The focus of this project is to provide methods for calibration of sensor nodes in sensor networks. The importance of the calibration problem is to compensate for the sensor reading drifts that occur due to systematic errors, noise or sensor degradation. The objective is to provide calibration techniques that apply on collaborative sensors autonomously, i.e. without supervision. Our work involves: a) distributed procedures to identify erroneous sensors; and b) a simulator framework for a sensor net drift calibration setups.
Overview• Motivation. The deployment of sensor networks has been growing in scale and scope with new applications in embedded and challenging environments. This progress has been facilitated by advances in nanoscale electronics that can be integrated with MEMS optical and biochemical technologies to build tiny sensor nodes, [1]. However, there are several issues and difficulties in building such sensor nets. In addition to reliability concerns of the sensor nodes and their communications, there is a significant problem with the drift of sensor readings from their correct values. A variety of sensor applications use MEMS for realizing the transduction mechanism. Along with degradations in the micromechanical parts due to environmental effects, wear and tear [2, 3], the signal conditioning and read-out circuitry also suffer from both systematic and random degradations inducing errors in the sensor reads.Sensor calibration has been used over time to correct the reading drifts. Traditionally, calibration is applied on sensors at the micro level, meaning on individual sensor nodes either at the factory or in the network off-line. However, micro-level calibration may not be feasible due to many difficulties such as remote access, security, size of sensor net and device degradations. There is need for network based calibration, i.e. autonomous calibration by collaborating sensor nodes.• Related Work. The problem of calibration of sensor nodes has received considerable attention in research works. Earlier calibration was performed on each sensor at the factory or in the field using simple built-in techniques [4]. Calibration in traditional sensor networks is still being done individually.Most of the recent research has focused on location discovery of sensor nodes. The early work on SpotON [5] modeled the signal to distance relation between transmitting and receiving sensors off-line. Another project targeting location discovery is Calamari, [6], which formulates a calibration approach to sensor localization as a parameter estimation problem. A two-phase collaborative technique is used in [7] where first all pair-wise calibration functions are found and then they are optimized to produce the global calibration function for the net. In the SCAAT project [8], a mathematical technique is used for tracking position and orientation while performing sensor autocalibration. Dynamic fine-grain localization is reported in [9]. More recently. a calibration approach to location estimation is taken in [10] using ...