Reliable sensor networks for planet exploration

摘要

Wireless sensor networks will play a critical role in space and planet exploration, allowing remote monitoring of non-easily accessible areas in preparation of human or robotic missions. Sensors, however, are fragile and can fail, reporting erroneous measurements, for example. Decisions derived from flawed sensor measurements can adversely impact the correctness of the overall sensor network findings and may jeopardize the success of the mission. Unfortunately, failed sensors in space cannot be easily diagnosed and replaced. To improve the reliability of decisions and minimize the impact of faulty sensor measurements, the preferred approach is to exploit data redundancy. In this paper, we present confidence weighted voting (CWV), a distributed technique that can greatly improve the data reliability and fault tolerance of sensor network applications. We evaluate CWV against traditional approaches (e.g., majority voting (MV) and distance weighted voting (DWV)), in the presence of flawed sensors. The results show that CWV consistently outperforms the other schemes by providing as much as 49% more resiliency.