Hjalmar Wennerström

A long-term study of correlations between meteorological conditions and 802.15.4 link performance

Outdoor wireless sensor networks are all exposed to a constantly changing environment that influences the performance of the network. In this paper, we study how variations in meteorological conditions influence IEEE 802.15.4 links. We show that the performance varies over both long and short periods of time, and correlate these variations to changes in meteorological conditions. The case study is based on six months of data from a sensor network deployed next to a meteorological research station running a continuous experiment, collecting both high-quality link and meteorological measurements. We present observations from the deployment, highlighting variations in packet reception ratio and signal strength. Furthermore, we show how the variations correlate with four selected meteorological factors, temperature, absolute humidity, precipitation and sunlight. Our results show that packet reception ratio and signal strength correlate the most with temperature and the correlation with other factors are less pronounced. We also identify a diurnal cycle as well as a seasonal variation in the packet reception ratio aggregated over all links. We discuss the implication of the findings and how they can be used when designing wireless sensor networks.

Location and density of rain gauges for the estimation of spatial varying precipitation

Abstract Accurate estimation of precipitation and its spatial variability is crucial for reliable discharge simulations. Although radar and satellite based techniques are becoming increasingly widespread, quantitative precipitation estimates based on point rain gauge measurement interpolation are, and will continue to be in the foreseeable future, widely used. However, the ability to infer spatially distributed data from point measurements is strongly dependent on the number, location and reliability of measurement stations. In this study we quantitatively investigated the effect of rain gauge network configurations on the spatial interpolation by using the operational hydrometeorological sensor network in the Thur river basin in north‐eastern witzerland as a test case. Spatial precipitation based on a combination of radar and rain gauge data provided by eteowiss was assumed to represent the true precipitation values against which the precipitation interpolation from the sensor network was evaluated. The performance using scenarios with both increased and decreased station density were explored. The catchment‐average interpolation error indices significantly improve up to a density of 24 rain gauges per 1000 km2, beyond which improvements were negligible. However, a reduced rain gauge density in the higher parts of the catchment resulted in a noticeable decline of the performance indices. An evaluation based on precipitation intensity thresholds indicated a decreasing performance for higher precipitation intensities. The results of this study emphasise the benefits of dense and adequately distributed rain gauge networks.

All Is Not Lost: Understanding and Exploiting Packet Corruption in Outdoor Sensor Networks

During phases of transient connectivity, sensor nodes receive a substantial number of corrupt packets. These corrupt packets are generally discarded, losing the sent information and wasting the energy put into transmitting and receiving. Our analysis of one years data from an outdoor sensor network deployment shows that packet corruption follows a distinct pattern that is observed on all links. We explain the patterns core features by considering implementation aspects of low-cost 802.15.4 transceivers and independent transmission errors. Based on the insight into the corruption pattern, we propose a probabilistic approach to recover information about the original content of a corrupt packet. Our approach vastly reduces the uncertainty about the original content, as measured by a manifold reduction in entropy.We conclude that the practice of discarding all corrupt packets in an outdoor sensor network may be unnecessarily wasteful, given that a considerable amount of information can be extracted from them.

Towards even coverage monitoring with opportunistic sensor networks

Opportunistic sensor networks typically rely on node mobility to monitor an area by collecting samples at different locations. In this paper we show how the mobility in combination with the periodic sampling of nodes causes large differences in the sensor coverage. We address this issue by leveraging simple heuristics based on local knowledge, employing an adaptive sampling scheme. The main insight is that areas where over-sampling is prevalent exhibit a high correlation with node contacts. Results obtained from both synthetic and real-world traces show that a dramatic decrease in oversampling of affected areas is achievable alongside a smaller increase of samples in more sparse areas.

Considering Multi-Contact Encounters in Opportunistic Networks

Opportunistic networks, often characterized by their intermittent connectivity and sparsity of nodes, has sparked routing in these networks to focus on isolated contact opportunities. Routing has predominantly been viewed as an exchange of messages between a pair of nodes. In this paper, we take a new look at contact opportunities, specifically focusing on the occurrence of having multiple simultaneous node contacts. Multi-contact encounters warrants a more holistic view of routing decisions, where more factors than just the features of a message-node tuple can be considered. We discuss these aspects and propose a protocol addition to leverage multi-contact opportunities with the notion of heterogeneous link quality, in order to limit energy consumption. The approach, implemented for the Epidemic, Spray-and-Wait and PRoPHETv2 protocols, re-evaluates routing decisions, weighting the routing metrics value against the estimated cost of the relay. Results indicate a two to three fold decrease in the number of messages lost, as well as a reduction in message relays, while maintaining a high delivery ratio for all three protocols.

IEEE 802.15.4 Channel Diversity in an Outdoor Environment

Low-power wireless link quality is known to be frequency-dependent because of multipath fading and other factors. We present a performance study of IEEE 802.15.4 radio links that quantifies and analyzes this frequency-specific performance in a clear-field outdoor environment. Using data from 16 channels on 240 links, we show that effect from channel selection on the average link is up to 4.89 dB, comparable with the effect from 38.7°C change in temperature. These results provide a performance baseline for other environments, as the diversity can be expected to further increase in environments with more obstacles and external interference.

A game theoretic approach to sensor data communications in an opportunistic network

Opportunistic communication coupled with a sensing task enables the collection and spreading of sensory information in areas without global connectivity, providing useful information in challenging environments. In this paper, we consider an opportunistic sensor network where the mobility of users enables both the measurement and spreading of sensor data. We motivate user participation through a game theoretic approach, which is designed to ensure a fair and efficient exchange of sensor messages. The message exchange is modeled as a two-player game where sensor measurements are exchanged between nodes in a contrite tit-for-tat manner. The proposed game captures the nodes desire to limit energy consumption while at the same time obtaining messages containing useful information. We show that the best response in the game is a Pareto optimal subgame perfect equilibrium. The game is evaluated through simulation in a realistic scenario and compared with three other approaches, generating the best overall efficiency by striking a balance between size and content of messages.

Auto-tagging of images in non-english languages using tag language conversion

Utilization of web images with social tags as training data has been a major trend for the development of automatic image tagging/classification systems. While the amount of information available on web sites such as Flickr is abundant, the majority of information obtained from such sites is in English, the dominant language on the web. This linguistic unbalance is expected to affect auto-tagging results in a negative way for non-English users, who demand image tags in their native languages. The objective of this research is to develop an image auto-tagging system which can generate tags in languages other than English. This paper examines the effect of linguistic unbalance in training data to construct auto-tagging systems, which aim to generate tags in minor languages. Furthermore, we propose methods which utilize an auto-tagging model generated on English training data, and convert the auto-tagging results to the target language. Subjective evaluations show that the proposed method is capable of generating auto-tagging results with better quality than conventional approaches.