Catálogo de publicaciones - libros

Specific problems were extracted in an experiment using a practical system in an attempt to transport an object cooperatively with two humanoid robots. The result proved that both body swinging during movement and the shift in the center of gravity, by transporting an object, caused a shift in the position after movement. We investigated the behavior of fundamental motions to make sure the impact of initial positioning on the robot operation. Consequently, it is found that position matching of motors is very difficult even using the same robot and even in the same motion, there occur errors in moving distance and direction. Therefore, we have proposed a learning method to revise a position shift while the cooperative transportation, and established a learning framework in a simulation. In addition, the obtained results were verified by using real robots in a real environment. In order to move towards the target position efficiently, it is necessary to perform the real learning by two robots. Therefore, it is important to discuss the approach for efficient movement and perform experiment with real robots. Since huge time is required for learning in real robots, it is important to reduce the time of learning in real environment using learning data in the simulator. In our future work, we want to study how robots can more to the target in the shortest path when there is an obstacle in the path or how to more in an L-shaped path.

In the previous sections we have seen how the evolutionary computations algorithms were successfully used to evolve many types of controllers for Khepera robot. It was used to evolve neural network synaptic weights in the obstacle avoidance behavior of experiment 1 and the battery recharging behavior of experiment 3. We have also seen how it can evolve the architecture of the neural network along with the synaptic weights as in the experiment of evolving light seeking behavior. Alternatively, it can evolve the learning rules and learning rate necessary for training the neural network synaptic weights. Other types of controllers were successfully evolved too, such as fuzzy logic controllers and computer programs. Many other experiments are conducted using evolutionary computations on different robotic platforms recently. In fact, evolutionary computation is a very promising approach for designing controllers for mobile robots.

In this chapter a methodology for circuit design using Multi-FPGA Systems has been presented. We have used evolutionary computation for all the steps of the process. Firstly, an Hybrid compact genetic algorithm was applied on achieving partitioning and placement for inter-FPGA systems and, for the Intra-FPGA tasks Genetic programming was used. This method can be applied for different boards and solves the whole design flow process.