Currently, with technological advancements, the presence of robots among humans and their performance in many applications has grown significantly. One of the important practical applications of robots is related to their motion and navigation. Therefore, there is a need for a motion planning algorithm to find feasible, safe, and collision-free paths for robots among humans, considering various factors and the environment. The goal of this motion planning is to efficiently transfer the robot (or agent) from an initial state or origin to a final state or destination, and the primary challenge in this field is to avoid collisions with pedestrians and other obstacles present in the environment. The various methods introduced in this field all share the capability to manage robot motion from the origin to the destination, and their performance eva‎luation criteria encompass metrics such as collision count, stuck, the number of steps taken to reach the destination, and the time agents need to reach their final states. This research examines several significant and high-performance methods previously applied in the domain of motion planning and collision avoidance. Subsequently, an intelligent motion planning approach for robots in densely populated human environments, named "Futuristic-Force based Motion Planning" (Futuristic-FMP), is introduced. This method is adaptable to two-dimensional environments with high population densities. While the application of the algorithm proposed in this thesis is primarily focused on two-dimensional spaces, Futuristic-FMP can also be extended to three-dimensional environments. The movement strategy of this approach is based on predicting the paths of neighboring pedestrians over time steps, and due to its computational nature, it incurs very low processing overhead. In this study, various experiments are designed and conducted to validate the performance of the Futuristic-FMP algorithm and compare it with other methods in order to determine its efficiency. Each of these experiments is deployed to assess a specific functional aspect of the eva‎luated methods. The results indicate that, in the experiments, Futuristic-FMP has achieved a minimum improvement of 4% and a maximum improvement of 12% compared to the FMP method (the baseline algorithm) in terms of algorithm convergence and pathfinding. Furthermore, in terms of the traveled path lengths, Futuristic-FMP demonstrates more optimal and closer-to-ideal performance compared to other algorithms.

سمانه حسيني

شادرخ سماوي , عليرضا بصيري