國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Chapter 1 Introduction
1.1 Background
The technology of Internet has been developed for a long time so far, and the most popular applications in daily life can mainly include two parts: First, the popularity of mobile devices. People can receive the Internet anytime, anywhere whether they use the evolution of Wi-Fi technology or 4G, 5G mobile networks;
Second, the emergence of social networking sites. People can use Facebook, Twitter and other web sites to establish a close relationship. They also help most kinds of information to be pushed more immediate and extensively.
In many previous studies, we tried to solve the influence maximization problem [1] in the online social network like in [2][3][4], or how to do trace data processing and data exploration effectively [5]. We can use the technology of Mobile Crowd Sensing(MCSC) [6]. However, most users are free to participate in MCSC environment [7]. So there is not enough incentive for users to upload sensor data using their own mobile network with no cost. In addition, the hardware resources and energy of mobile devices are also quite limited. So, data collection and sharing is a key issue in the situation of lack of infrastructure network architecture and limited hardware resources.
Mobile Crowd Sensing and Computing has a wide range of applications, such as temperature [8] and air quality detection in the city [9], restaurant recommendations [10] and so on. With the concept of MCSC, there are still many problems need to be solved. Mobile devices are different from ordinary large base stations, or high-spec computer equipment. If we can't handle too much information exchange while
‧
message transferring, can we derive the cycle and effect of message delivery and make the most effective solution?
1.2 Motivation
We assume a situation: There are all kinds of messages being spread on a campus. Basically, Messages are always broadcasted to the students by school, and the student are always receivers. In another common situation, the students in campus always transfer related messages by word of mouth. But each student's living habits may be different from those often visited. In Opportunistic Mobile Social Networks, the randomness is very large, and the node mobility is very high. Efficiency will be very low and the overhead will be too large if the message is random, with no rules.
Therefore, we hope to use the theme of our paper: “Mobile crowd sensing and computing”, including the key word "mobile crowd": different from the common centralized cloud control. It makes good use of the power of the masses. Especially the use of the mobile devices of the crowd, including mobile phones, tablets, wearable devices, etc. "sensing and computing": sensing includes "collecting" of nodes and
"encountering" with other nodes. And computing emphasizes that each node needs to do routing basic calculations to help optimize the transmission efficiency of the entire model.
In our situation, all the students on campus are nodes. They receive and transmit certain kinds of messages on mobile devices at any time. These messages all contain relevant interest content: some sports game information related to sports, some art exhibition relevant arts and culturally relevant information. Even some social related group activities we classify as social information. Students can bring this information to the building or department where they normally attend classes. Then the building
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
gather information and process other unnecessary messages. Finally, the buildings help broadcast the information that students might want, or the students who can help transferring through the building.
These processes are two-way. Each user can be a message carrier, a message producer, or one of the receivers interested in the message. Everyone in the mobile crowd sensing has an important transmission and routing role. This is the most important spirit of MCSC.
figure 1 MCSC in our paper
1.3 Concept of MCSC
We quote the paper "Mobile Crowd Sensing and Computing: The Review of an Emerging Human-Powered Sensing Paradigm"[21] in our research, and help us understand of the decentralized MCSC architecture. The article first mentions the development history of emerging human-powered sensing method, from crowd wisdom, crowdsourcing to Participatory sensing. The MCSC is analyzed in multiple aspects. The MCSC scale contains group, community, and urban components. The
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
difference between the masses is compared with the source of the collected data.
There are also many key determinants, including motivation of human participation, fusion of human-machine intelligence. User security and privacy, data quality,
heterogeneous, cross-space data mining affect the operation of MCSC that we should carefully consider.
This paper can be represented by the taxonomy graph below. At the same time, this picture also shows the core value of MCSC:
(1) The method mobile devices collect information, whether active or inactive, participatory or opportunistic.
(2) The process of collecting data and the composition of network nodes, including the operation of Infrastructure and Ad-Hoc networks.
(3) How to obtain information from the user's device, including user, ambient, social awareness.
Finally, there are several classic MCSC applications. There are cases using Environment monitoring, Urban dynamics sensing, Mobile social recommendation, Public safety, and so on.
‧ 國
立 政 治 大 學
‧
N a tio na
l C
figure 2 Concept of MCSCh engchi U ni ve rs it y
1.4 Relay Node Set
Because we don't know the exact destination of the message transmission in a DTN-based network environment, this is the obvious difference between DTN and the existing network topology. Therefore, we constantly update the information that the node encounters and the cost of the transfer to the next node as the message continues to store-and-forward[22]. With the calculated cost, it can help calculate the message transmission path in an environment where anycast forwarding does not specify a destination. This paper establishes a "relay node set" that computes the path that links nodes to other nodes based on anycast forwarding, and then links to destinations through these nodes. This paper uses the Bellman-Ford algorithm to calculate the cost
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
of the transmission path. And we uses the cost among the relay nodes to select the highest weight with highest priority, that is, the lowest cost path. This algorithm helps us find the best message forward method in polynomial time.
figure 3 Concept of Relay node Set