Multi-robot path planning using a hybrid dynamic window approach and modified chaotic neural oscillator-based hyperbolic gravitational search algorithm in a complex terrain

The current research aims to enhance the humanoid NAO's ability to plan their overall routes through static and dynamic terrains. The strategy is based on the fusion of the modified hyperbolic gravitational search algorithm and dynamic window approach (DWA) for the navigation of humanoids in various complex terrains. While the updated GSA enhances the fundamental design of basic GSA in terms of exploration and exploitation, the DWA aids in enhancing navigational velocity. The method helps improve overall computational time and, hence, the cost associated with path planning. The modifications to the hyperbolic GSA are carried out by introducing the chaotic neural oscillators to carry out a chaotic search in the immediate surroundings. Three different neural oscillators are chosen and applied to the path planning. The approach helps overcome the cons when the masses/agents show lesser movements in the population. Path planning is carried out in simulation and experimental terrains with and without dynamic obstacles. The proposed MGSA-DWA showed a significant improvement of more than 5% in the path length and time compared with the GSA-DWA. When merged with the Dining Philosophy model, the proposed model effectively avoided dynamic obstacles. The deviation between the simulation and experimental arena was below 6%. The robustness of the proposed model was obtained by comparing it with existing vision-based and other sensor-based approaches. Compared with the vision-based approach, the proposed model effectively traced an optimal path with a 30% improvement in time. The model greatly outperformed other sensor-based methods in a similar scenario.