GENERALIZED FRAMEWORK FOR THE NAVIGATION OF MULTI-VEHICLE SYSTEM BASED ON NONLINEAR MODEL PREDICTIVE CONTROL

GENERALIZED FRAMEWORK FOR THE NAVIGATION OF MULTI-VEHICLE SYSTEM BASED ON NONLINEAR MODEL PREDICTIVE CONTROL. PhD thesis, King Fahd University of Petroleum and Minerals.

[img]
Preview
PDF (2014- PhD Thesis - GENERALIZED FRAMEWORK FOR THE NAVIGATION OF MULTI-VEHICLE SYSTEM BASED ON NONLINEAR MODEL PREDICTIVE CONTROL)
Bilal_PhD_Thesis2014.pdf - Accepted Version
Available under License Creative Commons Attribution Non-commercial No Derivatives.

Download (3MB) | Preview

Arabic Abstract

نقوم في ھذا البحث العلمي بمعالجة عملیة تشكیل تركیبات ذات خاصیة القائد والأتباع للتحكم في المركبات ذاتیة الحركة المقیدة والتي تعمل في بیئة ذات عرض نطاق ترددي بالاضافة الى وجود تأخیر وبطء في الاتصالات. نقترح تصمیم نموذج تحكم تنبؤي غیر خطي، بحیث لا تحتاج فیھ الروبوتات إلى تقدیر حركة الروبوتات المجاورة مع ضمان تجنب الاصطدام. لضمان تخفیض حجم حزمة البیانات المرسلة، یتم ضغط البیانات باستخدام الشبكات العصبیة. وعلاوة على ذلك، فإن اقتراحنا یتمیز بتكیف الروبوتات مع قراءات لأجھزة الاستشعار بمعدلات مختلفة بالاضافة إلى التكیف مع دینامیات مختلفة وعوامل تنبؤ وتقیید مختلفة، مع كونھا ذات قدرة على تحمل بطء في الاتصالات وعدم الیقین من مسارات الروبوتات المجاورة. نقترح في ھذا البحث وسیلة مبتكرة في تحقیق تفادي الاصطدام. یتم عرض النتائج التحلیلیة التي تثبت رسوخ واستقرار نموذج التحكم المقترح أثناء تجربتھا في أنماط شبكیة مختلفة. توضح نتائج المحاكاة فعالیة نموذج التحكم المقترح.

English Abstract

Recent advances in MEMS-based sensors, low consumption actuators, as well as affordable and high performance computing and communication equipment has allowed mobile robots to advance rapidly towards development of multi-agent systems. Control system of the robot, which consists of the sensors to quantify measurable variables affecting it, the software which takes this information to dictate the actuators to achieve prescribed goals. For multi-agent systems, the key ingredient is communication among the agents to coordinate decisions and control actions. Coordination efficiency is dictated by communication bandwidth and reliability, as well as computational power available. This thesis addresses the formation control of teams of mobile robot - or multi-agent system of autonomousvehicles - by providing a rigorous generalized framework for distributed model predictive control of constrained nonlinear systems. We address leader-follower formation control of constrained autonomous vehicles operating in an environment where communication bandwidth is limited and transmission delays are present, along with other sources of uncertainty and disturbances. A number of sources of uncertainty are taken into account to provide robustness to the algorithms developed. In existing literature, usually only measurement / estimation errors or model mismatch are taken into account. We consider the simultaneous presence of six sources of uncertainty consisting of errors in estimation, modeling, prediction, data compression and loss of information due to delay. We provide detailed feasibility and stability analysis to derive closed form analytic expressions relating the growth of uncertainty along the prediction horizon, and its effect on recursive feasibility and robust stability. Nearly ten new algorithms are developed in this thesis for designing distributed robust NMPC controllers for multi-agent vehicle control based on a very general theoretical framework providing key insights in choosing design parameters for control design. The proposed algorithms can be divided into two main categories: offline and online algorithms. The offline algorithms are computationally intensive, but since they are executed offline, this is not a major concern. The online algorithms are fast processing and provide update to the receding horizon control strategy. We provide robustness by finding upper bounds on uncertainty growth and hence restricting the admissible states to tighter constraints. Recursive feasibility is shown to depend on controllability characteristics of system dynamics, which restricts the maximum allowable uncertainty growth. Our approach is dual-mode NMPC, where stability is ensured by suitable selection of terminal weighting factor, terminal constraint set and a linear terminal control law. We provide a method of maximizing this terminal constraint set, which is a measure of stability. Similarly, output feasible set of NMPC algorithm is determined with proposed min-max optimization technique. We also propose a method for data compression and trajectory tail estimation. We propose a practically stable (ultimately bounded) formulation of the distributed nonlinear model predictive controller (DNMPC), in which agents communicate compressed information to each other with propagation delays and collision avoidance is guaranteed, despite the presence of these delays and uncertainties. Data compression using neural networks approach is used ensuring a considerable reduction of the data packet size (as much as 75 %). Moreover, the approach allows the agents to be sampled locally at different rates as well as to have different dynamics, constraints and prediction horizons, while being robust to uncertainties and propagation delays. Collision avoidance is achieved by means of a novel spatial filter-based potential field. Analytical results proving Input to State Practical Stability (ISpS) and generalized small gain conditions are presented for both strongly connected and weakly connected networks. Extended analytical and simulation based examples are provided to show the efficacy of proposed algorithms.

Item Type: Thesis (PhD)
Subjects: Systems
Department: College of Computing and Mathematics > lndustrial and Systems Engineering
Committee Advisor: El-Ferik, Sami
Committee Members: Al-Sunni, Fouad and Zerguine, Azzedine and El-Shafei, Mustafa and Hawwa, Muhammad
Depositing User: SIDDIQUI B AHMED (g200405080)
Date Deposited: 12 Apr 2015 07:28
Last Modified: 01 Nov 2019 15:44
URI: http://eprints.kfupm.edu.sa/id/eprint/139439