Since I am a second-year undergraduate student, I need to undergo a summer internship. For this purpose my scientific adviser proposed the theme “Applications of neural networks to autonomous navigation systems”.
Autonomous (inertial) navigation systems use the coordinates of the initial location to determine the current position. They do not require external reference points and are more noise-resistant, in comparison with non-autonomous ones. Nevertheless, since the sensors used by these systems have imperfect characteristics, measurement errors occur that lead to a decrease in the accuracy of the location. Within the framework of this project, the possibilities and limits of the applicability of neural networks for processing information from gyroscopes, accelerometers, magnetometers, barometers and temperature sensors, of the middle class of accuracy in autonomous navigation applications will be considered. Figure 1 shows a block diagram of the hardware of the navigation system. The Raspberry Pi 3 Model B platform is chosen because of its relatively small gabarions and power consumption, at a sufficiently high computing power. If successful results will be obtained, the Raspberry Pi 3 Model B can be replaced by the Raspberry Pi Compute Module 3, which has smaller dimensions.
Figure 1 – block diagram of system hardware