Modeling and Velocity Control of A-Shape Microrobot With Adaptive Neural Network Controller

Design and control of micro robots have been one of the interesting fields in robotics in recent years. One class of these micro robots is the legged robots. Various designs of legged robots have been proposed in the literature. All designs rely on friction for locomotion. In this paper dynamic model of a planar two-legged micro robot is presented using Luger friction model, and an adaptive neural controller used to control the robot to improve robustness and velocity of the robot. As mentioned earlier, friction plays an important role in locomotion of the legged robots. However, especially in legged micro robots, it is difficult to model the frictional force correctly since environmental disturbances like dust and changes in shape of the test bed can significantly alter its value. Therefore, one needs to design a controller that adapt to new condition and had enough robustness so one chooses neural network controller. Result show that with updating weight of neural network robot could follow desired trajectory, and with change in friction coefficient training time was low enough to update weight at each step.