OVERVIEW

See also: NIMS Field Results, NIMS Indoor Lab Systems

NIMS is composed of a hierarchy of mobile nodes and supporting reconfigurable infrastructure. NIMS also introduces sensor diversity for adjusting and exploiting the sensor node population to reduce uncertainty in the 3-D environment. NIMS supports this with coordinated mobility for proactive and reactive, mobile probing of the environment for measuring uncertainty, establishing calibration, and finally optimizing fidelity. A primary NIMS embodiment includes mobile, aerial nodes moving on a cableway infrastructure that may be suspended in three-dimensional environments. NIMS node motion is horizontal and vertical, energy efficient, and precise. In contrast to wheeled or tracked vehicles that remain on the surface, NIMS explores the full 3-D space with quiet, low energy inspection capability. Also, NIMS may be placed in many typical environments that are non-navigable by wheeled or tracked vehicles.

NIMS Videos | NIMS Deployment Images

APPROACH

Infrastructure-Enabled Mobility
Physical reconfiguration achieved through proper forms of mobility may be required to circumvent sensing obstacles. However, here are conditions on the form of mobility that can enhance the full set of distributed sensing operational capabilities. For example, in addition to providing diverse location and perspective and providing navigation through complex environments, it is also essential that mobility methods be predictable and precise. Specifically, the mobility mechanism must reduce system-wide spatio-temporal uncertainty as opposed to increasing uncertainty as a result of errors or limitations in motion or navigation. As will be seen, this generally requires the introduction of an infrastructure.
The requirements for sensor mobility control for applications in environment monitoring are as follows: 1) Sensor mobility must permit a wide range of location and viewing perspectives. This requires the ability to change separation between sources and sensors over a wide range and choose a wide range of viewing or sensing perspectives. In the natural environment, this will require overhead viewing perspective. 2) Sensor mobility must be precise so that sensor location uncertainty does not degrade sensing uncertainty yet further. 3) Sensor mobility must accommodate complex terrain and surfaces that may incompatible with surface vehicle navigation (or may themselves be disturbed by vehicle passage). 4) Sensor mobility must also be sustainable in that energy requirements and the rate of system degradation must be low. At the same time, the impact of mobility on the environment (for example acoustic noise or powerplant exhaust emissions) must be minimized. 5) Finally, the sensor mobility system must also permit logistics for motion and delivery of components that may include physical samples, energy sources, replacement nodes, and other subsystems.

Figure 1.  (a) NIMS Systems include fixed and mobile nodes along with instrumented and adaptable infrastructure. NIMS nodes may be fixed to the infrastructure, may move on the infrastructure, or be delivered to locations and recovered by other nodes. (b) A schematic view of a NIMS deployment in a riparian stream environment with distributed sensing, sampling, and node transport

The addition of an infrastructure immediately addresses the above requirements in a way that would not be possible with other robotic forms. While many infrastructure types are anticipated, the “cableway” infrastructure discussed further below provides an example that meets these requirements, and is compatible with a broad range of environmental science applications. Further, it requires small logistics cost for deployment (that is a deployment cost no greater than deploying fixed sensors at elevation). The cableway infrastructure will be discussed with reference to the above requirements. 1) First, the cableway permits a wide range of location and viewing perspectives by allowing aerial suspension of nodes that may themselves probe a three-dimensional volume, as shown in Figure 1a. 2) The cableway provides precise sensor mobility. 3) Also, the cableway system allows sensor nodes to negotiate complex terrain. 4) The cableway system also enables sustainable operation. Energy requirements for mobility are modest and may be made vanishingly small when transport velocity is low and mass-balancing is employed to reduce gravity-work. 5) Finally, the cableway system provides a means to acquire physical samples and deploy sampling systems. It also permits low energy transport of massive payloads (if required) and permits the implementation of logistics for energy, node, and sample transport.

Distributed Physical Sampling
A limitation of the contribution of distributed sensors and even sensing diversity to information acquisition includes the limitations of fundamental sensing elements. A primary goal of enabling scalable deployment of distributed sensors has been that individual elements be compact and low in mass (to reduce the logistics cost of deployment) and to present low energy demands. Of course, it is also required that sensing elements provide reliability with adequate sensitivity (noise-equivalent signal spectral density) in the environment of interest. However, many environmental characterization problems involving chemical sensing (solid, liquid, or gas phase) confront the need for detection of trace elements within interfering media. In addition, these sensor systems may require subsystems for management of media flow and filtering. Also, in the event that trace element detection or isotopic analysis is required for an investigation, then compact sensors may not be available and laboratory-scale spectrometers may be needed. Taken together, these fundamental measurement requirements may limit the capability of conventional distributed sensor networks since the fundamental measurement may not be possible with distributed, compact sensors. However, again NIMS sensor diversity may be applied, but, now with physical sampling capability.

NIMS infrastructure enabled mobility provides another high precision method with the ability to acquire physical samples (solid, liquid, or gas phase) from the environment for transport to centralized assets for analysis. As shown in Figure 2 (below), this includes the ability to acquire a compact sample and in addition to re-provision sensors that may require entire replacement or replacement of materiel required for operations. In addition, the NIMS infrastructure can enable the accurate recover, re-calibration and replacement of sensor systems. The NIMS infrastructure effectively enables a distributed sensor to consist of two components, a remote forward area sampler, and a fixed base and possibly centralized, analysis system.

System Ecology
It is clear that physical reconfiguration (or a very high three-dimensional, volume density of deployed static sensor nodes) is required for enabling autonomous measurement and active reduction of sensing uncertainty in complex environments. Further, it is also clear that to achieve sustainable, precise, and capable sensing and sampling, infrastructure-enabled mobility is required. However, to achieve the ability to adapt to varying environments, and to scale to large deployments, an architecture is required that properly combines the advantages of fixed and mobile nodes and infrastructure. In particular, it is important to introduce hierarchy to enable scalability with the assets requiring the largest resource costs being sparsely distributed and yet supporting a high spatial density of less capable nodes. Further, this hierarchy of node architecture tiers must include standardized interfaces and methods for cooperation between tiers in order to exploit hierarchy in favor of scalable, sustainable, robust and high performance operations.  Specific applications may favor a larger distribution of elements in a specific tier and self-aware, self-adapting systems will adjust their own distribution to optimize application-specific resource costs and benefits.

The hierarchy of fixed and mobile nodes tiers along with interaction among tiers, forming a System Ecology, as shown in Table 1. Here, the resources exchanged between tiers along with system architecture define the System Ecology. These resources include data, samples, nodes assets, and energy. In some cases, resources are extracted from the environment (for example, sensor data, physical material samples, and solar energy) and in other cases these are supplied at the time of deployment.

Table 1. The NIMS System Ecology includes Fixed and Mobile Node and Infrastructure Tiers to enable the adaptation required to optimize the Dimensions of sensing fidelity, energy efficiency, and reach the largest spatio-temporal coverage. In this Table, the benefits contributed by each tier to these sensing dimensions are listed.

The lowest System Ecology level includes untethered fixed nodes, such as wireless sensor networks that can be precisely and autonomously deployed and maintained by NIMS for study of phenomena at appropriate spatial scales. The next level consists of tethered fixed assets such as wired suspension networks, mobility drive mechanisms, gateways (for energy and communications), position beacons, storage depots, and chemical analysis engines. Together, the three levels in this info-mechanical network provide a means for generating and transporting energy and information, where information may be in the form of bits or physical samples.

NIMS operation algorithms confront the challenges of rapid spatio-temporal formation of teams (linking multiple tiers) that enhance sensing and sampling capability by autonomously allocating appropriate tasks and roles. This is related to previous progress in homogeneous and to a lesser extent, heterogeneous teams for agents and robotics. NIMS, however, departs from previous development by including a System Ecology, organized hierarchically, with a diversity of communication pathways and sensing assets. NIMS operation also depended on a multi-objective optimization, (engaging all ecology dimensions), spatially distributed, and operates over a wide range of temporal scales (for example, defined by the speed of data transport and the speed of mechanical transport).

The System Ecology opens a complex design space that enables adaptation to application demands. For example, the relative demands of spatial sampling density and physical configuration latency both contribute to determining the required rate-distortion operating point. By exploiting the System Ecology, both at design-time and run-time, therefore, the distribution of static and mobile sensors with varying operation range may be selected to match evolving environmental and application demands. For example, at the cost of increased measurement latency, a slowly moving mobile sensor node may explore a region of space with a high sampling point density and at the cost of only few mobile assets. Alternatively, at the cost of node resources, static or mobile nodes may be relocated and remain resident at locations that best benefit the sensing task. Such adaptations may evolve in time and space.
Finally, the System Ecology may include both infrastructure-supported nodes as well as unsupported and freely moving surface-bound or aerial robotic systems that further augment monitoring capability.

Reactive and Proactive Coordinated Mobility
Sensor diversity enables a method for determining and reducing sensing uncertainty. Now, since sensing uncertainty arises from limitations associated with physical configuration of sensor network nodes, then physical reconfiguration in the form of articulation, mobility, and the distribution of new sensing assets is required for reducing uncertainty. However, this then creates the requirements for systems that combine sensor diversity based self-awareness to enable coordinated mobility for measurement of sensing uncertainty and methods for effecting its reduction.

The relocation of sensing assets may be in rapid response to a triggering event that results from physical phenomena directly or model-based analysis of phenomena. This exploits progress in multi-robot operations, however, with the new features of NIMS constrained and precise mobility. This is enabled by reactive coordinated mobility. However, the NIMS system many also proactively probe the sensor network environment to determine the spatio-temporal regions where sensing uncertainty is expected to be large. This forms a proactive coordinated mobility operating regime.

A domain specific application applies to the problem of detection of mobile objects (sources) in natural environments. For example, acoustic sensors may typically be deployed in environments where acoustic propagation is highly variable with source-sensor range, terrain foliage, and meteorological conditions. Yet, it is at the same time required that detection of sources remain effective throughout these variations. Figure 6 illustrates an example where acoustic sensors are able to detect that sources have moved through their area, however, due to obstacles to sensing, these acoustic sensors are not able to support detection of an important large aggregation of sources. A combination of both event detection and an awareness of sensing uncertainty level produce a trigger for reactive coordinated mobility of mobile sensors and redeployment of nodes. Figure 3 illustrates that coordinated mobility enables a potentially drastic advance in performance by optimizing sensor population and position with both mobile nodes (imaging devices with powerful viewing perspective) and redeployed sensors. In this example, a static node acts as a trigger and the system is able to physically relocate sensing assets to acquire data at higher resolution and diversity at the trigger

Examples of proactive coordinated mobility include those where mobile nodes may analyze historical data (obtained via sensor diversity algorithms) and realize that particular areas are mapped with less certainty, causing them to revisit those areas at higher frequency until they are better mapped. Another reason for opportunistic motion is exploration, where in the absence of triggers from the static sensors on the ground, the mobile nodes proactively explore their configuration space, to detect phenomena of interest. For example, it is a general occurrence that in situ sensors may not successfully detect features of interest (i.e. acoustic sensors may not detect sound sources or chemical sensors may not detect compounds for which they are sensitive.) However, of course, it cannot be concluded by the system that the lack of sensor signals means that no sources are present – they may simply be occluded by current environment conditions or be unexpected with respect to initial sensor deployment. Thus, proactive exploration of the sensing space is essential for establishing performance. Here, proactive coordinated mobility provides a constant background probing of system performance and at the same time surveying for unanticipated events and sources. Without this, the distributed sensor system may detect, at best, only those sources that were expected prior to deployment.

The underlying problem in coordinated mobility for any distributed actuated system is the selection of actions at the individual node level such that the entire system performance is optimized, or at least improved. Typically performance is measured using a task-specific objective function (give example from one of our applications here). This underlying action selection problem is widely studied in the mobile robotics community. Approaches fall broadly under one of two sets of algorithms: those that minimize spatial interference (e.g. avoid collisions between nodes at junctions), and those that focus on task allocation which dynamically assign nodes to tasks. Each of the two problems (interference and task allocation) exist for both regimes (proactive and reactive) in which NIMS operates.

ACCOMPLISHMENTS

Field System
Figure 4 shows an image of the first generation NIMS prototype system developed for forest environment monitoring. Its objective is the monitoring of critical parameters, including complex microclimate dynamics and also the spatiotemporally dynamic light environment that affect plant physiology and in particular, photosynthetic production by plants. The NIMS node also includes capability for imaging of the forest ecosystem. The NIMS node and its cable are shown suspended between trees. Both horizontal transport and vertical node transport have been implemented.

Figure 4. (Left Panel) A NIMS Node system deployed in a forest environment. This node includes embedded computing, wireless networking, horizontal transport, image sensing. This node also supports a vertically suspended meteorological sensing Node carrying atmospheric temperature, relative humidity, and photosynthetically active radiation (PAR) sensor devices. Wireless links provide access between the nodes and conventional wide area networks.

This prototype system includes an embedded processing platform (Linux operating system) and horizontal motion drive in a horizontally mobile node. This node also includes a two-axis articulated image sensor. The NIMS node also carries a vertical transport mechanism for a vertically-suspended NIMS node. This second node includes atmospheric temperature and relative humidity meteorological sensors along with an optical sensor for detection of downwelling photosynthetically active radiation (PAR). Wireless networking supports links between all NIMS nodes, fixed nodes, and gateway access points to the Internet that are distributed in the environment. While developed for forest monitoring, it is clear that this NIMS system is applicable in many other environments and is also one application-specific example of a very large configuration space of NIMS architecture choices.

NIMS has recently been deployed in both test environments for fundamental algorithm and system research as well as in a natural environment, the Wind River Canopy Crane Research Facility in the Wind River Experimental Forest in Washington. A view of the NIMS node suspended in the forest environment is shown in Figure 4.

Figure 5. The second generation prototype NIMS node.

Figure 5 shows two images of the second generation NIMS nodes. This system is similar to the first generation prototype node with slight upgrade modifications. This node is capable of withstanding different weather conditions. The above node is currently deployed in the James San Jacinto Mountain Reserve near Idyllwild, California. It has been in continuous operation since the end of March, 2004, enduring rain, wind, snow, and direct sunlight. This node is capable of operating over extended time periods. The plan is to operate this node continuously for a time period of at least several months. Unlike the previous prototype, no batteries are carried onboard either node. Power is distributed to the horizontal NIMS node through festooning cable similar to the first prototype node, and energy is delivered through the vertical support cable to the vertical hanging node.

FUTURE DIRECTIONS

Future work includes implementing additional types of NIMS nodes to coherently work as a team, including nodes that are fixed to the infrastructure and immobile nodes placed on the ground. Future NIMS nodes will be capable of obtaining and transporting physical samples from the environment and also reconfigure the network autonomously.

A NIMS system is planned to operate submerged underwater. This underwater NIMS system could be similar to a conventional NIMS system, utilizing embedded computer platforms, actuators, energy harvesters, and infrastructure enabled mobility.

PEOPLE

Faculty:

Prof. William J. Kaiser
Prof. Gregory J. Pottie
Prof. Mani Srivastava
Prof. Gaurav S. Sukhatme
Prof. John Villasenor
Prof. Deborah Estrin

Graduate Students:

Maxim Batalin
Jason Gordon
Lisa Gruzdas
David Jea
Aman Kansal
Duo Liu
Chris Lucas
Richard Pon
Mohammad Rahimi
Nithya Ramanathan
Arun Somasundara
Jeff Tseng
Yan Yu

Undergraduate Students:

Roja Bandari
Jamie Burke
Victor Chen
Willie Chen
Vishwa Goudar
Wendy Guo
Eric Lin
Kris Porter
Rachel Scollans
Lisa Schirachi
Mike Smith
Michael Stealey
Lynn Wang
Eric Yuen

Past Contributors:

Iman Ahmadi
Anita Chan
Ryan Fong
Gary Kao
Alex Liu
Richard Park
Bryan Ribaya
Jade Sche
Hassan Sharghi
Rocky Waugh
Shivesh Wangrungvichaisri