Real-Time Reinforcement Learning for Vision-Based Robotics Utilizing Local and Remote Computers