Real Time Mobile Robot Route Map Plan

Leow, Wei Qian (2019) Real Time Mobile Robot Route Map Plan. Final Year Project (Bachelor), Tunku Abdul Rahman University College.

Text LeowWeiQian.pdf Restricted to Registered users only Download (3MB)

Abstract

This project dissertation proposes a real time mobile robot route map planning system at indoor environment to localize the mobile robot and then generate path to guide it move toward the target position. Based on researches, most of the existing method produces inconsistent result of localization and mapping for indoor mobile robot path planning hence appropriate method is required. This project contains three objectives. First is to develop a real time path planning system for mobile robot. Secondly, an artificial intelligence technique called faster R-CNN (Region based Convolutional Neural Network) is used for the localization for mobile robot. Thirdly, implementing docking technique into mobile robot. This dissertations documents methodology used on achieving the project goal. In this project uses two cameras to capture the top-view of the environment. Concatenating technique was used to merge two images together. A faster R-CNN object detector was trained and used to localize the position of mobile robot and its heading from the image meanwhile using the template matching method to localize the target slot. A fixed cell decomposition method was used to convert the environment into digital form two-dimensional mapping then use developed path planning algorithm generate path based on the constructed map. The system perform position checking for the mobile robot to ensuring it reach the target slot like docking. The performance of system was determined based on the success rate on detection and navigation using top-view vision with merged image. This dissertation also documents the experimental investigation and discussion collected results such as image merging can get back the actual environment with negligible of error. Besides that, object placing at the merging line might has distortion that lead to loss of feature due to height of object. Furthermore, the developed path planning algorithm are tested able to generate path with minimal path cost and turns required in simple environment for obstacle avoidance. The performance of mobile robot navigation using top-view vision could be improved with position checking in environment with or without obstacle. In conclusion, this project had successfully developed the real time mobile robot route map planning system for indoor environment however still consists some limitations and require improvement in future work.