This document presents a model-free, off-policy actor-critic algorithm to learn policies in continuous action spaces using deep reinforcement learning. The algorithm is based on deterministic policy gradients and extends DQN to continuous action domains by using deep neural networks to approximate the actor and critic. Challenges addressed include ensuring samples are i.i.d. by using a replay buffer, stabilizing learning with a target network, normalizing observations with batch normalization, and exploring efficiently with an Ornstein-Uhlenbeck process. The algorithm is able to learn policies on high-dimensional continuous control tasks.