Online Self-Evolving Fuzzy Controller for Autonomous Mobile Robots Pouria Sadeghi-Tehran School of Computing and Communications Infolab21, Lancaster University United Kingdom, LA1 4WA p.sadeghi-tehran@lancaster.ac.uk Plamen Angelov School of Computing and Communications Infolab21, Lancaster University United Kingdom, LA1 4WA p.angelov@lancaster.ac.uk Abstract - In this paper, an online self-evolving fuzzy controller is proposed for an autonomous leader/follower. The self- evolving controller starts with a simple configuration and learns from its own actions while controlling the mobile robot during the leader following behaviour. A traditional Takagi- Sugeno type fuzzy controller is also implemented and compared with the proposed controller to verify the reliability and performance of the self-evolving controller. Experiments are carried out with a real mobile robot Pioneer 3DX at Lancaster University. Keywords-autonomous mobile robot; leader follower; human-robot interaction; fuzzy controllers; evolving systems I. INTRODUCTION During the last decade, control and coordination of distributed mobile robots have drawn an extensive research attention in control and robotic community [1], [2]. The role of autonomous mobile robots which can interact with other robots and humans has grown significantly in wide variety of applications such as industry, defence, security, etc [4], [5]. Nowadays, many autonomous mobile robots are implemented in different work environments like airports, office buildings, and particularly in hospitals as a courier to deliver materials and suppliers to support the medical staff [1]. In order to create reliable cooperative functions, new approaches must be developed. It is important that the mobile robots operate in such a way that are socially accepted to people and able to interact with them without using specific technical expertise. Having a robot to follow a leader reduces the amount of specific instructions the leader needs to give the robot and, for humans, allows the user to interact with the robot in a natural manner. A dual problem of robot leading a person has also numerous applications (reduce IED casualties, elderly care, rehabilitees, etc.). Following the leader may be easier when the final destination is pre-specified or the robot relies on GPS or pre-loaded maps. On the other hand, designing a controller to work online and robustly can be extremely challenging. In many cases the trajectory and kinematic model of motion is unknown or the leader frequently changes the place or moves in such ways that are unpredictable to the followers [2]. In these cases, the robot should smoothly follow the leader and maintain the desired distance to the leader in every situation. This paper focuses on the ability of autonomous mobile robots to follow a moving object (e.g. human or another robot) on a desired relative distance with respect to the leader. In order to increase the performance of the follower behaviour, an online self-evolving fuzzy controller is implemented which starts with a simple configuration and based on the input/output data collected during the following procedure the structure and parameters of the controller are modified. The remainder of the paper is organised as follows. First, in Section 2 some related works on object following and formation control of autonomous mobile robots is summarised. In Section 3, the outline of the proposed approach and control design for leader following is investigated; also, the architecture of the fuzzy controller is described and two types of fuzzy controllers one with fixed and another one with evolving structure are introduced. Section 4 displays the experiment results and also the comparison between two controllers. Finally, Section 5 provides conclusion with a brief outline of the future directions. II. Related Work Formation control has been investigated in variety of applications for unmanned aerial vehicles (UAVs) [4], aircrafts [5], autonomous underwater vehicles (AUVs) [6], unmanned ground-based vehicles (UGVs), etc. Overall, the formation control of mobile robots can be categories in three main methods [3]: behaviour based approach, virtual structure, and leader following. Most common approach for the formation control is the leader following method [7], [8]. In this method, one of the moving objects which can be another robot or human is selected as the leader. The follower robot needs to position itself to the leader and maintain a distance and heading of the follower directed towards the leader. Compared to the two other methods discussed earlier, the leader following approach is easy to understand and implement. Only the leader‟s motion and desired relative position and bearing between the leader and follower is needed to specify. Once