B.M.S.

College of
Engineering NLP AND GEN AI
23DS6PCNLP

Voice-Enabled AI Agent for Real-Time Price Comparison of Cab

Booking Platforms for Easier, Streamlined Service

Suhas Shenoy (1BM22CD065)

Sahithi Kandula(1BM22CD054)
Deeksha R Naik(1BM22CD021)
Hemanshikaa Jain(1BM22CD029)

Manoj Kumar
Assistant Professor
Department of CSE(Data Science)
1
BMSCE, Bengaluru
 B.M.S. College of
Engineering Table of Contents
1. INTRODUCTION
2. MOTIVATION
3. PROBLEM STATEMENT
4. IMPLEMENTATION
5. RESULTS
6. FUTURE ENHANCEMENTS

Department of Computer Science Engineering(Data Science) 2

 INTRODUCTION
Voice-Enabled Cab Price Comparison System
This project presents an intelligent voice-activated application that compares cab prices between Uber and Ola platforms using real-time
data and multi-agent AI architecture.
Key Features:
● Voice recognition for natural language input
● Real-time price comparison using OpenStreetMap APIs
● Multi-agent system powered by Google's Gemini LLM
● Text-to-speech feedback for hands-free interaction
● Direct booking functionality with confirmation
Technologies Used:
● Python with Tkinter GUI
● Google Speech Recognition API
● CrewAI multi-agent framework
● Google Gemini 2.0 Flash model
● OpenStreetMap services (Nominatim & OSRM)
● Text-to-speech synthesis

3
Department of Computer Science & Engineering(Data Science)
 MOTIVATION
Current Pain Points:
● Users need to manually check multiple apps for price comparison
● Time-consuming process of switching between platforms
● Lack of hands-free interaction while multitasking
● Difficulty in choosing the most cost-effective option quickly
Market Need:
● Growing demand for voice-activated applications
● Increasing adoption of AI assistants in daily tasks
● Need for unified interface across multiple service providers
● Accessibility requirements for users with visual impairments
Value Proposition:
● Save time by comparing prices from multiple platforms simultaneously
● Hands-free operation for improved user convenience e.x. elderly
● Intelligent location extraction from natural speech
● Automated recommendation based on price optimization

4
Department of Computer Science & Engineering(Data Science)
 PROBLEM STATEMENT
To develop an intelligent, voice-enabled transportation booking system that eliminates accessibility
barriers and decision fatigue by automatically comparing real-time cab prices across multiple
platforms (Uber, Ola) through natural language conversation, leveraging AI agents for speech
processing and location extraction, while providing seamless booking capabilities that save users
time and money compared to manually checking individual apps.

5
Department of Computer Science & Engineering(Data Science)
 IMPLEMENTATION
Multi-Agent Architecture & System Design
Agent-Based System: Key Implementation Features:
1. Voice Processing Agent - Extracts pickup/dropoff locations ● Fallback mechanisms for API failures (Haversine distance
from natural speech using Gemini LLM calculation)
2. Price Retrieval Agent - Fetches real-time pricing from ● Caching system for geocoded locations to optimize
Uber and Ola platforms performance
3. Comparison Agent - Analyzes prices and provides ● Rule-based location extraction as backup for LLM failures
cost-optimized recommendations ● Real-time surge pricing simulation with randomization
4. Booking Agent - Handles cab reservation on the selected ● Thread-based voice processing to prevent UI blocking
platform
Technical Stack:
Core Components:
● GUI: Tkinter with scrolled text areas and responsive layout
● MockCabAPI Class: Simulates real cab services with
● AI: Google Gemini 2.0 Flash model via CrewAI
OpenStreetMap integration
● APIs: OpenStreetMap (Nominatim + OSRM) for location
● Location Services: Nominatim geocoding + OSRM routing
services
for accurate distance calculation
● Voice: SpeechRecognition + pyttsx3 libraries
● Speech Interface: Google Speech Recognition + pyttsx3
for voice I/O
● CrewAI Framework: Orchestrates multi-agent
collaboration
6
Department of Computer Science & Engineering(Data Science)
 RESULTS
System Performance & Outcomes
Functional Achievements: System Robustness:
● Successfully processes natural language queries like "Book a cab ● Error handling for network failures and API timeouts
from BMS College to BTM Layout" ● Graceful degradation when LLM services are unavailable
● Accurate location extraction with both AI-powered and rule-based ● Location caching to reduce API calls and improve
fallback methods performance
● Real-time distance calculation using OpenStreetMap routing data ● Thread-safe voice processing to maintain UI responsiveness
● Dynamic price comparison with surge pricing simulation ● Comprehensive logging for debugging and user feedback
● Complete voice-guided booking flow with confirmation
Future Enhancements:
Performance Metrics:
● Integration with real Uber/Ola APIs
● Location geocoding accuracy: High precision using Nominatim API
● Multi-language support for voice recognition
● Distance calculation: Real-world routing via OSRM with Haversine
● Historical price tracking and trend analysis
fallback
● Integration with payment gateways for complete booking
● Response time: Sub-5 seconds for complete price comparison
flow
workflow
● Voice recognition: Integrated Google Speech API with timeout
handling
● Multi-platform support: Simultaneous Uber and Ola price retrieval
7
Department of Computer Science & Engineering(Data Science)
 FUTURE ENHANCEMENTS
1. Integrating other platforms, along with the usage of the Uber and Ola APIs for real time
price retrieval so that travel is cost-effective.
2. Multi-language handling for Indian audience, catering to crowds not well versed in
English.
3. Allowing the choosing of different modes of transport (change in seating capacity, bike,
autorickshaws, etc.)
4. Storage of custom locations like “home”, “work”, etc. and frequently travelled locations for
faster service.

8
Department of Computer Science & Engineering(Data Science)
B.M.S. College of
Engineering

THANK YOU

Department of Computer Science Engineering(Data Science) 9