Describe the data analysis method(s) used.Evaluate the appropriateness of the data analysis method (hint: focus on the extent to which it addressed the research questions and the limitations of the method).

In recent years, distributed intelligent microelectromechanical systems (DiMEMSs) have appeared as a new form of distributed embedded systems. DiMEMSs contain thousands or millions of removable autonomous devices, which will collaborate with each other to achieve the final target of the whole system. Programming such systems is becoming an extremely difficult problem.

The difficulty is due not only to their inherent nature of distributed collaboration, mobility, large scale, and limited resources of their devices (e.g., in terms of energy, memory, communication, and computation) but also to the requirements of real-time control and tolerance for uncertainties such as inaccurate actuation and unreliable communications. As a result, existing programming languages for traditional distributed and embedded systems are not suitable for DiMEMSs.

In this article, we first introduce the origin and characteristics of DiMEMSs and then survey typical implemen- tations of DiMEMSs and related research hotspots. Finally, we propose a real-time programming framework that can be used to design new real-time programming languages for DiMEMSs. The framework is composed of three layers: a real-time programming model layer, a compilation layer, and a runtime system layer.

The design challenges and requirements of these layers are investigated. The framework is then discussed in further detail and suggestions for future research are given.

Categories and Subject Descriptors: C.2.4 [Computer-Communication Networks]: Distributed Systems General Terms: Design, Algorithms, Languages
Additional Key Words and Phrases: Distributed intelligent MEMS, autonomous devices, distributed collab- oration, real-time programming language, programming framework

Typical examples of MEMSs include accelerometers (which are commonly used in
airbag deployment systems for modern automobiles or in consumer electronics devices such as mobile phones to detect abnormal acceleration, e.g., collision or free-fall), mi-cromirror devices (which are used in video projectors or optics devices for light deflection and control), and so on.

However, with the rapid development of embedded and distributed intelligent tech-
nologies in recent years, large numbers of devices are being designed with the ability to collaborate with each other and provide efficient processing capacity for complex tasks as a group.

In order to prevent the problem of a single point of failure in a centralized controlling mode from occurring, each device in the group should work in a totally distributed fashion; that is, there will be no centralized control unit in the system, and each device will independently perform its tasks and communicate with other devices to achieve the final target of the system.

Such distributed collaboration can improve the scalability, fault tolerance, and energy efficiency of the whole system. This new kind of MEMS is called a distributed intelligent MEMS (DiMEMS) [Bourgeois et al. 2013].

In DiMEMSs, thousands or millions of devices are organized as a whole to perform
their tasks. For example [Lakhlef et al. 2014], a large number of MEMS mini-robots
(which can be viewed as devices) are attached together to form a larger robotic struc-
ture, and the larger robot system can perform more complex tasks (e.g., search and
rescue, space exploration, etc.) than can be performed using individual robots. A larger robot system has greater strength to perform actuation, more diversity to reconfigure its shape, and a larger capacity to perform tasks of sensing and communication.

A survey on progress on research relating to DiMEMSs was recently published
[Bourgeois and Goldstein 2015]. In the survey, the research challenges of the soft-
ware and hardware used in DiMEMS were analyzed, and several typical projects were
introduced. Our article differs from that survey in the following aspects:

(1) We are more concerned about typical applications of DiMEMSs and related re-
search hotspots.

(2) We are more concerned about the real-time operations and challenges of
DiMEMSs.

(3) Our emphasis is on the problem of how to design new and practical real-time
programming languages for DiMEMS. Therefore, we propose a real-time programmingframework. The design challenges and requirements of the framework are also inves-tigated. We also discuss future research on the framework and put forward feasible suggestions.

There is an urgent need for users to program their DiMEMS applications [Bordignon
et al. 2011]. However, the task of programming the system presents great difficulties
and complexities. The reasons for this include:

(1) Each device has limited sensing, communication, computation, and actuation
capabilities due to its small size. Therefore, it is hard for a device to independently
achieve the system’s target. Collaboration among the devices is inevitable.

(2) The number of devices is very large, making collaboration among them a very
complex matter.

(3) The distributed feature of the system makes it difficult for the devices to acquire
global information such as data on system topology, the identification numbers of other devices, and so on.

(4) The topology of the system may change dynamically due to the motions of the
devices or to the addition of devices and damage to some devices.

(5) Motion constraints and uncertainties. Motion constraint is a common feature in
DiMEMSs. For example, a device is required to connect with at least one of its neighbors when it is moving, and it cannot carry other devices. A motion may produce inexact or unexpected results (e.g., irregular alignment with other devices, unpredictable energy consumption, and unstable network connection), which are called uncertainties.
ACM Computing Surveys, Vol. 49, No. 1, Article 20, Publication date: