Development of a motion planning and obstacle avoidance algorithm using adaptive neuro fuzzy inference system for mobile robot navigation

The autonomous navigation of robots is one of the most significant issues about robotics because of its difficulty and dynamism. This is because it relies on environmental situations such as the interface between themselves, individuals or any unexpected changes within the surroundings. It is necessary that the trajectory to the robots’ destination be calculated online, and throughout motion, to enable the robot to respond to variations within the environment. However, the essential difficulty in solving this issue may obstruct a sufficiently quick solution from being calculated online, given sensible calculation resources. These come from high dimensions of the exploration of space, geometrical and kinematic features of the obstacles. Especially their velocities, uncertainty, cost function to be improved, and the robot’s dynamic and kinematic model, This research focuses on the existing drawbacks and inefficiencies of the available path planning approaches within unknown dynamic environments. These drawbacks can be categorized as the problem encountered in this research into four categories, including inability to plan under uncertainty of dynamic environments, non- optimality, failure in crowded complex situations, and predicting the obstacle velocity vector. In this research, a new sensor-based online approach was proposed for generating a collision-free trajectory for differential-drive wheeled mobile robots, which could be applied to an unknown dynamic environment, in which the obstacles are moving and their speed profiles are not pre-identified. This approach depends on future predictive behaviour to predict the obstacles’ future route and priority behaviour to make decisions about the best navigation to reach the destination safely. This approach employs several intelligent techniques to improve the performance of the planner in terms of the quality of the resulted path, runtimes of the planner, ability to solve complex problems effectively and capability of planning in unknown dynamic environments. Firstly, a new sensor-based online approach is planned to reach the first and second objective of the research. This comprises planning in unknown dynamic environments and predicting the obstacle’s velocity vector in order to find safe and fast reactive trajectories. This is particularly true in unforeseen environments that contain both static and dynamic obstacles. After this, the third objective of the research is planning in a crowded complex situation to evaluate the risk of collision between the robot and the obstacle’s trajectory using a fuzzy logic controller. This would allow the FLC to generate a local path for an obstacle avoidance system unique to mobile robot navigation in dynamic environments. Finally, the last objective is to improve the optimality of the new approach using a robust Machine Learning strategy. An adaptive neuro-fuzzy inference system (ANFIS) was designed which constructs and optimizes a fuzzy logic controller using a given dataset of input/output variables in order for the mobile robot to learn. This depends on the previous outcomes to generate a short path with a low runtime for an obstacle avoidance system unique to mobile robot navigation in dynamic environments. The proposed multilayer decision approach successfully guides the robot in uncertain and ever-changing surroundings. It also efficiently predicts the obstacles’ velocity vector. The designed multilayer decision-based fuzzy logic model effectively solves the path planning queries in crowded and complex situations without any failure. Finally, the proposed ANFIS generated FLC successfully improves the optimality and reduces runtime rates of the proposed FLC planner. The present algorithm exhibits attractive features such as high optimality, high stability, low running cost and zero failure rates. The failure rate were zero for all test problems. The average path length for all test environments is 16.51 with standard deviation of 0.49 which gives an average optimality rate of 89.79%. The average runtime is 4.74 (standard deviation is 0.26)