Changman Son


Path finding algorithms, interrelation (probability of success and fuzzy/crisp entropies), fuzzy theory, neural network, machine intelligence, ultrasonic and vision sensors


Two intelligent path finding algorithms with fuzzy and crisp entropies and learning, using a robotic part bringing task as an example, to bring a part from an initial position to a target (assembly hole) for a purpose of a part mating in a partially unknown environment containing obstacles are introduced. Comparison through several criteria and discussion associated with the two path finding algorithms are then presented. In the first algorithm, a path finding is accomplished by cooperating a neural network strategy with a fuzzy optimal control model. The higher a probability an input pattern of a neural network to be identified as a desired output is, the lower a fuzzy entropy is. Through a fuzzy entropy, a degree of identification between an input pattern and a desired output of a neural network can be measured. In the second algorithm, by employing a learning, an uncertainty associated with a path finding task with a sensor fusion technique is reduced. The higher a probability of success related to the path finding task is, the lower a crisp entropy is. Entropy functions, which are useful measures of a variability and an information in terms of uncertainty, are introduced to measure their overall performances of task executions associated with the path findings. A degree of uncertainty associated with the path finding tasks is used as an optimality criterion, e.g., minimum fuzzy or crisp entropy, for a specific task execution. Interrelations between learning and (fuzzy and crisp) entropies are described. Results show interrelations between a probability of success related to a task execution of path finding and (fuzzy and crisp) entropies and also show effectiveness of above methodologies. The algorithms utilize knowledge processing functions. The proposed techniques are not only useful tools to measure a behaviour of learning, but applicable to a wide range of robotic tasks including pick and place operations, manufacturing and motion planning.

Important Links:

Go Back