The rapid advancement of connected and automated vehicles (CAVs) necessitates efficient coordination mechanisms for signal-free intersections in high-density, dynamic environments such as ports and logistics hubs. This paper presents a comprehensive review of cooperative control strategies for CAVs at signal-free intersections, focusing on single- and multi-intersection scenarios. For single intersections, two-stage methods decompose coordination into priority scheduling and trajectory planning, balancing efficiency and computational feasibility, while all-in-one methods synchronously optimize scheduling and trajectories via mixed-integer linear programming (MILP), model predictive control (MPC), and reinforcement learning (RL). Multi-intersection coordination is addressed through re-routing and decomposition approaches, leveraging distributed optimization and rolling-horizon mechanisms to manage coupled traffic flows. Key challenges include scalability limitations of centralized algorithms, communication reliability, and adaptability to heterogeneous traffic and dynamic disturbances. The analysis highlights the superiority of distributed frameworks in scalability and the potential of AI-driven methods like deep reinforcement learning for high-dimensional optimization. Future directions emphasize hierarchical decomposition with edge computing, mixed-traffic coordination, and enhanced robustness against communication uncertainties. By addressing these challenges, signal-free intersection management can evolve from theoretical advancements to scalable industrial solutions, enabling safer and more efficient CAV operations in closed-road networks.