Vous êtes ici : GIPSA-lab > Formation > Thèses en cours
SANI Mukhtar

Real-time control of mobile robots using model predictive control and game-theoretic approaches


Directeur de thèse :     Ahmad HABLY

Co-encadrant :     Bogdan ROBU

École doctorale : Electronique, electrotechnique, automatique, traitement du signal (EEATS)

Spécialité : Automatique et productique

Structure de rattachement : Université Grenoble Alpes

Établissement d'origine : Kano University of Science and Technology, Wudil Kano (Nigeria)

Financement(s) : Bourse campus france ; Sans financement


Date d'entrée en thèse : 01/10/2018

Date de soutenance : 04/03/2022


Composition du jury :
HABLY Ahmad (directeur de thèse)
ROBU Bogdan (co-directeur de thèse) M. GEORGES, Didier, Président, Université de Grenoble
Mme. LABBANI-IGBIDA, Ouiddad, Rapporteur, Université de Limoges
M. CASTILLO, Pedro, Rapporteur, CNRS, Université de Technologie de Compiègne
M. STOICAN, Florin, Examinateur, University Politehnica of Bucharest, Romania
M. CHEMORI, Ahmad, Examinateur, CNRS Laboratoire d''Informatique, de Robotique et de Microélectronique de Montpellier.
M. EL-RAFEI, Maher, Invité, Lebanese University, Lebanon.


Résumé : Autonomous control and navigation of mobile robots received a lot of attention due to the ability of robots to carry out sophisticated tasks in a complex environment with a high level of precision and efficiency. The majority of control problems related to mobile robots involved go-to-goal, object tracking, and path following consist a target with pre-defined behavior. As such, the control design does not take into account the future behavior of the target. In surveillance, interception, pursuit-evasion problems, the future behavior of the target must be taken into consideration. These problems where the agent plays against an adversary are best tackled using game theory which provides the best strategy for winning. However, game-theoretic algorithms required a lot of information on the opponent to take into account the optimal strategy of the opponent, which is the worst-case scenario from the perspective of the player. This information requirement often restricts the application of game theory on mobile robots. Also, the majority of the works found in the literature proposed offline solutions applied to holonomic systems. This PhD thesis proposed three different solutions to non-cooperative game problems based on the opponent’s information available to each player. The proposed solutions are online in nature with the ability to incorporate obstacles avoidance. Also, the controllers designed are applied on nonholonomic mobile robots, first in simulation and then validated experimentally. In the first part of the work, the point-stabilization problem in a complex environment was handled using Nonlinear Model Predictive Control(NMPC) with Static and dynamic obstacles avoidance which revolves around the target position. Secondly, the problem was modified to involve a moving target that has a conflicting objective to form a pursuit-evasion game. The problem was solved using NMPC such that only the current states of the opponent are known to each player. Next, a novel game-theoretic algorithms are developed depending on the level of opponent’s information availabe to each player. The methods are compared in terms of capture time, computation time, ability to incorporate obstacle avoidance, and robustness to noise and disturbance. Finally, a variant problem in the context of differential games which lies at the intersection between the point stabilization and pursuit-evasion problem was formulated and solved using game-theoretic model predictive control. Through various simulations and real-time experiments, the proposed agorithms proves to be effective and improved the state-of-the-art.

GIPSA-lab, 11 rue des Mathématiques, Grenoble Campus BP46, F-38402 SAINT MARTIN D'HERES CEDEX - 33 (0)4 76 82 71 31