Nowadays, smartphones have become ubiquitous and are playing a critical role in key aspects of people's daily life such as communication, entertainment and social activities. Most smartphones are equipped with multiple embedded sensors such as GPS (Global Positioning System), accelerometer, camera, etc, and have diverse sensing capacity. Moreover, the emergence of wearable devices also enhances the sensing capabilities of smartphones since most wearable devices can exchange sensory data with smartphones via network interfaces. Therefore, mobile sensing have led to numerous innovative applications in various fields including environmental monitoring, transportation, healthcare, safety and so on. While all these applications are based on two critical techniques in mobile sensing, which are data collection and data aggregation, respectively. Data collection is to collect all the sensory data in the network while data aggregation is any process in which information is gathered and expressed in a summary form such as SUM or AVERAGE. Obviously, the above two problems can be solved by simply collect all the sensory data in the whole network. But that will lead to huge communication cost. This dissertation is to reduce the huge communication cost in data collection and data aggregation in mobile sensing where the following two technical routes are applied. The first technical route is to use sampling techniques such as uniform sampling or Bernoulli sampling. In this way, an aggregation result with acceptable error can be can be calculate while only a small part of mobile phones need to submit their sensory data. The second technical rout is location-based sensing in which every mobile phone submits its geographical position and the mobile sensing platform will use the submitted positions to filter useless sensory data. The experiment results indicate the proposed methods have high performance.