The objective of a Reinforcement Learning Policy Gradient agent is to maximize the “expected” reward when following a policy

References

Policy Gradients in a Nutshell. Towards Data Science. Link.