Citation Context

...rinciples of "reactive robotics", distributed sensing systems tend to operate in a "reactive" mode in which their immediate behavioris heavily influenced by asynchronous and concurrent sensor inputs [=-=Brooks 1986-=-]. ACM Journal Name, Vol. V, No. N, June 2006.sEmstar: a Software Environment for Developing Heterogeneous Sensor Networks * 7 Layer 0: FUSDLow Level IPC Layer 1: GlibHandle events on IPC Layer 2: Dev...

Citation Context

...s in the development of wireless sensor network systems. EmStar “supports” several routing protocols, but it also makes it easy to invent your own. For example, the authors of Directed Diffusion [16] =-=[18]-=- have ported diffusion to run on top of EmStar. 2.3 EmStar Device Support EmStar includes native support for a number of devices, including sensors and radio hardware. HostMote and MoteNIC EmStar syst...

Citation Context

...s interface begin with a standard link header that specifies common fields and masks cosmetic differences in theactual over-the-air headers used by different MAC layers, such as 802.11, Berkeley MAC [=-=Hill et al. 2000-=-] and SMAC [Ye et al. 2002]. The command and status devicesprovide asynchronous access to the configuration of a stack module. 3.3.8 Impact of FUSD and Emstar Devices on Latency. While Emstar processi...

Citation Context

...d link header that specifies common fields. This link header masks certain cosmetic differences in the actual over-the-air headers used by different MAC layers, such as the Berkeley MAC [17] and SMAC =-=[22]-=- layers supported on Mica Motes. The command and status devices provide asynchronous access to the configuration of a stack module. The status device reports the current configuration of the module (s...

Citation Context

...networking systems field who are bringing new application needs to the table. A recent NSF Workshop report [4] details a number of these needs, building on early experience with deployments (e.g. GDI =-=[7]-=-, CENS [23], James Reserve [26]). Many of these applications lead to “tiered architecture” designs, in which the system is composed of a mixture of platforms with different costs, capabilities and ene...

Citation Context

...ks is the capacity to perform system-wide compile time optimizations. Because EmStar supports forms of dynamic binding that do not exist in TinyOS, many compiletime optimizations are ruled out. Click =-=[19]-=- is a modular software system designed to support implementation of routers. While Click is designed for a different application space, there are many similarities, including an emphasis on modularity...

Citation Context

...lation, emulation, and visualization of live systems, both real and simulated. EmSim/EmCee Transparent simulation at varying levels of accuracy is crucial for building and deploying large systems [9] =-=[11]-=-. Together, EmSim and EmCee comprise several accuracy regimes. EmSim runs many virtual nodes in parallel, in a pure simulation environment that models radio and sensor channels. EmCee runs the EmSim c...

Citation Context

...nterpreted on another node. Timesync can also compute relations between the local CPU clock and other clocks in the system, such as sample indices from an ADC or the clocks of other processor modules =-=[3]-=-. Routing EmStar supports several types of routing: Flooding, Geographical, Quad-Tree, and Diffusion. One of the founding principles of EmStar is that innovation in routing and hybrid transport/routin...

Citation Context

... James Reserve [26]). Many of these applications lead to “tiered architecture” designs, in which the system is composed of a mixture of platforms with different costs, capabilities and energy budgets =-=[5]-=- [21]. Low capability nodes, often Crossbow Mica Motes [24] running TinyOS [17], can perform simple tasks and provide long life at low cost. The high capability nodes, or Microservers, generally consu...

Citation Context

...e accessible from the shell and from most simple UNIX utilities and scripting languages without additional support. Ousterhout’s arguments on the relationships between scripting and system languages [=-=Ousterhout 1998-=-] apply here. Accessing Emstar services from scripts trades-off performance for significant ease of use and code re-use. By providing typeless ASCII interfaces in addition to typed binary interfaces, ...

Citation Context

...at plays a role similar to FUSD in Emstar. 5.2 Simulation Environments Several other simulation environments have been applied to sensor network systems, including TOSSIM [Levis et al. 2003], Avrora [=-=Titzer et al. 2005-=-] and ns-2 2 . TOSSIM and Avrora both support real-code simulation of Mote-based systems. The ns-2 simulation environment is commonly used to model the behavior of Internet protocols. Like EmTOS, TOSS...

Citation Context

...t contain both Motes and Emstar sys-tems, by running the Mote code inside an EmTOS wrapper [Girod et al. 2004b]. Other "real-code" simulation environments include TOSSim [Levis et al. 2003] andSimOS [=-=Rosenblum et al. 1997-=-], but Emstar is the only environment that readily supports heterogeneous networks and "emulation" using real hardware. 3.5.2 Remote Access Methods. As an IPC mechanism, FUSD has the benefit ofbeing f...

Citation Context

... but is extensible to support hardware-specific features in the interim before the uniform abstraction can be extended. This represents a different approach from the recently proposed SP abstraction [=-=Polastre et al. 2005-=-], in which all hardware-specific details are hidden behind a rigid abstract interface. Emstar implements several “link drivers”, providing interfaces to radio link hardware including 802.11, and seve...

Citation Context

... enable cross-process messagepassing, while at the same time exposing shell-accessible interfaces to internal state and control functions. In some respects, FUSD is similar to the AT&T Plan 9 system [=-=Pike et al. 1990-=-], but has the advantage of running over any platform running Linux rather than requiring the port of a complete OS to the latest embedded hardware. FUSD also has much in common with the procfs and sy...

Citation Context

...system restart model. Prior work has addressed robustness by partial rebooting in several contexts, including Microreboot [Candea and Fox 2003] [Candea et al. 2004] for distributed systems and Nooks [=-=Swift et al. 2004-=-] for driver restart within a Linux kernel. The conclusion of these efforts has agreed with our own conclusions: enabling restart is often the best approach to designing robust complex applications, b...

Citation Context

... areas in the development of wireless sensor network systems. EmStar “supports” several routing protocols, but it also makes it easy to invent your own. For example, the authors of Directed Diffusion =-=[16]-=- [18] have ported diffusion to run on top of EmStar. 2.3 EmStar Device Support EmStar includes native support for a number of devices, including sensors and radio hardware. HostMote and MoteNIC EmStar...

Citation Context

...solving relativelysimple problems in a one-off demo context, the challenges we describe here made it difficult to build more complex and robust solutions on these early successes [Girodet al. 2004b] [=-=Girod et al. 2002-=-]. This led to the design of Emstar. Emstar is a software framework and development/deployment environment designed for imple-menting complex sensing applications within a distributed network of Micro...

Citation Context

...timation service. 4. CASE STUDY: THE ACOUSTIC ENSBOX ACOUSTIC POSITIONING SYSTEM To show how we use Emstar to build a sophisticated system of Microservers, in thissection we describe Acoustic ENSBox [=-=Girod et al. 2006-=-] [Girod 2005], a system we recently built using Emstar. The Acoustic ENSBox system, shown deployed inthe James Reserve in Figure 9(b), is a platform for developing distributed acoustic sensing applic...

Citation Context

....In addition, as with many microkernels, the cost of crossing the kernel boundary on inter-service calls is expensive. However, newer microkernels such as L4 achievereasonably comparable performance [=-=Hrtig et al. 1997-=-] to monolithic kernels using memory mapping mechanisms to share memory for large data transfers. Such anapproach could also be used in Emstar for large memory transfers where buffer copies across ker...

Citation Context

...es Reserve [26]). Many of these applications lead to “tiered architecture” designs, in which the system is composed of a mixture of platforms with different costs, capabilities and energy budgets [5] =-=[21]-=-. Low capability nodes, often Crossbow Mica Motes [24] running TinyOS [17], can perform simple tasks and provide long life at low cost. The high capability nodes, or Microservers, generally consume mo...

Citation Context

... much more difficult to implement correctly than a full-system restart model. Priorwork has addressed robustness by partial rebooting in several contexts, including Microreboot [Candea and Fox 2003] [=-=Candea et al. 2004-=-] for distributed systems andNooks [Swift et al. 2004] for driver restart within a Linux kernel. The conclusion of these efforts has agreed with our own conclusions: enabling restart is often the best...

Citation Context

...ssing of high-rate audio: audible acoustic ranging, acoustic beamforming, and animal call detection are a few of the applications. We used earlier versions of EmStar to tackle a few of these problems =-=[10]-=-[15][8]. Referring back to the animal call localization application of Figure 1, we see how the “syncd” and “audiod” services collaborate so that “collab detect” can correlate events detected on nodes...

Citation Context

...ifferent nodes and uses linear regression to estimate conversion parameters among clocks that receive broadcasts in common. We chose RBS because techniques based on measuring send times, such as TPSN =-=[14]-=-, are not generally applicable without support at the MAC layer. Requiring this support would rule out many possible radios, including 802.11 cards. A key insight in RBS is that it is better to enable...

Citation Context

...ypothetical application for which EmStar is well-suited. In this example, several nodes collaborate to acoustically localize an animal based on its call—an improved version of our system described in =-=[8]-=-. The large dashed box shows how the system might be implemented by combining existing EmStar components (gray boxes) with hypothetical application-specific components (light gray dashed boxes). Becau...

Citation Context

...se EmStar at CENS in several current development efforts, including a 50-node seismic deployment and the ESS microclimate sensing system. We also support other groups using EmStar, including the NIMS =-=[13]-=- robotic ecology project and ISI ILENSE. Our current platform focus is the Crossbow/Intel Stargate platform, an inexpensive Linux platform based on the XScale processor. Stargates are much easier to c...

Citation Context

...de a simpler “straight-line” programming model when making immediate calls to a service. On the other hand, a sockets-based solution would not be subject to deadlock. Some libraries such as libasync [=-=Dabek et al. 2002-=-] enable asynchronous socket operations within an event-based system to appear as though they were “straightline” blocking code. However, the very fact that these systems allow asynchronous operations...

Citation Context

...simulation, emulation, and visualization of live systems, both real and simulated. EmSim/EmCee Transparent simulation at varying levels of accuracy is crucial for building and deploying large systems =-=[9]-=- [11]. Together, EmSim and EmCee comprise several accuracy regimes. EmSim runs many virtual nodes in parallel, in a pure simulation environment that models radio and sensor channels. EmCee runs the Em...

Citation Context

... capacity to perform system-wide compile time optimizations. Because Emstar supports forms of dynamic binding that do not exist in TinyOS, many compile-time optimizations are ruled out. Player/Stage [=-=Vaughan et al. 2003-=-] is a software system designed for robotics that supports “real-code” simulation. In many respects, Player and Emstar have a similar philosophy. Both focus considerable effort on ease of system integ...

Citation Context

...of both Motes and Microservers [Girod et al. 2004b]. Using EmTOS as a building block, we have developed more sophisticated Mote-Microserver collaboration tools, including the concept of Mote Herding [=-=Stathopoulos et al. 2005-=-]. 3.3.3 EmRun Services. EmRun is a program that parses a configuration file that describes an Emstar system, launches the system, and provides several centralized services. Among the services it prov...

Citation Context

... Emstar at CENS in several current development efforts, including the Acoustic ENSBox project [Girod et al. 2006] [Girod 2005], the Cyclops heterogeneous camera network, a 50-node seismic deployment [=-=Lukac et al. 2006-=-], and the James Reserve ESS microclimate sensing system. In support of these projects, we continue to enhance Emstar with new components and tools. We also support several other groups using Emstar, ...

Citation Context

...tonomously estimate a high-accuracy 3D position map for all nodes in the system. This map may then be used by higher-level applications such as bearing-based localization applications and algorithms [=-=Wang et al. 2005-=-; Wang et al. 2004; Chen et al. 2003; Ali et al. 2007]. 4.1.1 Leveraging Emstar’s Wealth of Services. The implementation of the Acoustic ENSBox software stack, shown in Figure 9, uses most of the comp...

Citation Context

...concurrently support multiple independent applications. The ability to support multiple applications was a requirement of some of the first instances of sensor nodes, notably the WINSNG 2.0 platform [=-=Merrill et al. 2002-=-] used in the DARPA SensIT program. In this program, several independent research groups ran applications on a single fielded network. Given that applications act independently, there was therefore a ...

Citation Context

...latively unre-liable nodes, the most expedient solution would simplify the end nodes by centralizing as much of the computation as possible. Unfortunately, interest in high ratesensors such as audio [=-=Wang et al. 2004-=-] [Chen et al. 2003] or video and the desire for low-latency actuation within the network [Merrill et al. 2004] often render thisstrategy impossible. In addition to increasing complexity, these sophis...

Citation Context

...ough the sophisticated application software shown in the architectural diagram in Figure 9. We have used this system to detect and localize animal calls in several field experiments [Ali et al. 2007; =-=Trifa et al. 2007-=-]. In a typical deployment, a cluster of ENSBoxes is positioned on tripods in an area of interest. Each node runs an Emstar stack and hosts a 4-channel microphone array and 4 piezo tweeters, capable o...

Citation Context

...USD provides an API through which those system calls can be handled [Elson 2002]. This API is modeled along the lines of the character device API in the Linux kernel, with a few key differences. See [=-=Rubini and Corbet 2001-=-] for an introduction to the Linux character device API. Similar to the Linux character device API, FUSD handles the character device system calls by calling handler functions specified by the server ...

Citation Context

...e support for networking, sensing, and time synchronization. Link and Neighborhood Estimation Wireless channels have a significant “gray zone” where connectivity is unreliable and highly time-varying =-=[6]-=-. Node failures are also common. Therefore, applications are brittle when they assume the topology is pre-configured. Dynamic neighbor discovery is a basic service needed by all collaborative applicat...

Citation Context

... is impossible when the application involves sensors with high data rates such as audio [Wang et al. 2004] [Chen et al. 2003] or video, or requires low-latency actuation within an unreliable network [=-=Merrill et al. 2004-=-]—these cases force a distributed implementation. These types of distributed system are notoriously difficult to deploy. Through our initial experiences in this area we identified several impediments ...

Citation Context

...from hardware through the sophisticated application software shown in the architectural diagram in Figure 9. We have used this system to detect and localize animal calls in several field experiments [=-=Ali et al. 2007-=-; Trifa et al. 2007]. In a typical deployment, a cluster of ENSBoxes is positioned on tripods in an area of interest. Each node runs an Emstar stack and hosts a 4-channel microphone array and 4 piezo ...

Citation Context

...t is also important for the user to be aware of faults that may have occurred in the deployment. To address these needs, Emstar includes StateSync, an efficient reliable statedissemination mechanism [=-=Girod et al. 2005-=-]. The StateSync protocol defines a local retransmission protocol, and is provided in two versions: the flooding versionand the tree-based multicast version. The flooding version is ideal for debuggin...

Citation Context

... has shown it is well worth the infrastructure investment. Users have consistently observed that using real radios is substantially different from our best efforts at creating a modeled radio channel =-=[2]-=-[6]. Even channels driven by empirical data captured using the ceiling array don’t seem to adequately capture the real dynamics. Although testing with EmCee is still not the same as a real deployment,...

Citation Context

...of the most promising signs in the field is a growing involvement by researchers outside the networking systems field who are bringing new application needs to the table. A recent NSF Workshop report =-=[4]-=- details a number of these needs, building on early experience with deployments (e.g. GDI [7], CENS [23], James Reserve [26]). Many of these applications lead to “tiered architecture” designs, in whic...

Citation Context

... systems field who are bringing new application needs to the table. A recent NSF Workshop report [4] details a number of these needs, building on early experience with deployments (e.g. GDI [7], CENS =-=[23]-=-, James Reserve [26]). Many of these applications lead to “tiered architecture” designs, in which the system is composed of a mixture of platforms with different costs, capabilities and energy budgets...