Implementation of Integrated CNN Models for Path Planning of Robot in Static Environments

Recently, there have been advances in mobile autonomous robots in fields such as exploration, warehouses and hospitals. Path planning is an important component, as effective path planning is crucial for autonomous navigation in dynamic environments. This study explores the application of convolutional neural networks (CNN) combined with bidirectional long short-term memory (BiLSTM) and gated recurrent unit (GRU) networks for path planning using grid maps. Traditional path planning methods find it challenging to implement dynamic obstacles and complex environments due to their dependence on static algorithms and limited flexibility. The spatial feature extraction capabilities of CNN and the temporal sequence learning strengths of BiLSTM and GRU networks are harnessed by Integrated CNN models. Among these models, CNN-GRU has the highest performance indices. To carry this exercise forward, analysis of the CNN-GRU model's performance with increased grid size and comparison of implementations with other models of deep learning with dynamic obstacles can be undertaken.