Table 7.

Summary of current efforts on ML applications for O-RAN resource management.

Year	Ref	Contribution
2022	[44]	Demonstrated the performance of on-policy and off-policy DRL methods in the form of PPO and ACER, respectively, and the performance was compared in an O-RAN setup for resource allocation optimization in a real-time surveillance video application.
2022	[45]	Employed a policy gradient-based RL algorithm as an alternative to the initially proposed ILP, to solve an MDP and address the challenge of fairly allocating limited resources to eMBB and URLLC users from multiple O-RUs while providing a significantly less complex solution.
2021	[46]	Proposed a RL-based dynamic resource controller leveraging policy differential semi-gradient SARSA to optimize traffic flow management by effectively and dynamically allocating per-flow resources within an O-RAN platform.
2021	[47]	Designed ML deployed in a Non-RT RIC to adapt the resource allocation policy to environment dynamics while satisfying various energy-driven criteria.