido, J. Sundaram, I.S. Choi, S. Hoffer, G.A. Somorjai, R. Langer, A reversibly switching surface, Science 299 (2003) 371.

Suggest Documents