🔒
Trusted Engineering Publisher
Serving Researchers Since 2012

Smart Emergency Response and Ambulance Assistance System

DOI : 10.5281/zenodo.24446484
Download Full-Text PDF Cite this Publication

Text Only Version

Smart Emergency Response and Ambulance Assistance System

1st Mr. Ashitosh Ingale

Department of Computer Engineering

D. Y. Patil Education Society (Deemed to be University)

School of Engineering and Management Kolhapur, India

2nd Ms. Siddhi Patil

Department of Computer Engineering

D. Y. Patil Education Society (Deemed to be University)

School of Engineering and Management Kolhapur, India

3rd Ms. Manasi Patil

Department of Computer Engineering

D. Y. Patil Education Society (Deemed to be University)

School of Engineering and Management Kolhapur, India

4th Ms. Samiksha Patil

Department of Computer Engineering

D. Y. Patil Education Society (Deemed to be University)

School of Engineering and Management Kolhapur, India

5th Ms. Tanuja Gurav

Department of Computer Engineering

D. Y. Patil Education Society (Deemed to be University)

School of Engineering and Management Kolhapur, India

6th Mr. Satish Madhukar Ranbhise

Department of Computer Engineering

Y. Patil Education Society (Deemed to be University)

School of Engineering and Management Kolhapur, India

Abstract – In a road accident or medical emergency the rst few minutes matter most, and they are often wasted. The victim may be unconscious and unable to say who they are or what conditions they have. Bystanders rarely know whom to call. If several of them phone for help about the same crash, dispatchers receive overlapping reports that someone has to sort out by hand, and language differences between callers and responders add more delay. We present ResQAI, an Android-based emergency coordination platform that tackles these problems with a single incident record shared by victims, bystanders and responders. Every registered user gets a unique Emergency ID and a matching QR code. Scanning the code shows a bystander or responder only what is needed for immediate help, not the rest of the persons data. Reports about the same event are matched and grouped under one Incident ID, which cuts down duplicate dispatch requests. The app records GPS location and basic incident details, supports several languages, and, if the user has given permission, can bring in veried health information such as ABHA records from the Ayushman Bharat Digital Mission. From this information the system estimates how severe the incident is and prepares a structured summary for the right service: ambulance, police, or rescue and trafc units. ResQAI is meant to work alongside Indias ERSS 112 service and professional medical care, not to replace them. We describe the architecture, the data ow and a six-step implementation process covering registration, reporting, QR identication, incident creation and duplicate linking, assessment, and response and closure. We also propose how the system should be evaluated: response time, in- formation accuracy, incident-matching accuracy, communication success and usability. This paper presents the design and the evaluation plan only; we report no measured results.

Index TermsEmergency response, incident management, An- droid application, emergency QR, duplicate detection, multi- lingual communication, ambulance coordination, Kotlin, Spring Boot

    1. INTRODUCTION

      Fast, coordinated response matters a great deal in road accidents and medical emergencies, but a few practical prob- lems keep slowing it down. Victims are often unconscious or confused and cannot tell anyone who they are or what their medical history is, and the people who stop to help usually cannot nd out either. Often several people at the scene call emergency services separately, so responders receive overlapping reports that they have to reconcile by hand.

      ResQAI addresses these gaps with one coordination layer that connects victims, bystanders and responders through a single incident record. It is intended to run alongside existing infrastructure such as Indias ERSS 112 service, as a layer for coordination and information rather than a substitute for professional emergency response.

      1. Novelty and Key Contributions

        Existing tools usually offer identication, location sharing or emergency reporting as separate features that are not tied to a shared incident. This work makes the following contributions:

        1. Incident-Centric Coordination: One Incident ID ties together the reports from victims, bystanders and respon- ders, so several reports about the same accident become a single record instead of several separate events.

        2. Emergency ID and Privacy-Preserving QR: Each user has a unique Emergency ID and QR code. Scanning it gives a bystander or responder only the information the user has authorized for emergencies.

        3. Multilingual Communication: Information is organized and shown in several languages, so reporters, victims and responders are less likely to misunderstand one another.

        4. Authorized Health-Information Context: Veried health information is added to a report only if the user has allowed it, so medical context is available without giving up privacy.

        5. Structured Emergency Assessment: The collected details are used to estimate severity and to send a structured report to the right ambulance, police, or rescue and trafc service.

    2. LITERATURE REVIEW

      Most earlier work looks at one piece of the problem at a time. Indias Ayushman Bharat Digital Mission issues the ABHA number, a digital health identier that allows authorized access to health records [1]. In an emergency such records could give useful medical context, as long as authorization and privacy are respected. On the response side, the Emergency Response Support System (ERSS) runs the single emergency number 112 and brings voice calls, SMS and SOS alerts into one central dispatch service [2]; its mobile app adds location-aware alerts [3]. These channels depend on the victim or a nearby person having a working phone and being able to describe what happened. Mobile SOS apps likewise offer location sharing, emergency contacts and alerts, but when several people report the same accident they can still produce duplicate or uncoordinated requests.

      Work on language models in healthcare adds another angle. A systematic review of large language models in patient care [4] points to growing interest in clinical settings. The World Health Organization (WHO) guidance on prehospital emergency care stresses timely, well-organized care before the patient reaches hospital [5], and help from bystanders matters most when the victim is unconscious, injured or unable to communicate.

      Wider surveys cover the applications, advances and chal- lenges of large language models in medicine and healthcare [6], and a scoping review summarizes how they are used in medicine [7]. Research on compact models shows that language models with under a billion parameters can be tuned

      to run on a device [8], and another review looks at transformer- based models in healthcare [9]. Together these suggest that language support and structured summaries are becoming realistic for time-critical uses such as emergency reporting, although their reliability has to be established before anyone relies on them for clinical decisions.

      At the population level, the WHO Global Status Report on Road Safety 2023 [10] documents how many people are harmed by road trafc injuries, which is why the speed and quality of the rst response to road accidents remain a public- health concern.

      Research Gap: Current systems tend to handle victim iden- tication, location haring, duplicate handling, health-record access and language support separately rather than around a common incident. What is missing is a platform that con- nects victims, bystanders and responders through one incident. ResQAI lls this gap by combining an Emergency ID, a privacy-preserving QR code, incident management, duplicate detection, location capture and multilingual communication.

    3. SYSTEM ARCHITECTURE AND WORKFLOW

      ResQAI uses a centralized architecture built around the inci- dent. Four kinds of participants take part: victims, bystanders, emergency responders and platform administrators. Instead of treating each report as a separate transaction, the system sends every report, location, Emergency ID reference and any autho- rized supporting information through one processing pipeline. The pipeline classies the incident and detects duplicates, then forwards a single consolidated, structured record to the appropriate response service.

      1. Workow Diagram

        At a high level, an accident leads to a report from the victim (through the app, a scanned QR code or their Emergency ID) or from a bystander (a GPS-tagged report). These reports go into an incident engine that detects duplicates, identies the victim and, where authorization allows, checks medical data against ABHA-linked records. The structured result goes to a decision layer that classies the emergency and routes it to the relevant service (ambulance/EMS, police, or rescue and trafc), and the hospital receives the relevant details as needed. Fig. 1 shows the whole ow.

      2. Implementation Architecture

        Fig. 2 shows how the prototype implements this workow in four tiers. The Android app contains everything a phone owner needs when there is no data connection: a Room database holding the local emergency prole, a Kotlin rule- based engine that classies the emergency through a short dialogue, live GPS with an optional ofine PIN-code lookup, an on-device message builder with Android text-to-speech, a call manager that dials the congured emergency number over the normal cellular network, and a WorkManager queue that retries pending reports when connectivity returns. A bystander uses an online client to scan the victims Emergency QR code,

        ACCIDENT

        Victim App / QR / ID

        INCIDENT ENGINE

        Bystander QR / GPS report

        Duplicate Detection

        Victim Identication

        Medical Data Check

        ABHA

        Authorized Data

        AI DECISION ENGINE

        No ABHA

        ResQAI Prole

        Emergency Classication

        Ambulance

        / EMS

        Police

        Rescue / Trafc

        HOSPITAL

        Fig. 1. ResQAI system workow: from incident occurrence through detection, classication, and response routing.

        pick the emergency type and submit the report together with their own GPS position.

        Online requests go to a Spring Boot backend. It validates the Emergency ID, checks that access to the victims stored medical prole is authorized, builds the incident context (vic- tim, emergency and location), and runs the call workow, in- cluding status and error handling. Data is stored in MongoDB Atlas, speech is generated with the Sarvam Speech API, and outbound emergency calls go through Twilio.

      3. Data Flow

        The operational data ow proceeds as follows:

        • A victim or bystander starts a report in the ResQAI app.

        • The app records the location, the Emergency ID if one is available, and a description of the incident.

        • The submitted data is stored securely and processed in the backend database.

        • The system compares each new report with existing records. If it matches one, the report is linked to that Incident ID instead of creating a new one.

        • The consolidated information is sent to the appropriate emergency service.

        • Responders use this single incident record to coordinate their work in the eld.

      Fig. 3 shows the two paths that run in parallel at the implementation level. In the ofine path, holding the SOS control loads the local prole, classies the emergency, gets live GPS and, if a veried dataset is installed, an ofine PIN-code lookup, and builds the emergency message. The message is played aloud through Android text-to-speech, and

      the call manager places a normal cellular call to the congured number. One limitation is that the text-to-speech audio plays on the device and is not automatically sent into the call. In the online path, the bystanders QR scan, emergency type and GPS position go to the backend, which validates the Emergency ID, retrieves the authorized victim prole, builds the incident context, generates speech and starts a Twilio call to the emergency recipient. Incident and report status are saved, and any report that could not be sent is retried by the WorkManager queue.

    4. IMPLEMENTATION METHODOLOGY

      The implementation follows a xed sequence of steps, chosen to keep the delay between report and response short and the system reliable.

      1. Methodology Framework

        Fig. 4 gives the lifecycle that follows once a user has registered, from the rst report to coordinating the response. The execution steps in Section IV-B describe the full process in more detail, beginning with registration.

      2. Execution Steps

        1. User Registration and Emergency ID Generation: When a user registers, they receive a unique Emergency ID and a privacy-preserving QR code. The code lets people identify the user quickly at an emergency scene without exposing sensitive personal details directly.

        2. Emergency Reporting: A victim or bystander reports an accident or medical emergency in the app and gives ba- sic details: the location, the victims apparent condition and how many people are involved.

        3. QR Scanning and Identication: A bystander, or a re- sponder who arrives later, scans the victims Emergency QR code. The system returns only the information that has been authorized for release, so unrelated personal or medical data stays hidden.

        4. Incident Creation and Processing: The platform creates a unique Incident ID for each new emergency. When several independent reports describe the same event, they are recognized and merged under that Incident ID, so one accident is not handled several times.

        5. Emergency Assessment and Communication: The col- lected information is analyzed to estimate how severe the emergency is and what kind of help is needed. Multilin- gual support is applied at this point so that the reporter and the responding team both receive the information clearly, which lowers the risk of misunderstanding across languages.

        6. Response Coordination and Closure: The nished, struc- tured incident record goes to the relevant emergency response service. It keeps being updated as the response proceeds and is closed once the emergency is resolved.

      Android app: phone owner (ofine-capable)

      WorkManager queue

      retry when online

      Call manager

      normal cellular call

      Message builder + TTS

      ofine speech playback

      GPS + PIN lookup

      ofine PIN dataset

      Rule-based engine

      Kotlin classication

      Room database

      local emergency prole

      Bystander client (online)

      Scan Emergency QR

      Select emergency type

      Capture bystander GPS

      Spring Boot backend

      Validate Emergency ID

      authorize data access

      Build incident context

      victim, emergency, location

      Call workow

      status and error handling

      MongoDB Atlas

      victim medical prole

      Sarvam Speech API

      speech generation

      Twilio

      outbound emergency call

      Online ow: bystander

      Scan Emergency QR: bystander at scene

      Report type: choose emergency, submit

      Capture GPS: bystander location

      Spring Boot API: validate ID, authorize

      MongoDB Atlas: victim medical prole

      Build context: victim, emergency, location

      Sarvam Speech API: generate speech

      Twilio: call emergency recipient

      Get location: live GPS, ofine PIN lookup

      Classify: Kotlin rule-based dialogue

      Load prole: Room database

      Ofine ow: phone owner

      SOS hold: phone owner trigger

      Fig. 2. Implementation architecture. Shaded frames are application tiers; dashed boxes are external data stores and services.

      Save status: incident, WorkManager retry

      Place call

      normal cellular

      Local TTS

      speech playback

      Build message: prole, type, location

      Limitation: TTS not sent into call audio

      Fig. 3. Implementation data ow. Dashed boxes are external services; the dotted box marks a known limitation of the ofine path.

      Stage 1: Emergency Reporting

      • Victim or bystander reports an accident or medical emergency

      • Basic details (location, condition, victim count) are collected

      Stage 2: Emergency Identication

      • Unique Emergency ID and privacy-preserving QR generated or used

      • QR identies the victim without exposing sensitive personal data

      Stage 3: Incident Processing

      • Incoming reports are checked against existing incidents

      • Duplicate reports are linked via a common Incident ID

      Stage 4: Emergency Assessment

      • Information is processed to classify the emergency and the required response

      • Multilingual communication organizes and conveys information clearly

      Stage 5: Response Coordination

      • Structured incident information is shared with appropriate services

      • Coordination takes place alongside existing services such as ERSS 112

      Fig. 4. Implementation methodology: ve-stage end-to-end operational life- cycle.

      TABLE I

      PROPOSED EVALUATION CRITERIA FOR THE RESQAI SYSTEM

      Criterion What is measured

      Response Time Time between the rst emergency report and the

      moment response coordination begins.

      with optional, authorized medical-record integration should make the information reaching responders clearer and more complete. ResQAI is not meant to replace infrastructure such as ERSS 112; it is a complementary coordination layer. The goal is to make emergency information sharing faster, better organized and more secure for everyone involved. Our next step is to carry out the evaluation described above and report measured results for each of the ve criteria.

      REFERENCES

      1. National Health Authority, Ayushman Bharat Digital Mission: ABHA Number, Government of India, New Delhi, India, 2026. [Online]. Accessed: Aug. 17, 2026.

      2. Ministry of Home Affairs, Government of India, Emergency Response Support System (ERSS) 112, New Delhi, India. [Online]. Accessed: Aug. 15, 2026.

      3. Ministry of Home Affairs, Government of India, Emergency Response Support System: Features, New Delhi, India. [Online]. Accessed: Aug. 15, 2026.

      4. F. Busch, L. Hoffmann, C. Rueger, et al., Current applications and challenges in large language models for patient care: A systematic review, Commun. Med., vol. 5, 2025.

      5. World Health Organization, Prehospital Emergency Care: Operational Guidance for Ambulance Systems. Geneva, Switzerland: WHO, 2025.

      6. D. Wang and S. Zhang, Large language models in medical and healthcare elds: Applications, advances, and challenges, Artif. Intell. Rev., vol. 57, p. 299, 2024.

      7. X. Meng, X. Yan, K. Zhang, et al., The application of large language

        Information Accuracy

        How correct the captured location, victim and incident details are compared with ground truth.

        models in medicine: A scoping review, iScience, vol. 27, no. 5, p. 109713, 2024.

        Incident Matching How accurately duplicate reports are detected and

        linked under one Incident ID.

      8. Z. Liu, C. Zhao, F. Iandola, et al., MobileLLM: Optimizing sub-billion parameter language models for on-device use cases, in Proc. 41st Int.

        Communication Success

        How often emergency information reaches responders and other stakeholders and is understood by them.

        Conf. Mach. Learn. (ICML), 2024, pp. 3243132454.

      9. S. Nerella, S. Bandyopadhyay, J. Zhang, et al., Transformers and large language models in healthcare: A review, Artif. Intell. Med., vol. 154,

      Usability Ease of use, reliability and user satisfaction.

      p. 102900, 2024.

      [10] World Health Organization, Global Status Report on Road Safety 2023.

      Geneva, Switzerland: WHO, 2023.

    5. EXPECTED OUTCOMES AND DISCUSSION

      ResQAI is expected to produce a working emergency- coordination app that supports reporting by victims and bystanders, unique Emergency IDs, privacy-preserving QR identication, incident management, duplicate detection, GPS location capture and a preliminary estimate of severity. We also expect it to make shared information better organized and clearer through multilingual support and, where the user has authorized it, veried health records.

      A. Evaluation Criteria

      Table I lists the criteria we propose for evaluating the system.

    6. CONCLUSION

  1. ResQAI takes an incident-centric approach to coordinating the response to road accidents and medical emergencies. Rather than relying on separate reporting tools, it links vic- tims, bystanders and emergency services through one shared platform. The unique Emergency ID and privacy-preserving QR code allow quick identication at the scene while limiting how much sensitive information is exposed. Grouping related reports under one Incident ID reduces duplicate handling of the same event, and multilingual communication together