Adaptive Self-Localization System for Low-Cost Autonomous Robot

Due to the massive growth in autonomous vehicles, mobile robots applications are more prevalent today. To implement intelligent behaviors, the robot must have the ability to locate itself and adapt to different environments. Despite the recent developments in self-localization, long-term navigation with low-cost robot is still an active area of research. This paper develops a new self-localization system based on Neural Network (NN) method that is fused into a fuzzy logic navigation system using low-cost encoders. The proposed system allows the autonomous mobile robot to adapt itself to different environments and improve its localization based on the trained model. In the experiment, the system is tested with PowerBot robot in different real environments, and compared with one of the most well-known self-localization method (i.e., dead-reckoning). The test is conducted in different set-up to confirm that the proposed system significantly improved the accuracy without the need for additional sensors other than the encoders. It was able to adapt to different environment and accumulatively improved the results.