Overview

This project involves development of an autonomous vehicle by refrofitting the exteroceptive and proprioceptive sensors on an electric vehicle (Kia Soul EV). For the exteroceptive sensors, two FLIR Blackfly cameras, radar along with one Velodyne Lidar have used for perceiving the environment. The precise inertial navigation of the autonomous vehicle is achieved by the integration of proprioceptive sensor in the sensors stack of the autonomous vehicle. For this purpose, the Novatel Global Navigation Satellite System (GNSS) is used which is a single device enclosing Global Positioning System (GPS) and Inertial Navigation System (INS). A CAN-BUS shield is used for reading the data from the CAN bus through OBD (On-board Diagnostic)-II connector. The design and development of control for an autonomous vehicle constitute towards the safe manoeuvring. In this respect, a drivekit which is an electronic controller specifically designed for drive-by-wire cars is used. It provides precise control by making an abstract layer on the top of the vehicle CAN bus, yet not affecting the vehicle own CAN bus. The drivekit provides its own CAN bus called Control-CAN which communicates with the vehicle CAN bus through a CAN Gateway for transferring and receiving the control signals.

The proposed autonomous vehicle is composed of localization, perception, planning and control modules. The design of a distributed system and incorporation of robust algorithms enable the autonomous vehicle to perform efficiently. The fusion of sensor data for localization in map generation and navigation and also in perception module enable reliable object detection, recognition and classification in a dynamic environment. In the planning module, the optimal path is devised by considering the lane, obstacle information, and upon which velocity and behaviour planning are executed. Finally, based on the planning results, the control module performs the lateral and longitudinal control of the autonomous vehicle. Finally, the efficacy of our autonomous vehicle is experimentally verified by deploying it as an automated taxi service in the constrained environment.

Architecture

  • Sensor Suite used for developing the autonomous vehicle
  • Overall architecture of our autonomous vehicle system. It includes sensors, perception, planning and control modules

Demonstration