Online Topological Mapping for Embedded Computers with Growing Neural Gas

This paper presents an online topological mapping method for embedded computers with growing neural gas. Currently, perceiving the real-world in 3D space has become increasingly important. In the context of robotics, understanding the real-world in 3D space enables the robots’ potential to perform intelligently and autonomously. However, a problem occurs when a large amount of data is generated, especially for embedded computers that are usually installed on mobile platforms. Dense maps are simply unpractical, and sparse maps are not suitable due to a lack of appropriate information. Topological maps, on the other hand, offer a balance between computational cost and accuracy. One of the most well-known unsupervised learning methods for topological mapping is growing neural gas (GNG). Unfortunately, it is difficult to increase the learning speed due to its traditional iterative method, making it challenging for real-time applications. Therefore, we propose a new method for topological mapping, called simplified multi-scale batch-learning GNG (SMB-GNG), by applying a mini-batch strategy in the learning process. The proposed method has been applied to a Raspberry Pi, one of the most popular embedded computers in robotics. It shows the capability for online topological mapping by maintaining its performance above 30 fps.