Proceedings of the International Conference of Control, Dynamic Systems, and Robotics Ottawa, Ontario, Canada, May 15-16 2014 Paper No. 50 50-1 Neuro-adaptive Formation Maintenance and Control of Nonholonomic Mobile Robots Omar Al-Buraiki, Sami El Ferik King Fahd University of Petroleum and Minerals, Systems Engineering Department P. O. Box 1137, Dhahran 31261, Saudi Arabia alburaiki@kfupm.edu.sa; selferik@kfupm.edu.sa Abstract - In this paper, the problem of fleet navigation with formation control and maintenance under leader- follower strategy has been addressed using simultaneous localization and mapping (SLAM)-based navigation unit and artificial potential field approaches. Agents are nonholonomic ground robots. In the proposed approach, the leader or virtual leader is assumed to be the center of the formation and its navigation is guaranteed using SLAM- based controller. On the other hand, potential field control strategy is used to position all followers in a certain formation around their leader. In this work, we represents the case where the dynamic is completely unknown. In such case, an on-line NN-based adaptive model estimator is used to approximate the robot nonlinear dynamic and guarantee the tracking of a desired path. In both cases, the control strategy adds an inner loop to the system’s configuration. Simulation results demonstrate the performance of the proposed approach. Keywords: Leader-follower, fleet formation, localization, potential fields, neural networks. 1. Introduction In recent years, cooperative multi-robot systems attracted the interest of the researchers because of the many advantages they offer compared to a single robot system. For example, multiple-agents robots can estimate their position faster and more accurately due to their ability to exchange information related to their positions, whenever they are in range and the connectivity is guaranteed. Cooperative control has diversified applications in many areas. They can be deployed for transportation, sensing, or military missions. In many situations, the mobile robots have to work cooperatively to accomplish certain tasks or actions. Several design strategies for controlling such a group of mobile robots have been proposed in the literature. The fleet formation problem is addressed in different stages depending on the desired mission. The first stage, agents should reach formation within a finite time (formation control). The second stage consists of maintaining the achieved formation, while the fleet navigates to track a desired trajectory (formation maintenance). Stage 3 aims at ensuring at all times a collision-free mobility of the different agents (collision avoidance). This paper addresses the problem of a collision-free fleet formation control and maintenance under leader-follower strategy for non-holonomic autonomous robot vehicles using simultaneous localization and mapping (SLAM) for the navigation control unit. The proposed framework is composed of two stages and will be built by integrating robot’s group formation based on the artificial potential fields combined with SLAM navigation system. In this work the leader represents a real leader or a virtual leader. The later situation can occur for instance during target tracking of a moving target (see for example Ma and Hovakimyan, 2013). In our approach, the first stage addresses the localization of a mobile agent at each time step when the group’s leader or target follow its trajectory based on the current observation and the last available knowledge about the navigation environment. This stage is achieved using simultaneous localization and mapping (SLAM) on the group leader. Many approaches have been investigated for formation control and maintenance. Artificial potential fields is one of the most used techniques. In the recent decade many control strategies for controlling the formation shape of a fleet of robots were proposed based on the potential fields approach. Song and Kumar (2002) analyzed the force equilibrium to show how potential fields can be used for different