The effort shows how low-cost drones and robotic systems--combined with rapid advances in machine learning--are making it possible to automate whole sectors of low-skill work. Avitas uses drones, wheeled robots, and autonomous underwater vehicles to collect images required for inspection from oil refineries, gas pipelines, coolant towers, and other equipment. Nvidia's system employs deep learning, an approach that involves training a very large simulated neural network to recognize patterns in data, and which has proven especially good for image processing. It is possible, for example, to train a deep neural network to automatically identify faults in a power line by feeding in thousands of previous examples.
The system is trained to automatically learn the internal representations of necessary processing steps, such as detecting useful road features, with only the human steering angle as the training signal. We train the weights of our network to minimize the mean-squared error between the steering command output by the network, and either the command of the human driver or the adjusted steering command for off-center and rotated images (see "Augmentation", later). Figure 5 shows the network architecture, which consists of 9 layers, including a normalization layer, 5 convolutional layers, and 3 fully connected layers. We follow the five convolutional layers with three fully connected layers, leading to a final output control value which is the inverse-turning-radius.
How do you satisfy the "one button" trick to help solve the menu and button configuration dilemma on the vehicle dashboard? One reason I chose to move into the automotive group at SAS is their proven approaches and experience with applying machine learning techniques to help these situations. Driverless vehicles, connected cars, e-hailing, car sharing, and other innovative offerings are reshaping our industry. In the case of getting the dashboard to work intuitively, conveniently and effectively with the driver, machine learning techniques are a wise choice.
The Internet of Things (IoT) and Machine Learning are two of the hottest technologies of our time. At first glance, I really like some of the ideas that are being proposed by the combination of Internet of Things and Machine Learning: smart light bulbs that know when to turn themselves on and off; smart kettles that will make sure you'll have freshly made coffee at the exact time you want it without having thought about it; smart fridges that will do the grocery for you; smart locks that will recognize you and unlock with the tap of a phone. I don't have to worry about leaving the door unlocked because my smart lock will automatically lock the door when it senses that the house is empty. Combined with the power of the fast evolving VR technology, IoT will enable us to travel to distant locations, feel things, meet people and do a lot more without ever setting foot outside our homes.