Patrick McDowell
Download 0.83 Mb.
|
| Figure 10. The m0.vi servo control screen. The individual servos are controlled by the sliders on the left hand side of the screen. The record and play functions occupy the lower part of the screen, while the current and next servo positions are shown at the upper right of the screen.
Using this program and the previously discussed method, Mickey took about three steps, but because of its mechanical configuration, it was difficult to control. The robot was barely statically stable and once it started moving the dynamics became unstable, so without some sort of good feedback loops it was evident that it was going to take a stroke luck to get good results. The torque problem also made it so that as the cat walked, it was difficult if not impossible to get it back to a standing position. The legs would simply get more and more sprawled because the servos lacked the power to raise the cat back up. Realizing that it was going to take more than just software changes to make the cat walk, a host of mechanical changes were made (discussed in the previous section) and the second prototype, Stubby, was born. Using the m0.vi program, Stubby showed much greater prowess, but “eyeballing” a walking sequence was still proving to be difficult. Eventually a nine frame sequence was developed with the intention to tune it with a genetic algorithm. The first frame was the standing position and next 8 frames held the servo positions for a simple walking sequence. The grow.vi program was created to hone the sequence over a number of generations into a viable walking gait. The program always left the stand frame alone, and assigned each of the frames in the walking sequence to a chromosome. It then created a population using mutation based on random numbers, maximum offsets, and chance of mutation. The individuals of the population were each run in the robot and the distance that it moved was used as a fitness function. After each member of the population had its chance, the individuals that walked the furthest were used to create a new population. Because this was not a simulation, running thousands of generations with large populations was not feasible, so the mutation factor was turned up. If a population had a worse average walking distance than the previous generation, the algorithm was halted and restarted using the best individual from the previous generation as the seed gait. By repeating this process the Stubby increased its walking distance in 17 frames from roughly 14 inches to about 25 inches in about 40 generations. At the end of the process Stubby walked, but it walked ugly. Ugly as in the servo movements were jittery, and it drug its feet. Figure 11 shows the control screen for the grow.vi program. 
Figure 11. The grow.vi program. The inputs in the lower left control the population size, and mutation. This program improved the walking gait of Stubby significantly. Directory: Academics -> Faculty -> pmcdowell Faculty -> Chapter 1 “Preliminaries” Reasons for Studying Concepts of Programming Languages Faculty -> Chapter 7 “Expressions and Assignment Statements” Faculty -> Deleuze and Analytic Philosophy by Faculty -> Curriculum Proposal Title of Proposal: Course Status Change Faculty -> Chapter 6: Arrays Faculty -> Chapter 5 Introduction pmcdowell -> Guide to Intelligent Systems; Second Edition; Michael Negnevitsky Faculty -> Carmen Sara Cañete Quesada Faculty -> Economic Conditions for Regional Development and for First Nations Download 0.83 Mb. Share with your friends:
|