There are typically three categories of multi-sensor systems. First, classical sensors system with different types of collocated sensors, e.g. a positioning system making use of a collocated inertial sensor, a pressure sensor and a GPS. Second, sensor joint systems wherein multiple same type of sensors coordinate to predict state of a system, e.g. estimating motion of a robotic or a human arm using multiple sensors attached to different positions, for capturing a versatile motion. The third kind of multi sensors system consists of collocated sensors with the same properties. The redundancy due to multiple sensors, results not only in enhanced noise performance of the system, but also allows the multi sensor system to achieve what single sensor system can not, e.g. a two dimensional array of accelerometers on a rigid circuit board can produce rotational information. On the one hand enhancing capabilities, shrinking size and reducing cost of MEMS sensors favor redundancy, but on the other hand data communication, processing and calibration compensation pose system level challenges.
The talk focused on technical merits of such multi-sensor systems. Talk covered the architecture of massive multi-IMU arrays with up to 288 measurement channels at 1 kHz, the engineering challenges associated with them including the requirements on on-node data processing, their merits and some applications.