[雙語翻譯]物聯(lián)網(wǎng)外文翻譯--物聯(lián)網(wǎng)的研究方向（英文）

1、Research Directions for the Internet of ThingsJohn A. Stankovic, Life Fellow, IEEE(Invited Paper)Abstract—Many technical communities are vigorously pursuing research topics that contribute to the Internet of Things (IoT)

2、. Nowadays, as sensing, actuation, communication, and control be- come even more sophisticated and ubiquitous, there is a significant overlap in these communities, sometimes from slightly different perspectives. More coo

3、peration between communities is encour- aged. To provide a basis for discussing open research problems in IoT, a vision for how IoT could change the world in the distant future is first presented. Then, eight key researc

4、h topics are enumerated and research problems within these topics are discussed.Index Terms—Cyber physical systems, Internet of Things (IoT), mobile computing, pervasive computing, wireless sensor networks.I. INTRODUCTIO

5、N THE notions Smart devices, Smartphones, Smart cars, Smart homes, Smart cities—A smart world—have been espoused for many years. Achieving these goals has been investigated, to date, by many diverse and often disjoint re

6、search communities. Five such prominent research communities are: Internet of Things (IoT), mobile computing (MC), pervasive computing (PC), wireless sensor networks (WSNs), and, most recently, cyber-physical systems (CP

7、S). However, as technology and solutions progress in each of these fields, there is an increasing overlap and merger of principles and research ques- tions. Narrow definitions of each of these fields are no longer approp

8、riate. Further, research in IoT, PC, MC, WSN, and CPS often relies on underlying technologies such as real-time com- puting, machine learning, security, privacy, signal processing, big data, and others. Consequently, the

9、 smart vision of the world involves much of computer science, computer engineering, and electrical engineering. Greater interactions among these com- munities will speed progress. In this paper, as a backdrop to identify

10、ing research questions, Section II briefly highlights a vision for a smart world. Section III then discusses open research questions categorized into eight topics. The research discussed is representative rather than com

11、plete. Two goals of the paper are: 1) to highlight a number of significant research needs for future IoT systems; 2) to raise awareness of work being performed across various research communities.II. VISION AND IOT SCOPE

12、Many people [8], including myself [28], [29], hold the view that cities and the world itself will be overlaid with sensing and actuation, many embedded in “things” creating what is referred to as a smart world. But it is

13、 important to note that one key issue is the degree of the density of sensing and actuation coverage. I believe that there will be a transition point when the degree of coverage triples or quadruples from what we have to

14、day. At that time, there will be a qualitative change. For example, nowadays, many buildings already have sensors for attempting to save energy [7], [38]; home automation is occurring [3]; cars, taxis, and traffic lights

15、 have devices to try and improve safety and transportation [9]; people have smartphones with sensors for running many useful apps [2]; industrial plants are connecting to the Internet [1]; and healthcare services are rel

16、ying on increased home sensing to support remote medicine and wellness [11]. However, all of these are just the tip of the iceberg. They are all still at early stages of development. The steady increasing density of sens

17、ing and the sophistication of the associated processing will make for a significant qualitative change in how we work and live. We will truly have systems-of-systems that synergisti- cally interact to form totally new an

18、d unpredictable services. What will be the platform or platforms that support such a vision? One possibility is a global sensing and actuation utility connected to the Internet. Electricity and water are the two utilitie

19、s that can be used for a myriad of purposes. Sensing and actuation in the form of an IoT platform will become a utility. IoT will not be seen as individual systems, but as a critical, integrated infrastructure upon which

20、 many applications and services can run. Some applications will be personalized such as digitizing daily life activities, others will be city-wide such as efficient, delay-free transportation, and others will be worldwid

21、e such as global delivery systems. In cities, perhaps there will be no traffic lights and even 3-D transportation vehicles. Smart buildings will not only control energy or security, but integrate personal comfort, energy

22、 savings, security, and health and wellness aspects into convenient and effective spaces. Individuals may have patches of bionic skin with sensing of physiological para- meters being transmitted to the cloud which houses

23、 his/her digital health, and to the surrounding smart spaces for improved com- fort, health, efficiency, and safety. In fact, smart watches, phones, body nodes, and clothes will act as personalized input to optimize city

24、-wide services benefiting both the individual and society. Consequently, we will often (perhaps 24/7) be implicitly linked into the new utility. Some examples of new services include immediate and continuous access to th

25、e right information for the task at hand, be it, traveling to work or a meeting, exercising, shopping, socializing, or visiting a doctor. Sometimes theseManuscript received January 03, 2014; accepted March 09, 2014. Date

26、 of publication March 18, 2014; date of current version May 05, 2014. This work was supported by the National Science Foundation under Grant CNS-1239483, Grant CNS-1017363, and Grant CNS-1319302. The author is with the C

27、omputer Science Department, University of Virginia, Charlottesville, VA 22904 USA (e-mail: stankovic@cs.virginia.edu). Digital Object Identifier 10.1109/JIOT.2014.2312291IEEE INTERNET OF THINGS JOURNAL, VOL. 1, NO. 1, FE

28、BRUARY 2014 32327-4662 © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more inform

29、ation.to turn ON all the lights. On the other hand, the energy manage- ment application may decide to turn OFF lights when no motion is detected. Detecting and resolving such dependency problems is important for correctn

30、ess of operation of interacting IoT systems.C. Creating Knowledge and Big DataIn an IoT world, there exists a vast amount of raw data being continuously collected. It will be necessary to develop techni- ques that conver

31、t this raw data into usable knowledge. For example, in the medical area, raw streams of sensor values must be converted into semantically meaningful activities performed by or about a person such as eating, poor respirat

32、ion, or exhibit- ing signs of depression. Main challenges for data interpretation and the formation of knowledge include addressing noisy, physical world data, and developing new inference techniques that do not suffer t

33、he limitations of Bayesian or Dempster–Shafer schemes. These limitations include the need to know the a priori probabilities and the cost of computations. Rule-based systems may be used, but may also be too ad hoc for so

34、me applications. The amount of collected data will be enormous. It can be expected that a very large number of real-time sensor data streams will exist, that it will be common for a given stream of data to be used in man

35、y different ways for many different inference purposes, that the data provenance and how it was processed must be known, and that privacy and security must be applied. Data mining techniques are expected to provide the c

36、reation of important knowledge from all this data. Enabling streams to act as primitives for unexpected future inferences is an interesting research problem. In addition, the overall system solution must deal with the fa

37、ct that no inference method is 100% correct. Consequently, uncertainty in interpreted data can easily cause users not to trust the system. Trust is one important aspect of the usefulness of big data. Security and privacy

38、 are essential elements of trust and these are discussed in Sections III-F and III-G. However, as a basis for trust, it is also necessary to develop new in-field sensor calibra- tion techniques and reliable transport pro

39、tocols. Without these basic underlying system-level capabilities, further inference might be operating with wrong or too much missing data, resulting in wrong conclusions. If these wrong conclusions drive actuators, then

40、 serious safety problems can occur. One approach is to ensure that all inferred information is accompanied by a confidence level in the form of a probability that the information is correct or incorrect and use that info

41、rmation to guarantee safe actuator operation. In many applications, informing users how information was derived is necessary. Another main challenge is making good (control) decisions using the created knowledge. However

42、, in making decisions, it is necessary to minimize the number of false negatives and false positives and guarantee safety; otherwise, the system will be dismissed as unreliable. Many IoT applications will be designed to

43、work for a particu- lar person. It is necessary to perform correct data association ensuring that the collected data and subsequent inferences are associated with the correct individual or individuals. This is a very cha

44、llenging problem for many situations. When users are wearing RFIDs or when cameras with pattern recognition are used, then the problem is solved (except for the privacy issues).However, in many other situations, it will

45、be necessary to combine a set of current sensor readings with a trace of the recent past readings and utilize a history of a given user’s activities and personal characteristics to arrive at an accurate data assignment.

46、More research is necessary on this problem.D. RobustnessIfourvisioniscorrect,manyIoTapplicationswillbebasedona deployed sensing, actuation, and communication platform (con- necting a network of things). In these deployme

47、nts, it is common forthedevicestoknowtheirlocations,havesynchronizedclocks, know their neighbor devices when cooperating, and have a coherent set of parameter settings such as consistent sleep/ wake-upschedules,appropria

48、tepowerlevelsforcommunication, andpair-wisesecuritykeys.However,overtime,theseconditions can deteriorate. The most common (and simple) example of this deterioration problem is with clock synchronization [18]. Over time,

49、clock drift causes nodes to have different enough times to result in application failures. While it is widely recognized that clock synchronization must reoccur, this principle is much more general. For example, some nod

50、es may be physically moved unexpectedly. More and more nodes may become out of place over time. To make system-wide node locations coherent again, node relocalization needs to occur (albeit at a much slower rate thanforc

51、locksync).Thisissuecanbeconsideredaformofentropy where a system will deteriorate (tend toward disorder) unless energy in the form of rerunning protocols and other self-healing mechanismsisapplied[35].Notethatcontrolofact

52、uatorscanalso deteriorateduetotheircontrollingsoftwareandprotocols,butalso due to physical wear and tear. In other words, how can a long- lived, dynamic, and mobile IoT be maintained? The required coherence (entropy) ser

53、vices must combine with many other approaches to produce robust system operation. This includes formal methods to develop reliable code, in situ debug- ging techniques, online fault tolerance, in-field maintenance, and g

54、eneral health monitoring services [23]–[25]. These problems are exacerbated due to the unattended operation of the system, the need for a long lifetime, the openness of the systems, and the realities of the physical worl

55、d. The goal is for this collection of solutions to create a robust system in spite of noisy, faulty, and nondeterministic underlying physical world realities. Another problem barely addressed to date is that in some IoT

56、applications, especially safety critical ones, run time assurances must be given to authorities, e.g., to (re)certify that the system is operating as expected. Consider a fire-fighting system deployed in a sky scraper of

57、fice building to detect fires, alert fire stations, and aid in evacuation. Periodically, it is necessary to demonstrate to certification authorities that this system meets these require- ments. Such IoT applications will

58、 need services that can support run-time certification.E. OpennessTraditionally, the majority of sensor-based systems have been closed systems. For example, cars, airplanes, and ships have had networked sensor systems th