Bertels, Koen, Cardoso, João M. P., Vassiliadis, Stamatis, Gonzalez-Gomez, J., Gonzalez, I., Gomez-Arribas, F., and Boemo, E.
In this paper, a locomotion algorithm designed for an eight modules worm-like robot has been successfully tested on three different FPGA-embedded processors: MicroBlaze, PowerPC and LEON2. The locomotion of worm-like robots, composed of a chain of equal linked modules, is achieved by means of wave propagation that traverse the body of the worm. The time the robot needs to generate a new motion wave, also known as the gait recalculation time, is the key to achieve an autonomous robot with real-time reactions. Algorithm execution time for four different architectures, as a function of the total number of articulations of the robot, are presented. The results show that a huge improvement of the gait recalculation time can be achieved by using a float point unit. The performance achieved using the LEON2 with FPU is 40 times better than LEON2 without FPU, using only 6% of additional resources. [ABSTRACT FROM AUTHOR]