WP4 SafePASS Passenger Chatbot for Emergency Audio Navigation
The SafePASS Passenger Chatbot is a hardware-software component, part of the lifejacket’s Smart Environment, whose role is to provide real-time audio instructions to passengers for navigation within the ship during evacuation.
The Chatbot, based on the hardware components shown below, has been custom designed to fit into the Smart Lifejacket in terms of size and weight, and provides all the required capabilities for its scope.
Development Board and Custom Design of the Passenger Chatbot
Given its small dimensions (i.e. 4x4 cm), its processing capabilities necessitate that less complex algorithms can be run on it and, thus, the module should focus more on functionality and less on accuracy, since its main goal is to provide simple audio instructions to the passenger. Another factor which has been taken into account for its design concerns the connectivity with other SafePASS components in order to de-centralize its operation and make it more efficient in managing its processing power. This is the reason why the personalized evacuation routes are not locally calculated but at its centrally deployed backend (Chatbot Backend) which then connects to the other SafePASS components, i.e. Core Engine, LDER etc.
B) Overview of Features
When the cruise ship’s management system activates the evacuation signal, i.e. the ship enters into emergency evacuation mode, the passengers are instructed to reach the indicated closest muster station, where a personalized lifejacket is handed to them by the ship’s crew. Once the Smart Equipment, mounted on the lifejacket, is turned on then the Chatbot application begins its operation, aiming to safely navigate the passenger to the emergency exit indicated per passenger via comprehensive audio instructions.
The Chatbot initially requests the passenger’s ID and default language (from the SafePASS Core Engine) and the passenger’s current location in the ship (from the Smart Lifejacket’s UWB component) and receives the Active Evacuation Route for the specific passenger (from the Crowd Simulator/LDER component). The Chatbot then starts its audio navigation functionality by playing back audio tracks which contain instructions dependent on the passenger’s current location and the indicated personalized evacuation route. The passenger can listen to these instructions through the wired earplug connected to the Chatbot’s hardware board.
The overall scope of the Passenger Chatbot is to navigate the passenger using audio instructions, which explain the personalized evacuation route that the passenger needs to follow until the emergency exit is reached. Along the way, if the passenger deviates from the indicated evacuation route or if the evacuation route changes for other reasons (e.g. a nearby exit is blocked or flooded), then the Chatbot dynamically updates its instructions in order to guide the passenger to the correct (updated) exit in the ship.
Additionally, the Chatbot supports different languages which can be selected by the passenger in case the default one (which has been pre-selected during the passenger’s check-in into the cruise ship) is not the desired one.
C) Description of Functionalities
The Passenger Chatbot has been designed to provide the functionalities presented in detail below:
Audio Instructions Playback: playback audio tracks which have already been pre-recorded and stored on its embedded SD card.
Each audio track consists of 2-5 seconds of instructions in different languages for the passenger to be guided from point A to point B. Initially two sets of audio tracks (in Greek and English) have been recorded, while more languages will be supported soon (including French, Spanish and German). In this way, the passenger’s movement between a pair of Points of Interest (POIs) in the ship corresponds to a different recorded audio track, which is named based on the POIs it connects and the language, e.g. audio tracks “10-11_EN.mp3” and “11-10_EN.mp3” are different since they correspond to instructions (in English) for the same POIs connection but for different directions.
Language Selection: passengers to be able to change the default language of the instructions and select one of the supported ones.
In the current Chatbot prototype, the passenger can manually change between English and Greek by simply clicking the earplug’s button. In the final Chatbot prototype, automatic language detection will be available based on the passenger’s voice with the use of advanced Artificial Intelligence (AI) techniques, i.e. “keyword spotting” technique. In this way, the Chatbot will recognise various spoken words and understand the language they are part of, thus changing the audio instructions language for all the supported languages, including Greek, English, French etc.
Location-based Audio Navigation: real-time navigation via audio instructions based on the passenger’s location, direction and speed.
This functionality necessitates connection to other SafePASS systems available in the Core Platform (Core Engine and LDER) or the Smart Lifejacket (UWB component). Based on this information, the Chatbot matches the passenger’s personalized evacuation route to the defined POIs of this route and, thus, a POI-to-POI list is created which is then matched with the pre-recorded audio tracks on its embedded SD card. The Chatbot’s audio instructions consist of continuous playback of the appropriate audio tracks in the correct order from one POI to the next, while considering the selected language and the passenger’s walking speed so that the playback rhythm is adjusted.