Obstacle detection for self driving car in Pakistan's perspective

Abstract

This paper reviews methods of obstacle detection for self-driving cars and proposes the better approach with reference to Pakistan's perspective. Pakistan's environment is lot more different from rest of the world where automated vehicles are being tested. Multi-sensor based obstacle detection methods in fusion with laser range finder and vision algorithms proved good for obstacle detection, velocity estimation of dynamic obstacles and estimation of rich semantics of obstacles. These are the key factors for making a robust system to make intelligent decisions for planning and navigating by itself, in an area where obstacles can pop up any time without any prior notice. Vision algorithms are also effective for making environment models with precision. Basically this paper provides a review of different methods used successfully for obstacle detection and concludes which method is more appropriate than others in Pakistan's environment.