1 Describe everything aboutthe relationship betweenintellig...

Question

1  Describe everything aboutthe relationship betweenintelligent contro...

Accepted Answer

1 Ultimately the problem of Artificial Intelligence and thus of Neural Nets comes down to that of making a sequence of decisions over time so as to achieve certain goals AI is thus a control problem at least in a trivial sense but also in a deeper sense This view is to be contrasted with AIs traditional view of itself in which the central paradigm is not that of control but of problem solving in the sense of solving a puzzle playing a board game or solving a word problem Areas where the problem solving paradigm does not naturally apply such as robotics and vision have been viewed as outside mainstream AI I think that the control viewpoint is now much more profitable than the problem solving one and that control should be the centerpiece of AI and machine learning research If both AI and more traditional areas of engineering are viewed as approaches to the general problem of control then why do they seem so different In the 1950s and early 1960s these fields were not clearly distinguished Pattern recognition for example was once a central concern of AI and only gradually shifted to become a separate specialized subfield This happened also with various approaches to learning and adaptive control I would characterize the split as having to do with the familiar dilemma of choosing between obtaining clear rigorous results on the one hand and exploring the most interesting powerful systems one can think of on the other AI clearly took the latter more adventurous approach utilizing fully the experimental methodology made possible by digital computers while the more rigorous approach became a natural extension of existing engineering theory based on the pencilandpaper mathematics of theorem and proof See the figure This is not in any way to judge these fields The most striking thing indicated in the figure is not that some work was more rigorous and some more adventurous but the depth of the gulf between work of these two kinds Most AI work makes absolutely no contact with traditional engineering algorithms and vice versa Perhaps this was necessary for each field to establish its own identity but now it is counterproductive The hottest spot in both fields is the one between them The current enormous popularity of neural networks is due at least in part to its seeming to span these twothe applications potential of rigorous engineering approaches and the enhanced capability of AI Intelligent control is also in this position My conclusion then is that there is indeed a very fruitful area that lies more or less between Intelligent Control and Machine learning including connectionist or neural net learning and which therefore presents an excellent opportunity for interdisciplinary research Dynamic Programming and A search These two techniques have long been known to be closely related if not identical Nevertheless the complete relationship remains obscure More importantly many results have been obtained independently for each technique How many of these results carry over to the other field Amazingly such interrelations remain almost completely unexplored at least in the open literature Backpropagation Backpropagation is a connectionist neural net learning technique for learning realvalued nonlinear mappings from examples that is for nonlinear regression see Rumelhart Hinton Williams 1986 Such a function has many possible uses in controlfor learning nonlinear control laws plant dynamics and inverse dynamics The important thing is not backpropagation as a particular algorithmits clearly limited and will probably be replaced in the next few yearsbut the idea of a general structure for learning nonlinear mappings This will remain of relevance to intelligent control TemporalDifference Learning This is a kind of learning specialized for predicting the longterm behavior of time series It was first used in a famous early AI program Samuels checker player Samuel 1959 and since has been used in Genetic Algorithms Holland 1986 and in adaptive control in the role of a learned critic Barto Sutton Anderson 1983 Werbos 1987 The basic idea is to use the change or temporal difference in prediction in place of the error in standard learning processes Consider a sequence of predictions ending in a final outcome perhaps a sequence of predictions about the outcome of a chess game one made after each move followed by the actual outcome A normal learning process would adjust each prediction to look more like the final outcome whereas a temporaldifference learning process would adjust each prediction to look more like the prediction that follows it the actual outcome is taken as a final prediction for this purpose If the classic LMS algorithm is extended in this manner to yield a temporaldifference algorithm then surprisingly the new algorithm both converges to better predictions and is significantly simpler to implement Sutton 1988 The Perfect Model Disease Ron Rivest has coined this term to describe an illness that AI and to a lesser extent control theory has had for many years and is only now beginning to recover from The illness is the assumption and reliance upon having a perfect model of the world In toy domains such as the blocks world puzzle solving and game playing this may be adequate but in general of course it is not Without a perfect model everything becomes much harderor at least much differentand so we has been reluctant to abandon the perfect model assumption The alternative is to accept that our models of the world will always be incomplete inaccurate inconsistent and changing We will need to maintain multiple models at multiple levels of abstraction and granularity and at multiple time scales It is no longer adequate to view imperfections and inconsistencies in our models as transients and to perform steadystate analysis we must learn to work with models in which these imperfections will always be present This means certainty equivalence approaches are not enough and dual control approaches are needed Control without Reference Signals The dogma in control is to assume that some outside agency specifies a desired trajectory for the plant outputs in such a way that controls or control adjustments can be determined For many problems however this is simply not appropriate Consider a chess game The goal is clearly defined but in no sense does one ever have a desired trajectory for the game or the moves to be made Suppose I want a robot to learn to walk bipedally Producing target trajectories for the joint angles and velocities is a large part of the problem a ....

1. Describe everything aboutÂ Â the relationship between intelligent control and AI. 2. Describe everything about the:Basic concept of...

90.2K

Verified Solution

Question

Mechanical Engineering

1. Describe everything aboutÂ Â the relationship betweenintelligent control and AI.
2. Describe everything about the:Basic concept of control systemstability.
3. Write your opinion about the Artificial intelligence in robotcontrol systems.

Answer & Explanation Solved by verified expert

Get Answers to Unlimited Questions

Membership Benefits:

Other questions asked by students