Mobile robots are one of the most important research topics in the field of robotic systems. They are still being designed and developed to accomplish service tasks based on the basic structure of wheeled mobile robots. Therefore, they have been applied in many real-life applications, such as industry, entertainment, education, and healthcare [1], as shown in Fig. 1. For example, these applications include indoor cleaning, security patrols, assistance to elderly people, and supplying food and medicaments to COVID-19 patients in hospitals [2,3]. Also, applications of wheeled mobile robots include logistic services of mining and transportation in industrial environments such as ports and warehouses [4]. Recently, autonomous navigation of robots has become essential for the development of autonomous personal vehicles. However, this field of research still has open challenges [5–8].

To perform autonomous navigation, a robot must be equipped with electronic sensors to allow adaptive interaction with its environment. Therefore, the visual navigation of mobile robots is considered one of the important research topics of robotics in the last years, for instance, indoor mobile robot navigation [9,10], crop picking robots [11], and robotic mobility devices [12]. Vision information can be obtained from a monocular camera to achieve tracking and localization tasks with potential obstacle avoidance [13,14]. Hence, real-time image processing algorithms are required to support the positioning, recognition, and tracking of the robot, based on acquired image measurements and segmentation procedures [15]. In this study, a monocular imaging system as an external surveillance camera has been used to guide the wheeled mobile robot for the colored target location. Moreover, automatic control systems present a main module of the robot navigation system to verify the desired robot performance to reach the target location successfully. These control systems apply adequate input velocities to correctly drive the robot towards its desired destination. However, controlling mobile robots still has open problems because of nonlinear and time-variant multivariable driftless system, motion constraints, delayed closed-loop system, and uncertain variability in applied forces and torques to the robot [16,17]. Therefore, this study presents an adaptive control methodology to handle the possible challenges and difficulties of operating wheeled mobile robots accurately and efficiently.

Fig. 8 depicts the experimental setup of this study, as described above in Section 2. The robot operates in a 140 cm × 100 cm rectangular workspace. The size of the workspace is selected such that it is both large enough for the robot (whose dimensions are: 20 cm × 15 cm) to maneuver within and small enough to match the viewing angle of the camera. The colored regions are chosen to be circular shapes to allow being approached from any direction with equal advantage. They are located as far as possible from each other with enough space left for the robot to turn upon reaching the boundaries of the workspace.

The experiments of this study have been conducted in twelve different scenarios of our controlled robot for tracking colored objects. In each scenario, the robot begins its trip from a common starting location and visits three colored regions in a specified order. Visually, the robot is ordered to follow a three-step trajectory in each scenario. For instance, in the first scenario, the robot moves from the starting location (S) to the blue region (B) as a first step, and then from the blue region (B) to the green region (G) as a second step, and finally from the green region (G) to the yellow region (Y) as a third step. A picture of the workspace when the robot is about to start to follow one of the scenarios is presented in Fig. 8.

Each experimental scenario is run once for each controller of interest, namely, classical PID, adaptive fuzzy, and custom PID. The parameter values of all tested controllers are illustrated in Table 4, including orientation and distance control, as depicted above in Fig. 5. All controller gains are fixed during the experiments, but adaptive fuzzy control modifies these gains automatically to achieve accurate performance results.

In each experiment, three performance metrics are measured for each step: the traveling time, the traveling distance, and the final error. The traveling time is measured in seconds and refers to the time taken by the robot to reach the destination of the current stage. The traveling distance (measured in centimeters) refers to the length of the path followed by the robot to reach the destination of the current step. The final error (measured in centimeters) is defined as the distance between the center of the robot and the center of the destination region at the end of the current step. Table 5 illustrates the evaluation results obtained by running the twelve different scenarios that are considered in this study with each of the three controllers. A comparison between the performance of the three controllers is visualized in Fig. 8. On average, the custom controller achieves less traveling-time and distance. However, when the fuzzy controller is engaged, the robot tends to become more accurate and get relatively closer to the center of the target region.

The above practical results demonstrated different successful scenarios of our proposed image-based mobile robot for tracking colored objects. The mobile robot has been controlled using three control methods, which are classical PID, fuzzy control, and custom PID. As shown in Fig. 9, classical PID control achieved the worst results in this study. Therefore, it is not sufficient for the tracking tasks of the mobile robot. In contrast, the proposed fuzzy controller, as given in Fig. 5, achieved the best minimal error among the other two controllers with the longest traveling-distance to the colored targets, as shown in Fig. 9. In addition, our custom PID control achieved the shortest traveling-time and distance path to reach the targets.

In Table 5, three evaluation metrics, namely traveling-time, distance, and final error have been applied to evaluate the control performance of the image-based mobile robot. Twelve scenarios of three colored object combinations have been tested by the controlled image-based robot tracking, as depicted in Fig. 8. Classical PID control could not handle these tracking tasks successfully, resulting in the longest traveling-time of 14.3 s, traveling-distance of 83.2 cm and the final error of 14.3 cm. Nevertheless, adaptive fuzzy PID control (see Fig. 6) achieved the best accurate results with the minimum final error of 13.8 cm to reach the colored targets successfully. The impact of the fuzzy rule base plays an important role in adapting PID control gains (see Tables 3 and 4), achieving the best average errors. However, it requires more time and a traveling distance of 86.2 cm to reach these targets because of cost computations, as illustrated in Table 5. The proposed custom PID is a simpler rule-based controller than fuzzy control. It is efficient to save both time and traveling distance of 6.6 s and 14.3 cm, respectively, but its final error is relatively high (14.1 cm) because of the designed non-linear if-then rules to operate the mobile robot. For both proposed intelligent controllers, the designed mobile robot tracker achieved better performance in tracking RGB-colored objects than classical PID control because of their adaptive features to handle the targeted robot tasks like experienced human operators instead of traditional mathematical approaches, as given in Table 5 and Fig. 9.

The control performance of the proposed image-based mobile robot can be enhanced by using other fuzzy control schemes, such as Type-2 and intuitionistic fuzzy approaches [37,38]. Also, integrated bio-inspired optimization techniques with robot controllers have been proposed to estimate the optimum path [39]. Additionally, optimal extremum machine learning (ELM) with a firefly optimization algorithm can be applied to overcome the decoupling problem between the camera coordinates and the image moment features for a vision-based robotic system [40]. However, these advanced approaches require more cost computations by extending the hardware resources including the memory size and processors for practical applications. Therefore, the proposed fuzzy and custom PID controllers are still valid to achieve good performance of image-based mobile robot tracking, as presented in Fig. 9.

In this paper, we presented a new intelligent mobile robot to accomplish multi-tasks by tracking objects in a real experimental field. This development was achieved by using real-time image processing, fuzzy-tuning and custom PID controllers. The proposed approach was used to recognize and segment a colored object in an image. The adaptive fuzzy logic achieved an accurate tracking result as a practical intelligent controller. In general, our developed multi-purpose robotic system’s promising results can be employed in various applications to reduce human workloads in different environments, such as delivering medical tablets for patients with infectious diseases [41].

For future work, this research can be extended to track different colored shapes with changing the floor color [42], such as industrial robots for painting applications or tracking balls and trajectory prediction using sports robots [43,44]. Also, we aim to improve the performance of developed image-based robot controllers by implying various lighting disturbances of reflection and shadows. Adding obstacles and moving colored targets will also be considered in the prospects of this study.