Guidance on map matching

General talk
1

Hello Folks,

It’s a pleasure to connect with you. I’m Sivasubramaniam, an undergraduate student majoring in AI and Data Science.

I’m leading a team of enthusiastic students participating in the Smart India Hackathon (SIH). We are working on a project to optimize the map-matching algorithm using AI and ML techniques to accurately distinguish vehicular movement on highways and service roads, even in cases where GNSS (Global Navigation Satellite System) data is intermittent or has large biases.

I strongly believe that any of your guidance or mentorship would provide invaluable insights to help steer our project toward success and make a meaningful contribution to solving a real-world problem.

Thank you for considering my request. I would greatly appreciate any suggestions, mentorship, or resources (articles, papers, etc.) you could share.

Looking forward to your response.

2

We are participating in SIH 2024 under ISRO’s (Indian space research organisation) problem, working on a project aimed at developing a map-matching algorithm that leverages AI and ML techniques to accurately distinguish vehicular movement between highways and adjacent service roads.

Problem Statement:

Develop a robust map-matching algorithm using AI and ML techniques to accurately distinguish vehicular movement on a highway versus a parallel service road, The main challenge is ensuring accurate differentiation even with intermittent GNSS signals or significant biases in GNSS coordinates.

Expected Solution:The algorithm will:

  1. Utilize coarse GNSS position data to plot vehicle movement on a digital map.

  2. Apply advanced map-matching techniques to identify whether the vehicle is on the highway or the service road.

  3. Integrate AI/ML models to enhance map-matching accuracy, compensating for GNSS signal loss or bias.

Application: The solution will improve the accuracy of GNSS-based tolling systems, ensuring fair toll collection by correctly identifying vehicles’ positions.

1 Like
3

Map matching is an interesting and quite broad topic, especially when combined with AI.

Some research papers:

“Transformer-based map-matching model with limited labeled data using transfer-learning approach”

“L2MM: Learning to Map Matching with Deep Models for
Low-uality GPS Trajectory Data”

4

Thank you so much for your reply and the detailed information you provided. It’s greatly appreciated! Your insights will be invaluable as we continue to work on this project. We look forward to applying the knowledge and techniques you’ve shared to enhance the accuracy and effectiveness of our map-matching algorithm.

Thanks again for your support!

5

hi I’m sorry if it bothers you but im trying to work on the same project and from the given contact info i sent a mail but i did not receive the required dataset yet, if you have, could you please share it with me ?

6

@ImreSamu Your previous suggestion on the problem statement helped our team move forward with L2MM. However, after reviewing the problem statement multiple times, we now have a different perspective. Specifically, we’re considering framing it as a binary classification problem to predict whether the road type is Highway or Service, using low-quality GPS data. As a result, we’ve divided our team: one group is continuing with L2MM, while the other is exploring alternative approaches. They found a paper titled Map-Matching Error Identification in the Absence of Ground Truth. Your insights on this new approach would greatly assist us in finalizing our approach to the problem statement.

1 Like
7

Hope you get the dataset from ISRO