The position of the sensor nodes in a WSN is very important for many applications, since measurements without a location where they were gathered are generally useless. Moreover, several routing protocol or self-organizing algorithms assume that the position of nodes is known. We consider the problem of real-time sensing and tracking the location of an indoor moving object.
Micro-Electro-Mechanical Systems sensors are used nowadays in a wide range of applications, from robotics to entertainment. Recently, this technology has been used also to equipe Inertial Measurement Units (IMUs) and Inertial Navigation Units (INSs) replacing traditional electro-mechanical sensors. However, all the data read from the sensing elements are affected by noise; therefore, the positioning process is subject to errors and the outcome could quickly diverge from the actual value.
On the other hand, although GPS (global positioning system) is a popular location estimation system, it does not work indoors, it consumes a notable amount of energy and it is quite expensive to get integrated in a dense network of energy constrained sensor nodes. Discovering the coordinates of sensor nodes using physical-layer measurements, such as pressure waves, time-of-flight or Ultra Wide Band technology, could enable indoor localization but would also require additional/dedicated hardware and it could be more energy-demanding. On the other hand, received signal strength indicator (RSSI) based localization techniques are attractive because of their simplicity: RSSI-based localization techniques does not require additional hardware, and does not have a significant impact on local power consumption, sensor size and thus cost since RSSI has become nowaday a standard feature in most wireless devices. Unfortunately, experimental studies have revealed that most of the state of the art localization algorithms which make use of the RSSI, once deployed in real testbeds, achieve much worse performance than what predicted by the simulation analysis, in particular in indoor scenarios.
Therefore, we propose WHYLOST (Wsn HYbrid LOcalization SysTem), an hybrid navigation system which combines positioning information obtained from an inertial navigation system and a short-range wireless reference system. The data produced by the INS lead to accurate short-term position estimates, but due to the drifts inherent to the system, these estimates are useless after some time. To solve this problem, we also get estimates using a wireless reference system to achieve accurate long-term position estimates.
We use some a data fusion algorithm based on Kalman ﬁltering with forward/backward smoothing to optimally combine the short- and long-term position estimates. We have implemented this localization system in a real-time testbed. The measurement results, which we obtained using the proposed method, show considerable improvements in accuracy of the location estimates.
First of all, we have developed a low-cost embedded IMU running on the top of a real-time kernel and a low-cost embedded microcontroller. The ERIKA enterprise real-time kernel is used to manage the computing tasks that perform sensor data sampling and processing. Moreover, low-cost MEMS accelerometers, gyroscopes and magnetometers provide the required data and a Kalman filter is used to process raw values and provide position estimations in real-time. This fully-functioning device is the main component of a complete inertial guidance system: data reported by the IMU are used to calculate current position based on velocity and time using a method known as dead reckoning.
Moreover, we believe that RSSI is still a promising indicator and it could be used to achieve accurate and low-cost indoor positioning in many case. Hence, we have realized two testbeds in two different environments (which could be representative of somehow typical indoor scenarios) in order to better understanding of the actual potentialities and limits of three common range-based localization algorithms: Min-Max, Multilateration an Maximum Likelihood. Our work complements other well-known researches and investigates some open issues in IEEE 802.15.4-based WSNs, such as
A lot of works have studied this issues in the past, mainly from a theoretical point of view or through simulations. On the other hand, a limited number of paper shows actual results obtained on the field and we tried to fill this gap through real tests.
Ongoing works are devoted to identify the optimal data-fusion technique in order to combine the 802.15.4 wireless-based localization systems and the estimates produce by the inertial unit.
All the software and the schematics will be freely available and released under the GNU GPL License.