Fuzzy System applied to Autonomous Navigation (Short paper)
The mobile robot navigation is one of the main topics of autonomous robotics and has two main tasks: reach the target and avoid obstacles that may be on the way to its target, [1]. Navigation is a result from the combination of three basic skills of mobile robotics [2]: Mapping, localization and path planning. The combination of mapping and path planning results in exploration. Exploitation is the guiding task of the robot during mapping and it allows that the robot coverage the environment through its sensors. For that reason, the approach's selection to exploration can contribute to the robot's navigation process [3]. Furthermore, the robot must deal with uncertainties in its sensors. Those uncertainties received by sensors promote different reactions on the robot, or different behavior [4]. A traditional solution to deal with this navigation's problem is the fuzzy systems [5].
Fuzzy systems are widely used by researchers because they are able to perform inferences under uncertain environments [6][7]. In mobile robots, It is common to combine a reactive approach with fuzzy methods. This approach enables the robot to react instantly by measures received at its sensors. Thus, the robot doesn't need to store all the navigation's planning, saving time and computational effort. However, many sources of uncertainties arise in the operational environment during the robot's navigation. That uncertainty is received from the robot's sensors, which operate in an environment where knowledge is limited by the dynamic elements and environment variables. To deal with approximate information, uncertain or incomplete, the fuzzy logic that is inspired by the ability reason perceived information by humans [8], became a satisfactory approach in many works [9], [10], [11], [7], [12]. Many of those uncertainties influence the performance of the autonomous navigation system. To overcome those uncertainties, some studies [13], [14], [15] has presented the Type-1 Fuzzy Inference System (T1-FIS) as a robust and flexible approachable to handle real time navigation. However, T1-FIS has limited capacity to maintain the uncertainties directly [16]. For using a defined set of inappropriate manner for direct modeling uncertainties, T1-FIS is not able to maintain a good performance for incomplete and uncertainty information. Nevertheless, the Type-2 Fuzzy Inference System (T2-FIS) creates the possibility to model and maintain such uncertainties.
The purpose of this study is to provide a reactive approach to robotic exploration using an occupancy grid map and a fuzzy system in an unknown indoor environment. To elaborate the state of the art of this work, a survey of related works was performed. The work in [6] provides an interval type-2 fuzzy neural method (IT2NN) for obstacle avoidance and stable position of mobile robots with wheels. The authors developed a robot to navigate in two scenarios, with three obstacles each. In results, experimental and simulated, the robot was able to avoid the obstacles and reach a stable position, in relation to its goal. After analysis the velocities, angular and linear, the authors compared the performance of the proposed method with the type-1 fuzzy neural method (T1FNN) measuring the error of the distance between the robot and its goal and the distance traveled by the robot. Thus, the authors concluded that the method proposed using interval type-2 fuzzy resulted in a shorter moved distances, a smoother movement and less control effort for the task, in relation to type-1 fuzzy method. In [17], it's showed an evolutionary Interval Type-2 Fuzzy Inference System(IT2-FIS) e T1-FIS for a cascaded architecture controller that enables a mobile robot to trace a path in the environment. The work highlights, in methodology, the position on y-axis and orientation axis. The paper presents an experiment where the PIONEER 3-DX navigates in an environment and follows a reference trajectory. The experiments highlighted four corners in the environment. The experiments showed that the approach using IT2-FIS resulted in smaller deviations in most corners traveled by the robot. The authors in [18] provided a two wheeled robot (0.12 m x 0.16 m x0. 23 m) in three simulated scenarios. Every scenario, it starts in chose point and navigates the entire scenario until it reaches the stop point. While it doesn't reach the stop point, it must navigate as close as possible to the safety line. The safety line is established from the time when the robot detects the obstacle. In each scenario, the robot performed a navigation behavior with T1-FIS and IT2-FIS. At the end of each simulation the following metrics were analyzed: RMSE (Root Mean Square Error) in relation to the safety line, navigation time and the distance covered by the robot. In their conclusion, the authors claim that the IT2-FIS method for safe navigation in uncertain environments is an attractive approach.
In this work, we propose a IT2-FIS approach in robotic exploration using an occupancy grid map. Here, the exploration algorithm was a result from the combination of Closest-Frontier algorithm, proposed in [19], and the proposed planning algorithm in [20]. This approach is simulated in unknown environments in order to build occupancy grid maps during the exploration. Assuming that the position is known, a wall-following robot is launched in three scenarios for exploration. Simulations are performed by submitting the robot to three fuzzy configurations: close fuzzy configuration (CFC), moderate fuzzy configuration (MFC) and far fuzzy configuration (FFC). Those configurations were designed in order to analyze the performance of T1-FIS and IT2-FIS in the exploration process. The following metrics were studied: time, distance, linear velocity, angular velocity, acceleration, power consumed by the robot and degree of occupancy of cells.
The mobile robot navigation is one of the main topics of autonomous robotics and has two main tasks: reach the target and avoid obstacles that may be on the way to its target, [1]. Navigation is a result from the combination of three basic skills of mobile robotics [2]: Mapping, localization and path planning. The combination of mapping and path planning results in exploration. Exploitation is the guiding task of the robot during mapping and it allows that the robot coverage the environment through its sensors. For that reason, the approach's selection to exploration can contribute to the robot's navigation process [3]. Furthermore, the robot must deal with uncertainties in its sensors. Those uncertainties received by sensors promote different reactions on the robot, or different behavior [4]. A traditional solution to deal with this navigation's problem is the fuzzy systems [5].
Fuzzy systems are widely used by researchers because they are able to perform inferences under uncertain environments [6][7]. In mobile robots, It is common to combine a reactive approach with fuzzy methods. This approach enables the robot to react instantly by measures received at its sensors. Thus, the robot doesn't need to store all the navigation's planning, saving time and computational effort. However, many sources of uncertainties arise in the operational environment during the robot's navigation. That uncertainty is received from the robot's sensors, which operate in an environment where knowledge is limited by the dynamic elements and environment variables. To deal with approximate information, uncertain or incomplete, the fuzzy logic that is inspired by the ability reason perceived information by humans [8], became a satisfactory approach in many works [9], [10], [11], [7], [12]. Many of those uncertainties influence the performance of the autonomous navigation system. To overcome those uncertainties, some studies [13], [14], [15] has presented the Type-1 Fuzzy Inference System (T1-FIS) as a robust and flexible approachable to handle real time navigation. However, T1-FIS has limited capacity to maintain the uncertainties directly [16]. For using a defined set of inappropriate manner for direct modeling uncertainties, T1-FIS is not able to maintain a good performance for incomplete and uncertainty information. Nevertheless, the Type-2 Fuzzy Inference System (T2-FIS) creates the possibility to model and maintain such uncertainties.
The purpose of this study is to provide a reactive approach to robotic exploration using an occupancy grid map and a fuzzy system in an unknown indoor environment. To elaborate the state of the art of this work, a survey of related works was performed. The work in [6] provides an interval type-2 fuzzy neural method (IT2NN) for obstacle avoidance and stable position of mobile robots with wheels. The authors developed a robot to navigate in two scenarios, with three obstacles each. In results, experimental and simulated, the robot was able to avoid the obstacles and reach a stable position, in relation to its goal. After analysis the velocities, angular and linear, the authors compared the performance of the proposed method with the type-1 fuzzy neural method (T1FNN) measuring the error of the distance between the robot and its goal and the distance traveled by the robot. Thus, the authors concluded that the method proposed using interval type-2 fuzzy resulted in a shorter moved distances, a smoother movement and less control effort for the task, in relation to type-1 fuzzy method. In [17], it's showed an evolutionary Interval Type-2 Fuzzy Inference System(IT2-FIS) e T1-FIS for a cascaded architecture controller that enables a mobile robot to trace a path in the environment. The work highlights, in methodology, the position on y-axis and orientation axis. The paper presents an experiment where the PIONEER 3-DX navigates in an environment and follows a reference trajectory. The experiments highlighted four corners in the environment. The experiments showed that the approach using IT2-FIS resulted in smaller deviations in most corners traveled by the robot. The authors in [18] provided a two wheeled robot (0.12 m x 0.16 m x0. 23 m) in three simulated scenarios. Every scenario, it starts in chose point and navigates the entire scenario until it reaches the stop point. While it doesn't reach the stop point, it must navigate as close as possible to the safety line. The safety line is established from the time when the robot detects the obstacle. In each scenario, the robot performed a navigation behavior with T1-FIS and IT2-FIS. At the end of each simulation the following metrics were analyzed: RMSE (Root Mean Square Error) in relation to the safety line, navigation time and the distance covered by the robot. In their conclusion, the authors claim that the IT2-FIS method for safe navigation in uncertain environments is an attractive approach.
In this work, we propose a IT2-FIS approach in robotic exploration using an occupancy grid map. Here, the exploration algorithm was a result from the combination of Closest-Frontier algorithm, proposed in [19], and the proposed planning algorithm in [20]. This approach is simulated in unknown environments in order to build occupancy grid maps during the exploration. Assuming that the position is known, a wall-following robot is launched in three scenarios for exploration. Simulations are performed by submitting the robot to three fuzzy configurations: close fuzzy configuration (CFC), moderate fuzzy configuration (MFC) and far fuzzy configuration (FFC). Those configurations were designed in order to analyze the performance of T1-FIS and IT2-FIS in the exploration process. The following metrics were studied: time, distance, linear velocity, angular velocity, acceleration, power consumed by the robot and degree of occupancy of cells.
Figure 1
Figure 2
