System, Design and Experimental Validation of Autonomous Vehicle in an Unconstrained Environment.

Area of Science:

• Robotics
• Artificial Intelligence
• Automotive Engineering

Background:

• Technological advancements in electric vehicles, sensors, and AI have spurred autonomous vehicle (AV) development.
• Current AV sensor suites can be prohibitively expensive for widespread consumer adoption.
• There is a need for cost-effective AV solutions without compromising safety and efficiency.

Purpose of the Study:

• To develop a full-stack autonomous vehicle system utilizing a minimal sensor configuration.
• To design a modular AV architecture comprising sensor, perception, planning, and control layers.
• To validate the system's performance in diverse, unconstrained environments and as an autonomous taxi service.

Main Methods:

• Integrated exteroceptive and proprioceptive sensors in the sensor layer.
• Implemented environmental perception, including localization and object detection, in the perception layer.
• Developed mission and motion planning algorithms with route information, velocity replanning, and obstacle avoidance in the planning layer.
• Incorporated lateral and longitudinal control mechanisms in the control layer.

Main Results:

• Demonstrated the efficacy of individual modules: localization, object detection, planning, and control.
• Successfully validated the autonomous vehicle's performance in an unconstrained environment.
• Showcased the practical application of the system through autonomous taxi service trials.

Conclusions:

• The proposed full-stack autonomous vehicle system effectively utilizes a limited sensor suite.
• The modular architecture enables robust performance in localization, perception, planning, and control.
• The system's successful validation and taxi service application highlight its potential for enhancing road safety and traffic efficiency.