Researchers Use Reinforcement Learning to Teach Autonomous Underwater Vehicles to Navigate Without GPS
• Researchers developed machine learning models to help autonomous underwater vehicles (UUVs) navigate without GPS signals underwater.
• They used deep reinforcement learning to teach the UUVs to navigate more accurately under difficult conditions.
• The models learn by taking random actions, observing results, and reinforcing positive actions while avoiding negative ones.
• The researchers tweaked the training process to mimic human learning by weighting recent, successful experiences more heavily.
• The adapted training technique allowed the models to train faster and consume less power, saving time and money when deployed.