Path planning for mobile robots

Abstract

In many industrial fields, mobile robots are widely used these days. Research on the mobile robot’s path planning is one of the most important aspects of improvements on the mobile robot field. A mobile robot’s path planning involves finding a collision-free trajectory, through the robot’s environment with obstacles, from a specified starting location to a desired destination while meeting certain optimization criteria. Despite much progress in the development of path planning methods in the field of mobile robots, the lack of a versatile path planner being able to handle the uncertainties or changes in the environment remains a significant problem: path planners still trap in local minima or are not able to satisfy optimization criteria when there are unmapped or moving objects in the environment. In this work, we used the YouBot from KUKA as a test platform. YouBot, an omnidirectional mobile robot from KUKA is intended for research and education. This research is begun by few modifications, including a dynamic mutation operator, to the cuckoo optimization algorithm (COA) as MCOA in order to improve the performance of this algorithm aiming at deploying this method for the mobile robot application. A comparative study of the problem of path planning using evolutionary algorithms is presented for a holonomic mobile robot compared to classical methods such as the A∗ algorithm. Gridbased mapping is used effectively to score paths so that collision-free trajectories can be determined from the initial position to the target position. This research takes into account the MCOA and genetic algorithm (GA) evolutionary algorithms as a global planner to discover the shortest safe path. Also, A new non-uniform motion coefficient is introduced for MCOA in order to increase the performance of this algorithm as EMCOA. This new motion coefficient try to make a trade-off between exploitation and exploration search capabilities of the algorithm pursuing for reaching an optimal solution without trapping in the local minimums. To validate the performance of the EMCOA algorithm, some experiments are conducted involving different scenarios of the environment configuration. The present work also demonstrates the application of an approach to detect corrupted information in observed data as outliers or noise. This method is based on the information theory using a statistical ratio to find a threshold for a critical region which states whether outliers are detected or not. Then, a probability-based approach is adopted to eliminate outliers from observed data. To validate the performance of the proposed method, one experiment and a simulation are conducted. The experiment considers a path planning problem in a noisy environment which includes an static obstacle. This test demonstrated that the outlier-removal preprocessing step has effectively removed the outliers from observed data without detecting the obstacle as outliers. To sum up, this thesis modifies, develops and contributes an algorithm and a method to improve the performance of both global and local planners in the robotic applications.

Date	8 Jun 2021
Original language	American English
Awarding Institution	École de technologie supérieure
Supervisor	Vincent Duchaine (Supervisor) & Tony Wong (Co-supervisor)