CROWDSOURCE-BASED SIGNAL STRENGTH FIELD ESTIMATION BYGAUSSIAN PROCESSES
the authors apply a Gaussian Process (GP) to model the RSS(received signal strength), and for estimation they use measurements from known locations，which based on a log-normal path loss model and perfectly knowledge of the user locations.
In this paper, the author proposed an extended GP implementation and took into account the inaccuracy of user locations and unknown of path loss exponent.
The parameters of the path loss Gaussian are estimated according to the empirical Bayesian approach.
Steering Crowdsourced Signal Map Construction via Bayesian Compressive Sensing
incentive mechanism is essential
challenges:
- missing value inference
- crowdsourcing quality estimation
- incentive mechanism (incentive distribution map)
cost and reward
? how to construct projection matrix (using correlations)
mentioned:
- auction-based incentive designs
RMapCS: Radio Map Construction From Crowdsourced Samples for Indoor Localization
Use site survey to construct a radio map
offline radio map construction, Online fingerprinting localization
Challenges:
- Inaccurate sample annotation
- Measurement device diversity
- Nonuniform spatial distribution
Quality-Aware Sparse Data Collection in MEC Enhanced Mobile Crowdsensing Systems
将边缘计算和群智感知结合起来：
边缘节点进行一部分的计算，推断出未采样格子的状态，降低冗余和向服务器传输的数据量。
边缘节点能够实时感知用户移动模式，从而判断当前区域的参与者密度并进行调节。
在云端：
通过压缩感知恢复数据
