Mid Sweden University

The ease of acquiring hardware-based link quality indicators is an alluring property for fast channel estimation in time- and safety-critical Wireless Sensor Network (WSN) applications, such as closed-loop control and interlocking. The two rudimentary hardware-based channel quality metrics, Received Signal Strength (RSS) and Link Quality Indicator (LQI), are the key constituents of channel estimation and a plethora of other WSN functionalities, from routing to transmit power control. Nevertheless, this study highlights three deficient aspects of these two indicators: 1) overall deceptiveness, i.e. the inability to reveal the presence of interference, falsely indicating excellent channel conditions in an unacceptably high fraction of cases; 2) the burstiness of missed detections, which compromises the attempts to eliminate the deceptiveness by averaging; 3) high mutual discrepancy of the two indicators, observed in 39-73% of packets, throughout different scenarios. The ability of RSS and LQI to indicate IEEE 802.11 interference is scrutinized in a variety of scenarios in typical industrial environments, using commercialoff- the-shelf hardware and realistic network topologies, giving the findings of this study a high general validity and practical relevance.

The knowledge of error nature in wireless channels is an essential constituent of efficient communication protocol design. To this end, this paper is the first comprehensive bit- and symbol-level analysis of IEEE 802.15.4 transmission errors in industrial environments. The intention with this paper is to extract the error properties relevant for future improvements of wireless communication reliability and coexistence of radio systems in these harsh conditions. An extensive set of bit-error traces was collected in a variety of scenarios and industrial environments, showing that error behavior is highly dependent on the cause of packet corruption. It is shown that errors inflicted by multipath fading and attenuation exhibit different properties than those imposed by IEEE 802.11 interference. The statistical behavior of these two patterns is concurrently investigated in terms of differences in bit-error distribution, error burst length, channel memory length, and the scale of packet corruption. With these conclusions at hand, abiding to the computational constraints of embedded sensors and the statistical properties of bit-errors, a Reed-Solomon $(15,k)$ block code is chosen to investigate the implications of bit-error nature on practical aspects of channel coding and interleaving. This paper is concluded by a number of findings of high practical relevance, concerning the optimal type, depth, and meaningfulness of interleaving.

Wireless sensor network communication in industrial environments is compromised by interference, multipath fading, and signal attenuation. In that respect, accurate channel diagnostics is imperative to selecting the adequate countermeasures. This paper presents the lightweight packet error discriminator (LPED) that infers the wireless link condition by distinguishing between errors caused by multipath fading and attenuation, and those inflicted by interfering wideband single-channel communication systems (e.g., IEEE 802.11b/g), based on the differences in their error footprints. The LPED uses forward error correction in a novel context, namely, to determine the symbol error density, which is then fed to a discriminator for error source classification. The classification criteria are derived from an extensive set of error traces collected in three different types of industrial environments, and verified on a newly collected set of error traces. The proposed solution is evaluated both offline and online, in terms of classification accuracy, speed of channel diagnostics, and execution time. The results show that in ≥91% of cases, a single packet is sufficient for a correct channel diagnosis, accelerating link state inference by at least 270%, compared with the relevant state-of-the-art approaches. The execution time of LPED, for the worst case of packet corruption and maximum packet size, is below 30 ms with ≤3% of device memory consumption. Finally, live tests in an industrial environment show that LPED quickly recovers from link outage, by losing up to two packets on average, which is only one packet above the theoretical minimum.

Communication reliability is the ultimate priority in safety-critical wireless sensor network (WSN) communication. Surprisingly enough, the enormous potential of error control on direct sequence spread spectrum (DSSS) chips in IEEE 802.15.4 has been completely overlooked by the WSN community, possibly due to incorrect presumptions, such as the concerns about computational overhead. Current error-correction schemes in WSN counteract the error process once the errors have already propagated to bit- and packet-level. Motivated by the notion that errors should be confronted at the earliest stage, this work presents CLAP, a novel method that tremendously improves the error correction in WSN by fortifying the IEEE 802.15.4 Physical layer (PHY) with straightforward manipulations of DSSS chips. CLAP is implemented on a software-defined radio platform, and evaluated on real error traces from heavily WLAN-interfered IEEE 802.15.4 transmissions at 3 different environments. CLAP boosts the number of corrected packets by 1.78-6.88 times on severely interfered links, compared to two other state-of-the-art schemes. The overhead in terms of computational complexity is about 10% of execution time of the OQPSK demodulator in the legacy IEEE 802.15.4 receiver chain.

The requirements of safety-critical wireless sensornetwork (WSN) applications, such as closed-loop control andtraffic safety, cannot be met by the IEEE 802.15.4-2006 standardnor its industrial WSN (IWSN) derivatives. The main problem inthat respect is the communication reliability, which is seriouslycompromised by 2.4-GHz interference, as well as multipathfading and attenuation (MFA) at industrial facilities. Meanwhile,communication blackouts in critical WSN applications maylead to devastating consequences, including production halts,damage to production assets and can pose a threat to safetyof human personnel. This work presents PREED, a method toimprove the reliability by exploiting the determinism in IWSNcommunication. The proposed solution is based on the analysisof bit error traces collected in real transmissions at four differentindustrial environments. A case study on WirelessHART packetformat shows that PREED recovers 42%-134% more packetsthan the competing approaches on links compromised by WLANinterference. In addition, PREED reduces one of the most trivialcauses of packet loss in IWSN, i.e. the corruption offrame lengthbyte, by 88% and 99%, for links exposed to WLAN interferenceand MFA, respectively.