This thesis presents a new approach to the study of multiple contact dexterous manipulation using qualitative analysis and simulation techniques. By constructing a qualitative model that symbolically abstracts the mathematical model in the domain of manipulation, new techniques for synthesis, analysis and simulation of complex robotic manipulations have been developed. The qualitative techniques facilitate fundamental analysis of the mechanics of manipulation, without the need for precise information about tne physical world. The qualitative model is abstracted from an existing mathematical model using well-defined mappings.

The principle in the development of a qualitative model is to classify the range of all possible values that a stale variable in a mathematical model can assume into a finite, totally ordered set of physically meaningful qualitative states. Since qualitative synthesis is performed using this finite set, complexity of a synthesis problem is greatly reduced. Further, qualitative methods simplify static analysis of multi-body systems by eliminating the frame transformations that are involved in the compulation of the resultant force. Also, qualitative techniques can explain failures and generate on-line strategies for error recovery by reasoning from first principles.

However, the qualitative approaches can provide only coarse grain solutions. On the other hand, the mathematical approaches which are more complex can provide precise, fine grain solutions. By complementing the qualitative techniques with the mathematical techniques it is possible to simplify the process of achieving precise solutions to complex manipulation problems. In particular, it is possible to arrive at an interval (or a neighbourhood) which contains the solution to a synthesis problem using simplified qualitative techniques. If the more detailed mathematical techniques are applied in this local neighbourhood, a precise solution can be generated without getting intractable. Such a co-existence has been recommended, that strikes a balance between the more complex hut precise mathematical and the less complex but imprecise qualitative methods.

The main contributions of this research are the development of a qualitative model of multiple contact manipulation and the associated qualitative analysis and simulation techniques that are needed for applying the model to off-line synthesis as well as real time control. A further contribution is the integration of the proposed qualitative formalisms with the existing quantitative models in an intelligent control architecture. Experimental evidence of the enhanced capabilites of a robot that uses the proposed qualitative techniques have been presented in order to illustrate the scope and promise of applying these techniques in real situations.