Source-linked AI summary
From Participatory Sensing to Mobile Crowd Sensing
Bin Guo, Zhiwen Yu, Daqing Zhang, Xingshe Zhou
TL;DR
The paper addresses how mobile-device sensing and mobile Internet data can support a broader crowd-sensing paradigm than participatory sensing alone. It reviews MCS’s history and issues, proposes a reference framework, and clarifies human-machine intelligence fusion. MCS is characterized by explicit and implicit participation and by combining mobile sensing with mobile social-network data.
Problem
Existing participatory sensing does not encompass both mobile social-network data and implicit as well as explicit participation in mobile-device crowd sensing.
Method
The paper reviews MCS’s literary history, defines its evolution from participatory sensing, proposes a reference framework, and discusses human-machine intelligence design and future research.
Results
MCS is characterized by crowd-powered data collection, cross-space data mining, low-quality data analysis, and data from both mobile sensing and mobile social networks.
Takeaways & Limitations
MCS provides a framework for combining human participation, mobile sensing, and mobile social-network data in people-centric sensing systems.
Abstract
from arXiv · showhide
The research on the efforts of combining human and machine intelligence has a long history. With the development of mobile sensing and mobile Internet techniques, a new sensing paradigm called Mobile Crowd Sensing (MCS), which leverages the power of citizens for large-scale sensing has become popular in recent years. As an evolution of participatory sensing, MCS has two unique features: (1) it involves both implicit and explicit participation; (2) MCS collects data from two user-participant data sources: mobile social networks and mobile sensing. This paper presents the literary history of MCS and its unique issues. A reference framework for MCS systems is also proposed. We further clarify the potential fusion of human and machine intelligence in MCS. Finally, we discuss the future research trends as well as our efforts to MCS.
I. INTRODUCTION
Mobile Crowd Sensing (MCS) extends mobile sensing by using citizens’ mobile devices and user-contributed online data for large-scale, people-centric sensing. The paper introduces MCS, illustrates its applications, identifies its research challenges, and proposes a reference framework.
- MCS uses mobile phones, wearable devices, and smart vehicles to acquire local knowledge such as location, context, noise, and traffic conditions.
- MCS aggregates and fuses mobile-device data and reused mobile social-network data in the cloud for crowd intelligence and people-centric services.
- Urban itinerary planning and noise mapping require combining mobile sensing with complementary information from location-based social-network data.
- The paper identifies challenges in data collection, incentives, user-data quality, cross-space fusion, and hybrid human-machine intelligence.
- The paper contributes a literature history and definition of MCS, clarifies human-machine intelligence fusion, proposes a reference framework, and discusses future trends.
II. FROM PARTICIPATORY SENSING TO MOBILE CROWD SENSING
MCS emerged by extending participatory sensing as mobile sensing and mobile Internet broadened crowd problem-solving with mobile devices. Its defining extension combines physical and online data sources and recognizes both explicit and implicit participation.
- Crowdsourcing solicits services or content from large online groups, whereas participatory sensing and MCS combine human and machine intelligence.
- Participatory sensing tasks everyday mobile devices to form interactive networks for gathering, analyzing, and sharing local knowledge through explicit participation.
- MCS extends participatory sensing by combining sensed mobile-device data with user-contributed mobile social-network data.
- MCS recognizes explicit primary-purpose participation and implicit second-purpose use of data originally contributed for online social interaction.
- Compared with participatory sensing, MCS enables cross-space fusion of online and offline data.
III. KEY FEATURES OF MCS
Mobile Crowd Sensing opens a crowd-powered sensing paradigm in which citizens contribute to sensing, and the section characterizes its key features.
- MCS is presented as a new crowd-powered sensing paradigm whose key features are characterized in this section.
A. Citizen Participation: Explicit or Implicit
MCS draws data from both mobile sensing and mobile social networks, while distinguishing whether participants are aware that their data serve a sensing task. Human participation also creates a human-machine intelligence design issue.
- A. Citizen Participation: Explicit or Implicit: MCS has two data-generation modes: mobile sensing from human-companied devices and user-generated data from mobile social-network services.
- A. Citizen Participation: Explicit or Implicit: Mobile social-network data provide large-scale information about city and societal dynamics and constitute MCS’s second data source.
- A. Citizen Participation: Explicit or Implicit: MCS classifies sensing style by participant awareness: sensing tasks are explicit when data collection is primary-purpose and implicit when data are reused secondarily.
- A. Citizen Participation: Explicit or Implicit: Human participation mixes human and machine intelligence, making their combination an important MCS system-design issue.
B. User Motivation
MCS participation depends on economic and non-economic motivations, while resource limits and sensitive data make incentives and privacy especially important.
- Participants may leave MCS data collection unless their return on investment exceeds expectations.Financial gain, interest, entertainment, socializing, glory, and recognition are identified as possible motivators.
- Financial gain, interest, entertainment, and social or ethical reasons can motivate participation.
- Limited device resources and sensitive information increase the importance of a solid economic model.The passage specifically mentions constrained energy and storage capacity and highly sensitive revealed information.
- Sharing personal data such as location and points of interest raises privacy concerns because the information is sensitive and vulnerable to attacks.MCS applications therefore need privacy-protection techniques that still allow reliable data contribution.
C. Dealing with Low-Quality Data
Human participation introduces uncertainty into MCS data, including incorrect, redundant, inconsistent, and condition-dependent observations. MCS therefore needs data-quality handling while extracting several levels of crowd intelligence.
- Anonymous participants may submit incorrect, low-quality, or fake data, while contributions can also be redundant or inconsistent.
- The same sensor can report different observations of an event under different conditions, such as when a phone is in a pocket or in hand.
- MCS extracts user, ambient, and social awareness from large volumes of user inputs.These dimensions cover personal contexts, space status, and group or community-level phenomena.
D. Heterogeneous, Cross-Space Data Mining
MCS combines data from offline and online communities whose differing coverage, infrastructure, operating times, interaction patterns, and social knowledge create heterogeneous cross-space mining challenges.
- Offline and online communities differ in geographical coverage, infrastructure support, and function time.
- Their distinct interaction patterns and implicit social knowledge provide heterogeneous inputs for cross-space data mining.Examples include online comments and likes, offline co-location, online friendship or trust, and offline movement patterns.
IV. A REFERENCE FRAMEWORK
The paper proposes a reference architecture for MCS as a starting point for the field. Its five layers organize sensing, transmission, collection, processing, and applications.
- IV. A REFERENCE FRAMEWORK: The proposed MCS architecture is intended as a starting point for advancing the research area.
- IV. A REFERENCE FRAMEWORK: The architecture contains five layers: crowd sensing, data transmission, data collection, crowd data processing, and applications.
- IV. A REFERENCE FRAMEWORK: Data transmission supports ad hoc, opportunistic, and infrastructure-based networks while tolerating network interruptions.
- IV. A REFERENCE FRAMEWORK: The data collection layer gathers data from selected sensor nodes and provides privacy-preserving mechanisms for contributors.
- IV. A REFERENCE FRAMEWORK: Crowd data processing applies machine learning and logic-based inference to transform low-level single-modality data into expected intelligence.It mines frequent patterns to derive integrated user, ambient, and social awareness.
- IV. A REFERENCE FRAMEWORK: The applications layer provides MCS-enabled services, including data visualization and user interfaces.
V. THE HYBRID HUMAN-MACHINE INTELLIGENCE DESIGN IN MCS
The paper frames MCS as a setting where human and machine intelligence contribute complementary abilities across system layers. It recommends application-centric trade-offs and formal design patterns for combining them.
- Human intelligence provides context, cognition, perception, and social interaction, but is limited by memory, speed, and variable data quality.
- Data collection layer: Machine intelligence can decompose complex sensing tasks, allocate them to suitable human nodes, and provide platforms for information sharing.
- DietSense combines automatic image processing with manual image review for recognition tasks that are complex or ambiguous.
- The paper recommends dynamically trading off human and machine intelligence according to application needs.
- It also calls for formal models and design patterns informed by social science, management, and computer science.
VI. FUTURE RESEARCH TRENDS AND OUR EFFORTS
The paper identifies unresolved MCS challenges in sensing-task assignment, networking, human grouping, and cross-community data integration. It presents these issues as future research opportunities for the still-emerging field.
- The study of MCS remains at an early stage, leaving numerous challenges and research opportunities for coming years.
- A Generic Framework for Data Collection: MCS task collection must select suitable mobile nodes using criteria such as region, time window, and acceptance conditions.
- Hybrid Mobile Networking: Infrastructure-based and opportunistic networks have distinct environments and benefits, but their interconnection remains underexplored in MCS.
- Varied Human Grouping: Future MCS systems should support interaction and varied human grouping, including virtual teams formed using skills and social networks.
- Cross-Community Sensing and Mining: Cross-community sensing integrates complementary data from online mobile social networks and offline mobile sensing communities.
- Cross-Community Sensing and Mining: Web knowledge can assist physical-world activity recognition, while heterogeneous community data can support new social applications.
VII. CONCLUSION
The conclusion characterizes MCS as an extension of participatory sensing that combines broader participation and data sources while introducing new design challenges.
- MCS extends participatory sensing through both implicit and explicit participation and data from mobile sensing and mobile social network services.
- The paper characterizes MCS through crowd-powered data collection, cross-space data mining, and low quality data analysis.
- It proposes a reference framework and discusses balancing human and machine intelligence in MCS system design.
- The conclusion identifies hybrid networking, varied user grouping, and cross-community sensing and mining as ongoing MCS issues.