Approach for Detecting Junctions in Roundabout Using Map Data

Abstract

In logical terms, a 2D map is a polymorphic form of linear objects made from practical features like roads, cartos (geometrical representation of a collection of buildings or specific features), roundabouts, highways, and many more. As autonomous vehicle disrupts transportation and navigation, there is a need to detect and identify map features at a granular level with no human intervention. This paper focuses on one of the practical problems, i.e. determining whether a junction is part of roundabouts using the Map domain, which is a novel approach to the best of the knowledge of an author. In a roundabout, the junction is the node having entry or exit of roads in or away from it. Closed loops in roads are extracted by traversing each link and checking for the common junctions using a deterministic algorithm. Loops can be formed not only by roundabouts it can be present at any intersection of roads, from all the universal sets of junctions; identifying the pattern and finding the right junctions of a roundabout is a very challenging and exciting problem. The authors have solved this problem in two parts; data extraction is done by collecting all the roads of America and Europe and then features are extracted, which depend on logical and domain knowledge. Finally, for the fuzzy part of features, a machine model that provides an efficient accuracy of 81% is created as it reduces lots of manual effort for verifying and correcting junctions. This applied methodology is easy to implement.

Keywords:

• Algorithms
• Autonomous vehicle
• Junctions
• Machine learning
• Maps
• Roads
• Roundabout
• Transport

ACKNOWLEDGEMENTS

We would like to thank here Solutions India Pvt. Ltd. for providing data crucial toward the completion of this research.

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the author(s).

Additional information

Notes on contributors

Rakesh Singh

Rakesh Singh is pursuing his PhD in the Department of Computer Science and Engineering from Thapar Institute of Engineering and Technology, Patiala, Punjab, India. He has done his master's from the Indian Institute of Information Technology, Gwalior, in information security. His research interests include machine learning, optimization, computational mathematics and cryptography. Corresponding author Email: [email protected]

Prashant Singh Rana

Prashant Singh Rana completed his PhD from the Indian Institute of Information Technology, Gwalior, and is currently working as an associate professor at Thapar Institute of Engineering and Technology, Patiala. His specializations are machine learning and data mining, modeling and simulation, parallel algorithms, optimization and computational biology. Email: [email protected]

Neeru Jindal

Neeru Jindal received her PhD degree from the Department of Electronics and Communication Engineering, Thapar University, Patiala, Punjab, India. She has been involved in various research activities in the area of image and video processing, machine learning, and deep learning. She is guiding many PG and PhD students. She is a reviewer of many prestigious journals. Currently, she is working as faculty in Department of Electronics and Electrical Engineering, Thapar Institute of Engineering and Technology Patiala, Punjab, India. Email: [email protected]