Developing and maintaining a collision-free trajectory is one of the most challenging tasks for autonomous robotics in unpredictable and dynamic situations. The current work describes the development of a mechanism for navigating humanoid robots. Sensors are used to determine the positioning of the developed system. The breadth-first search algorithm has yielded positive results in the task of step adjustment to avoid dynamic barriers. Rule-based machine learning provides two driving angles based on the position evaluation of static and moving barriers and the goal. Further, the footstep is optimized based on the JAYA algorithm. It takes the outputs of machine learning as its input and, provides an optimum driving angle to avoid static and dynamic barriers and generates a smooth trajectory. The proposed step planner is integrated into humanoid NAO for autonomous navigation. For avoiding dynamic barriers (same robot), a separate algorithm is required that will solve the conflicts during the navigation. In this paper, the dining philosopher controller is used. For evaluating the robustness of the proposed controller, various types of scenarios has been adapted. These environments are static with a single target, static with multiple targets, environment with dynamic obstacles and multiple robots in a single environment. It is evaluated in simulated environments and verified in experimental environments. The deviation observed is under 5 % that displays a validated relationship between them. Its acceptability is demonstrated based on torque generation at different joints in reference to the default controller of humanoid NAO. Ultimately, the robustness of the proposed step planner is evaluated by comparing it with a conventional path planner in a dynamic environment where it shows superiority.
Kashyap et al. (Fri,) studied this question.