Linux-based COMs drive remote data acquisition strategies

3Experimental data acquisition, particularly remote data acquisition, requires custom-designed equipment that supports redesign and re-deployment. The availability of sophisticated open source software packages for small form factor systems tempts designers to push increasing system complexity onto remote monitoring stations. These remote stations appear more and more like servers with high-accuracy, high-fidelity, and high sampling frequency requirements that push against the constraints of power consumption, battery life, and physical size.

Such diverse fields of endeavor as oceanography, aerial data-collection, agriculture, and low-earth orbit are united in tackling the issues of system design for challenging environments. Many projects use an intelligent remote platform as a core element in a system architecture to solve the problem of design iteration. In the following application examples, open source software designed for desktops but deployed on miniature Computer-On-Module (COM) systems allowed each of these teams to iterate rapidly and deploy successful systems capable of generating data adequate for peer-reviewed journals in their disparate fields.

Acquiring underwater data

In Australia, an international team of ocean scientists building Smart Environmental Monitoring and Analysis Technologies (SEMAT)[1] (Figure 1) needed to build a marine multi-station monitoring system to gather environmental health data for use in developing sustainability strategies and policies. The team behind the project determined a plug-and-play Wireless Sensor Network (WSN) – a collection of intelligent wireless sensors that transmit gathered data over a network for analysis – to be the best setup for their system.

Figure 1: The Smart Environmental Monitoring and Analysis Technologies (SEMAT) system used a WSN and open source systems to collect data about a marine environment for analysis.
(Click graphic to zoom)

In a harsh marine environment, many existing systems are cost-prohibitive. As in all scientific projects, cost management was critical but the team also needed a powerful enough approach to allow flexibility. Creating SEMAT with cheaper, modular, open source technologies meant the team could also focus on creating a user-friendly system for analysis rather than being bogged down with the technical knowledge and skill of building a complex network from scratch.

Three iterations were developed to come up with the best long-term data collection system. The first iteration, Mk1, used a closed hardware solution, and was therefore inflexible to expansion beyond the initial trial’s usage. Mk2, the second prototype, was developed as an open standard system. The Mk2’s computation and communications subsystem consisted of a COM with an onboard Wi-Fi chipset, and an expansion board for various peripherals including USB. During testing it became clear to the team that the onboard Wi-Fi was not powerful enough to cover the 1.7 km distance covered by the system, so a more powerful substitute was plugged in via USB. Mk2 was successful, but suffered from some environmental and data transmission shortcomings.

The Mk3 system used the next generation of the same COM platform as the Mk2. The software infrastructure was also reused because of the plug-and-play nature of hardware elements, even though the software and high-level reporting functions grew in scope. As sensor or communications limitations were discovered, the high-level COM-based architecture allowed substitution of new devices with minimal to no alterations in various data acquisition, data management, database, or network communications software.

Data acquisition and UAVs

Unmanned Aerial Vehicles (UAVs) that work in a team require fast and precise data gathering and distribution among units. The Tactical Agentfly project examined in a Prague university Mobile Ad Hoc Network (MANET) Ph.D. thesis[2] involved multiple real and simulated UAVs, each with its own sensors and actuators, that perform tactical missions as a unit. The system included onboard and ground-level components. The UAVs in the system needed to be able to quickly distribute mission data from a human operator to fulfill tasks, and precisely measure coordination to avoid collisions in a tight flight space. The UAVs attach and release from network connections where detection of a nearby UAV unit should trigger the onboard sensors to increase their transmission power while in range for increased reliability. Ensuring reliable communication was key to the UAV team carrying out its tasks.

Choosing a remote data collection server that could collect the data, manage data streams, and manage the topology of MANET systems in a small, lightweight package is crucial to a successful system in this application. Because the ground control unit was not sufficient to carry out missions, COMs were installed in the UAVs to host control algorithms, and Xbee modules were added for additional bandwidth. The COM itself met the Size, Weight, and Power (SWaP) requirements necessary for UAV systems; it measured in at 17 mm x 58 mm x 4.2 mm and consumed a mere 5 W of power. The project used a Linux operating system for ease of operation.

Selecting a system with low power requirements but high-level software let the system designer perform network management on a remote server in a distributed network. Theoretically, very large remote areas can be covered with low-cost, low-power 802.3 sensor networks that don’t require fixed-location base stations for access by intermittently connecting to UAVs and up-regulated power during the flyovers.

Investing between $100 and $200 for a COM solution appears more expensive than using a dedicated, lower-performance chip that does not support Ubuntu or Android; however, that investment is dwarfed by the higher and unpredictable cost of software development for the embedded systems without high-level open source software.

Collecting data in the ground …

On remote semi-arid farms in India, data acquisition systems are being used to measure highly varying soil moisture, rain data, and other information vital to crop yields. But remote areas with variable access to the power grid and a high illiteracy rate pose many challenges. The following application involves multiple village farms with sensor networks that connect to a centralized hub outside of the village where the data can be analyzed and relayed to farmers to give personalized agricultural advice[3].

The WSN network deployed suffered from many outages due to unreliable grid power, and power generated from solar panels was off the table due to visible sensor nodes attracting attention leading to theft and vandalism. The solution was to use a human-powered bicycle to power the nodes and opportunistically transfer data over a “Wi-Fi-GPRS” bridging system when power was available through the bicycle generators.

Flexibility and low cost in the setting of a small or semiarid agricultural setting requires an inexpensive development and deployment cycle in addition to robustness and ease-of-use. Readily available commercial sensors can provide a wide variety of speeds, signal-to-noise ratios, and power consumption profiles that meet the baseline requirements for deployment in remote areas. A tiny server like the Gumstix Overo COM with a SUMIT expansion board requiring less than 2 W of power suits the requirement to be powered by a bicycle-driven generator, yet with Linux it provides ease of use with the availability of device drivers when sensors need upgrading or replacing as needs of the system evolve. The ability to take these fully functional computer modules and operate them with the ease of managing any other Linux server makes them reprogrammable in-system. This applies to communication capabilities as well, for GSM, 802.11(g), or Ethernet. Ability to use a well-known Linux distribution means the community is available with support and related developments for sharing.

… and in space

Perhaps the greatest remote deployment conceived is where data transmission costs are most severe: satellite image data[4]. In the following in-development project that deals with asteroid imaging via satellite, onboard data analysis and transmission systems analyze gathered image data for potential interest to researchers, then compress the images to minimize bandwidth requirement. To address the challenges of this application, system designers can push data management tasks, in this case Model Based Compression (MBC), onto powerful processors in a CubeSat miniature satellite form factor (typically 100 mm3) with COTS components to minimize transmission costs from low-Earth orbit. The project team is working on getting their Model-Based Data Transmission (MBDT) system on a COM, which serves SWaP constraints for a system deployed in a small satellite. At 17 mm x 58 mm x 4.2 mm and drawing approximately 200 mA at 3.3 V in a standard boot of Linux, the 1 GHz Overo WaterSTORM COM provides a DSP that could satisfy the computational requirements of satellite deployment (Figure 2).

Figure 2: The Gumstix 1 GHz Overo WaterSTORM COM meets SWaP requirements for small satellite systems with its 17 mm x 58 mm x 4.2 mm size and sub-1 W power draw.
(Click graphic to zoom by 1.9x)

Open source Linux and COMs meet remote data acquisition needs

Tiny but fully functional Linux COMs have been shown in each of these examples to provide sophisticated computing power to the task of remote data collection. By maximizing remote processing power, the teams each added flexibility to sensor selection, eased the upgrading of their system designs as their projects evolved, and minimized development, debugging, and deployment costs.


[1] SEMAT – The Next Generation of Inexpensive Marine Environmental Monitoring and Measurement Systems.

[2] Distributed topology control in MANETs.

[3] Data Gathering and Information Dissemination For Semi-Arid Regions.

[4] Reducing Link Budget Requirements with Model-Based Transmission Reduction Techniques.

W. Gordon Kruberg, M.D. is President and C.E.O. at Gumstix, Inc.

Gumstix, Inc.