sibly based on input data that the computer is given (such as sensor data). The program consists of assignments (eg, x = 10), loops (repeat an action while or until something is no longer true) and branches (eg, if and else if statements). The computer does exactly what it is programmed to do, no more and no less. One can use terms like “greater than or equal to” or group items into pre-defined sets so that the computer can be pro- grammed to take action based on a range of values, rather than explicitly programming every possible value, but this still means that the computer cannot deal with something it has never seen. With traditional program- ming, the computer cannot generalize and determine that a value not explicitly pro- grammed is “most like” something it under- stands and has been programmed to handle. This approach can be used for portions

of the automated driving task, but not all of it. Imagine trying to use this approach, for example, to program an automated vehicle to locate all signs in the field of view of its cameras, and then how to read the signs. A computer can’t “read” text unless it has been programmed to do so. Even with somewhat standardized fonts used for road signs, the signs have various fonts and colors, and the computer has to deal with day and night and various lighting conditions, as well as par- tially obscured signs, just as human drivers must. And this is just one small component of what’s needed to drive a car. Artificial Intelligence pioneer Arthur Sam-

uel defined machine learning as providing computers with “the ability to learn with- out being explicitly programmed.” Machine learning is not magic. Machine learning pro- grams are built using these same basic build- ing blocks as other programs, but they allow the computer to adjust and select param- eters in order to best meet some target goal, using a set of “training data” and then apply the algorithms, using those optimized parameters to new data that the computer has not previously seen. The desired result is typically either a classification (eg, determin- ing what type of sign has been detected) or a value (eg, the amount of braking required). The approach of particular interest in

vehicle automation is deep neural net- works. Deep neural networks are a form


 Figure 2: Artificial Neural Network (ANN)

“Deep neural networks have many hidden layers and many nodes at each layer. They require an extremely large set of training cases, requiring huge amounts of data and processing time”

of artificial neural networks (ANNs). ANNs have been around for decades, and have gone through several boom and bust cycles in terms of research popularity. ANNs, as shown in Figure 1, have layers of nodes, beginning on the left with an input layer, then one or more hidden layers, and

an output layer. Data on each input vari- able is put into the system, and then the calculation proceeds from left to right, with different weightings on the connections (the arrows shown in the figure), and the result, such as classifying which character is represented in an image, or whether or not a pedestrian is present, is output on the left. During the training phase, the sys- tem is provided with inputs and the correct output, and the weightings on each con- nection are adjusted based on whether or not the system, in its present state, pre- dicted the correct outcome. This weighting adjustment proceeds backwards from the output layer back to the input, and hence is referred to as back propagation. This is how the system is trained and learns. Neural networks have an input node for

Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36  |  Page 37  |  Page 38  |  Page 39  |  Page 40  |  Page 41  |  Page 42  |  Page 43  |  Page 44  |  Page 45  |  Page 46  |  Page 47  |  Page 48  |  Page 49  |  Page 50  |  Page 51  |  Page 52  |  Page 53  |  Page 54  |  Page 55  |  Page 56  |  Page 57  |  Page 58  |  Page 59  |  Page 60  |  Page 61  |  Page 62  |  Page 63  |  Page 64  |  Page 65  |  Page 66  |  Page 67  |  Page 68