Advanced Pedestrian Dead Reckoning and Map Filtering for Worker Localization (PDR)

Sort Description

This technology was developed to locate and track the upright and walking mobile worker with high accuracy and update rate. A commodity MMU (Motion Measurement Unit), for example the XSens MTI (consisting of triads of MEMS accelerometers, rate gyros and fluxgate magnetormeters) is firmly attached to footwear.
Using commodity sensor fusion software (from XSens) plus custom routines, individual step length and directions can be estimated with very low error. Distance travelled is accurate to ~1-2%. Outdoor paths are very accurate as the Earth’s magnetic field can be used for heading estimates and occasional GPS position fixes further increase accuracy. Indoors, local magnetic disturbances perturb heading estimates. Consequently, map filtering algorithms along with facility plans can be used to estimate very accurate paths. Depending on the detail available on the maps and on the availability of other coarse local positioning systems, e.g. WiFi ranging, location accuracies down to the meter level are possible.

Status

Proof-of-concept and off-line demonstrations have been performed. Integration of the MMU into end-user footwear/clothing and software into a PDA can be done in a matter of weeks. There is currently a wired connection from the foot-mounted IMU to an external computer via USB. A developed Firefighter “BlackBox” can provide preprocessing, storage and wireless relaying to an end-user computer, e.g. to a PDA, and could be rapidly configured for real-time use. Cork’s modular Map Filtering backend server software is ready for commercial exploitation.
If a paying and viable commercial venture were forthcoming, this technology could be taken to the industrial COTS state quite quickly. Miniaturization of the MMU and incorporation of a wireless data relay and power conditioning would required 6-12 man-months of effort but is relatively low-risk with currently available technology. However, qualifying such a wearable module for industrial or rugged use (e.g. explosion, water and safety testing for firefighting use) may require several man-years of effort and would mply a long-lead time. Custom sensor fusion algorithms adapted to the PDR problem (i.e. dispensing with XSens’s libraries) would definitely improve autonomous (i.e. non-map filtering) positioning performance but creating and testing these routines will require 6-12 man-months of effort from highly qualified personnel nowledgeable in GPS/INS/PDR Kalman Filtering state-of-the-art methods. Again, only a commercially viable project would justify this effort level.