Topological based Environmental Reconstruction for Efficient Multi-Level Control of Robot Locomotion

Multi-legged locomotion is generated from multilevel control integration, from lower to higher. It is still become challenging for a robotic developer. This research builds the loco-motion model that integrates embodiment, perception, cognition, and knowledge building. The proposed model considers internal sensory information and external sensory information. It involves a multi-level control to solve the complexity of multi-modal system integration, a neuro-science and ecological psychology approach to developing the proposed system architecture, and a topological approach to enable knowledge building and external sensory processing. This paper focuses on the environmental reconstruction module based on topological based approach. The Topological based approach represents the data flow from sensing to knowledge building. We use dynamic density growing neural gas algorithm as the based of reconstruction module. It implies the dynamic granularity of topological structure of reconstructed environment. The module presents continuous real-time environmental reconstruction building from topological information generated by dynamic density growing neural gas. The reconstructed topological map composes as 3-D map nodes position and normal vector of the node, and their edges. We conducted several experiments showing efficient locomotion behavior could be realized using the proposed model for validating our proposed model.