Design, Perception, And Decision-Making In Autonomous Cars: A Systems-Level Review And Implementation

12 Aug
Posted on Author

Authors: Santosh Kumar Dash

Abstract: Abstract- Autonomous vehicles (AVs) promise to transform transportation by improving road-safety, increasing mobility access, and enabling new mobility services. This paper presents a comprehensive systems-level review and practical draft design for contemporary autonomous cars, covering levels of driving automation, sensor suites, perception pipelines, localization and mapping strategies, motion-planning and control architectures, simulation and evaluation approaches, and safety/ethical considerations. We summarize current industry and standards perspectives on automation levels, then present a modular architecture that integrates multi-sensor perception (camera, LiDAR, radar), real-time sensor fusion and object tracking, simultaneous localization and mapping (SLAM), behavior planning (route and tactical), and trajectory generation and low-level control. Important algorithmic choices—classical (A*, RRT, MPC) and learning-based (deep perception networks, reinforcement learning for decision-making)—are compared with their strengths and failure modes. The draft includes recommended software/hardware stacks, data pipelines, validation approaches (closed-track testing and large-scale simulation), and metrics for safety and performance evaluation. It also examines practical failure modes (sensor occlusion, adverse weather, distributional shift), regulatory and ethical constraints, and socio-economic impacts. Where possible the design favors explainable, verifiable methods that enable safety cases supported by reproducible testing. Finally, the paper outlines a staged roadmap for development from Level 2/3 driver-assistance prototypes to Level 4 operational design domains (ODDs). The review and draft aim to be a pragmatic blueprint both for research teams and startups seeking to build safe, testable autonomous driving systems while acknowledging open research challenges and policy needs.

DOI: http://doi.org/10.5281/zenodo.16811475