Performance evaluation on the accuracy of the semantic map of an autonomous robot equipped with P2P communication module

Semantic mapping plays an important role in the mobile robotic area, helping robots to perform numerous complicated tasks in both industry and daily life. According to the semantic map, the robot can automatically operate some challenging tasks such as path-planning, place localization, and human-robot interaction. Also, the terminologies of a group of homogenous or heterogeneous robots that perform a general purpose to solve a heavy task with resource limitation become more popular in robot society. To build a multi-robot system, the communication protocol is an essential component to cooperate with a team of robots. Therefore, we propose a multi-robot mapping system, which adopts the P2P communication protocol to solve the mapping problem and map segmentation method from the given mapping or floor plan. The P2P network is constructed based on a centralized overlay network to be scalable and broadcast the control command from the base station to all robots. For generating a semantic map, there have been many works to cope with this task; however, they are still lack of performance. Thus, we perform both feature extraction and feature selection from Voronoi node to enhance the performance of map segmentation. Then to support vector machine (SVM) algorithm and artificial neural network (ANN) are adopted for classifying the segmented areas from Voronoi Graph. The experiment of SVM and ANN shows that they accomplish in map segmentation; however, the SVM algorithm has slightly higher accuracy.