South Yorkshire Fire & Rescue
Search and Rescue 2020
scentro
WORKING FOR A SAFER SOUTH YORKSHIRE
Wearable Computing for Sensing and Navigation The aim of this project is to develop a novel wearable device based around a modified fire fighter's helmet that allows search & rescue workers to carry it into disaster areas both for local navigation and to relay information back to a control post. A key technology is the use of ultrasound sensors to detect nearby objects in environment and vibrators that transfer distance information to fire fighters by vibrating against the user's head.
Internet
Radio Satellite GSM/GPRS Wired Digital Voice
Aircrafts
Satellite
Operational Agents
Management Centre
Mobile Management Centre
Operational Field Vehicle
Vibrators Front Obstacle
Ultrasound Range Finders
Search Robotics Incident
Actuator Driver Board
Actuator Commands
Actuator Signal
Ultrasound, IMU data Microcontroller Board IMU Data
S&R Dogs Task Force Leader Manager
Mobile Centre Manager
IMU Command Ultrasound beam
Field Manager
Ultrasound, IMU, Actuator commands
Ultrasound data Ultrasound command
Hamideh Kerdegari:
[email protected]
Role of Voice Communication in Command and Control
Localisation and Mapping of Search and Rescue Assets
Audio stream
PRIOR Acoustic Analysis
Emotion Detection
Speech Recognizer Speech + Noise
N-best lists / Lattes of Word Sequences
Rescoring using Semantics
Convert to user-friendly format
Intent Detection Information Extraction
Identifying and structuring all of the incoming data from the dynamic environment of search and rescue mission can Semantic Knowledge maximise information content and provide situational awareness for the mission. Voice channels carry the most valuable information during crisis response among responders. Therefore, automatic speech recognition and conversation analysis can significantly enhance incident awareness. Variety of high acoustic noises makes it one of the most challenging environments for the current ASR systems. Integration of low-level information flowing on the voice channels with high-level information of the search and rescue context can generate the most likely interpretations in response to the spoken communications. Saeid Mokaram:
[email protected]
SLAM Update prior map using sensor readings
Rubble Ambiguous?
Improve position estimate based on updated map Input: Sensor readings
Furniture Ask operator to check
FINAL MAP Rescuer-friendly map showing danger areas (red) and survivor locations (orange)
Errors
The aim of this research is to provide improved Simultaneous Localisation And Mapping (SLAM) in partially collapsed buildings by tele-operated search and rescue robots to help make rescue operations safer and more efficient. This will be achieved by introducing two novel elements: using architectural building drawings to improve prior knowledge of the environment and using human input to distinguish between clutter/furniture and mapping errors. Christina Georgiou:
[email protected]