كيميا خاكباز

The objective of this thesis is to design an‎d implement an efficient algorithm for path planning an‎d path tracking of a wheel–legged robot in the presence of static an‎d dynamic obstacles. In the first stage, the problem of tracking a known reference path using a Nonlinear Model Predictive Controller (NMPC) is investigated in order to prevent collisions with obstacles. In the second stage, global path planning an‎d tracking without a predefined reference path are addressed through the integration of the RRT* an‎d PSO algorithms together with NMPC. In the proposed approach, global path planning is performed by the hybrid RRT*_PSO algorithm, where an initial path is generated in the known workspace using RRT*, an‎d subsequently optimized by PSO with respect to path length, smoothness, an‎d safety margins from obstacles. Thereafter, the nonlinear model predictive controller functions as a local planner an‎d trajectory tracker, continuously predicting an‎d following a new locally optimal path in dynamic environments an‎d under obstacle motion. The use of NMPC enhances tracking accuracy an‎d closed-loop stability in complex dynamic conditions due to its ability to model nonlinear robot behavior, incorporate motion constraints, an‎d predict the future evolution of the system. Simulation results demonstrate that the robot successfully follows a circular obstacle-free path with a mean tracking error of 0.014 m. In the presence of a moving obstacle, the deviation from the reference path increases to 0.08 m to allow obstacle avoidance, after which the robot smoothly returns to the desired trajectory. Furthermore, the hybrid RRT_PSO algorithm significantly reduces computational time compared with using PSO alone. Experimental validation on a real wheel-legged robotic platform shows a tracking error of 0.041 m for a circular obstacle-free path, 0.008 m for a straight obstacle-free path, an‎d 0.042 m for a straight path with an obstacle. Overall, the results confirm that the proposed hybrid framework, combining global RRT*_PSO-based planning with NMPC, provides safe, smooth, an‎d efficient performance in complex two-dimensional an‎d three-dimensional environments.

ربات چرخ‌دار-راه‌رونده , كنترل‌كننده پيش‌بين مدل غيرخطي , عدم برخورد با موانع , برنامه‌ريزي و رديابي مسير , الگوريتم RRT*_PSO

wheel-legged robot , nonlinear model predictive controller (NMPC) , obstacle avoidance , Path planning an‎d tracking , Hybrid RRT*_PSO algorithm

kimia khakbaz

Prof. M. Habibnejad Korayem