Crowd-sourced collaborative sensing in highly mobile environments

2016-06-20Networked sensing has revolutionized various aspects of our lives. In particular, it has allowed us to minutely quantify many aspects of our existence: what we eat, how we sleep, how we use our time, and so forth. We have seen such quantification from the smart devices we use daily, such as smartphones and wearable devices. Those smart devices usually have more than ten high precision sensors to sense both internal and external information. Another domain that will likely to see such quantification in near future is automobiles. Modern vehicles are equipped with several hundred sensors that govern the operation of internal vehicular subsystems. Those sensors from both smart devices and automobiles, coupled with online information (cloud computing, maps, traffic, etc.) and other databases as well as crowd-sourced information from other users, can enable various forms of context sensing, and can be used to design new features for both mobile devices and vehicles. We abstract those aspects for context sensing into three parts: mobile and vehicular sensing, cloud assistance and crowdsourcing. Though each part itself comes with different challenges, accurate context sensing usually requires a careful combination of one or more of the three aspects, which brings new challenges for designing and developing context sensing systems. In this dissertation, we focus on three challenges, Programmability, Accuracy and Timeliness, in designing efficient and accurate context sensing system for mobile devices and vehicles. We will leverage the mobile and vehicle sensors, cloud information and crowdsourcing, collectively to ease context sensing programming, improve context sensing accuracy and timeliness. ❧ First, for Programmability, we focus on programming context descriptions using information from cloud and vehicle sensors. As more sensor-based apps are developed for vehicular platforms, we think many of these apps will be programmed using an event-based paradigm, where apps try to detect events and perform actions on detection. However, modern vehicles have several hundred sensors, these sensors can be combined in complex ways together with cloud information in order to detect some complicated context, e.g. dangerous driving. Moreover, these sensor processing algorithms may incur significant costs in acquiring sensor and cloud information. Thus, we propose a programming framework called CARLOG to simplify the task of programming these event detection algorithms. CARLOG uses Datalog to express sensor processing algorithms, but incorporates novel query optimization methods that can be used to minimize bandwidth usage, energy or latency, without sacrificing correctness of query execution. Experimental results on a prototype show that CARLOG can reduce latency by nearly two orders of magnitude relative to an unoptimized Datalog engine. ❧ Second, for Accuracy, we focus on automotive positioning accuracy. Positioning accuracy is an important factor for all kinds of context sensing applications for automobiles. Lane-level precise positioning of an automobile can improve navigation experience and on-board application context awareness. However, GPS by itself cannot provide such precision in obstructed urban environments. We propose a system called CARLOC for lane-level positioning of automobiles which carefully incorporates the three aspects in context sensing. CARLOC uses three key ideas in concert to improve positioning accuracy: it uses digital maps to match the vehicle to known road segments; it uses vehicular sensors to obtain odometry and bearing information; and it uses crowd-sourced location estimates of roadway landmarks that can be detected by sensors available in modern vehicles. CARLOC unifies these ideas in a probabilistic position estimation framework, widely used in robotics, called the sequential Monte Carlo method. Through extensive experiments, we show our system achieves sub-meter positioning accuracy even in obstructed environment, which is an order of magnitude improvement over a high-end GPS device. ❧ Finally, for context sensing applications, Timeliness is another important problem we need to take care of. We consider how to ensure the timeliness and availability of media content from mobile devices. Motivated by an availability gap for visual media, where images and videos are uploaded from mobile devices well after they are generated, we explore the selective, timely retrieval of media content from a collection of mobile devices. We envision this capability being driven by similarity-based queries posed to a cloud search front-end, which in turn dynamically retrieves media objects from mobile devices that best match the respective queries within a given time limit. We design and implement a general crowdsourcing framework called MediaScope that supports various geometric queries and contains a novel retrieval algorithm to maximize the retrieval of relevant information. In experiments on a prototype, our system achieves near optimal performance under different scenarios