The energy consumption issue of mobile devices has been widely studied recently. For instance, Smartphone users may recharge their mobile devices multiple times a day especially when users surf on the Internet through Wi-Fi or 3/4G networks. Although smartphone producers have put much effort to prolong the daily usage time of smartphones, e.g., equipping smartphones with large-capacity batteries, this would increase the weights and sizes of smartphones. In this work, we propose a new method to schedule the usage of network based on users' habits by designing an APP named HOST on Android OS, which could maximally avoid opening inactive network operations. To test the efficiency of proposed algorithms, we conduct extensive experiments on smartphones with Android OS. The experimental results show that HOST reduces the battery loss by 64.2% in average when network is active, and enhances network bandwidth utilization by 23% in average. The percent of undesired interruption is around only 1%.

Hash table is a valuable data structure that is expected to provide constant amortized access time. Although there are a lot of researches on hashing, it seems there is no enough practical study on its stability with large data set. In this paper, we conducted a few experiments to study the performance of hashing with a large set of data and compared the results of different collision approaches. Our experiments revealed a few new phenomena. The experiment results leans to close addressing than open addressing by a huge edge and deem linear probing impractical due to its low performance. When items are randomly distributed with keys in a large space, different hash algorithms might produce similar performance. Increasing randomness in keys does not help hash table performance either. These discoveries might be able to provide heuristics to programmers on how to design software products using hash tables.

This study continued to investigate programming habits based on keystrokes, software quality, and code format. In previous work, we studied programmer's performance when they were working without pressure. In this work, we conducted an experiment by asking nineteen junior undergraduate students to complete a programming task under pressure. We also used a software tool to record the keystroke frequency and designed criteria to evaluation program quality. The experiment results demonstrate that while novice programmers are diverse in terms of programming styles, good ones tend to control execution in finer granularity and produce more solid code such as exception handling. Source code format seems to be a useful indicator for programming performance. It seems that there is still no direct correlation between the frequency of keystrokes and the quality of programs.

Coverage is a fundamental problem in wireless sensor networks (WSNs). From both economic and applicable concerns, designers always would like to provide guaranteed QoS of coverage of WSNs. In this paper, we address two path-coverage problems in WSNs, maximum k-support path coverage (a.k.a. best case coverage) and minimum k-breach path coverage (a.k.a. worst case coverage), in which every point on the desired resultant path is covered by at least k sensors simultaneously while optimizing certain objectives. We present two polynomial-time approaches to find optimal solutions for both maximum k-support coverage problem and minimum k-breach coverage problem. The time complexity of both algorithms are O(k2n log n), where n is the number of deployed sensor nodes and k is the coverage degree. In addition, a number of properties of kth-nearest point Voronoi diagram are presented, which is new to the literature.