Abstract
This brief proposes a novel control scheme for tracking a differential drive wheeled mobile robot. A neural network-based tilt integral derivative controller is introduced to track the wheeled mobile robot. Conventional control algorithms fail to satisfy the desired performance in terms of accuracy, total control effort, and maximum control input. By properly training a neural network, optimal values of the parameters of the tilt integral derivative (TID) controller are obtained. The proposed controller is used to track circular, lemniscate, straight line, and B spline trajectories. Performance of the proposed controller is compared with conventional TID and PID controllers using different performance analysis indices. Simulation studies show that the integral squared error and total control effort are small for the proposed controller as compared to the TID and PID controllers.
Similar content being viewed by others
Keywords
- Trajectory tracking
- Mobile robot
- PID controller
- TID controller
- NN-based TID controller
- Kinematics
- Tracking accuracy
1 Introduction
Robots are replacing humans in various fields in order to achieve accuracy, efficiency, and speed as part of modernization. The terrestrial robots are classified as wheeled, stationary, and legged robots. Wheeled mobile robots (WMR) are less complex, cheaper, and have less balance issues as compared with legged robots. Motion control algorithms are required to track the trajectory planned by the guidance systems. Highly robust control algorithms ensure a safe navigation in restricted environments.
In literature, a number of control algorithms have been proposed. In [1], an output feedback controller is proposed using adaptive sliding mode controller and is robust against un-modeled dynamics. But the tracking speed is affected by the measurement noise. In [2], the desired trajectory is exponentially converged with minimum tracking error but model uncertainties, effect of saturation, and actuator failures during motion are not considered. A controller in which the gain of the proportional controller is adaptively tuned is developed in [3]. The tracking accuracy is poor in the presence of noise and also stability is not ensured. An adaptive PID controller [4] is used to increase the flexibility of the control system. Only the tracking accuracy is focused rather than considering the stability of the vehicle body control system. Fuzzy logic is used to design a controller for the safe Robotino navigation [5]. A fuzzy-based PID controller [6] is proposed to track the path of a WMR. Tracking accuracy is improved while compared to the PID controller but the time taken to track the desired trajectory is more. Fractional order calculus and sliding mode controller are combined in a F-AIHMC [7]. High accuracy and high precision motion of the robot are ensured here. A combination of PI controller and neural network controller is used in [8] to control the omni-directional WMRs and the tracking error is reduced as compared to conventional PID. A backstepping and fuzzy sliding mode controller (BFSMC) is proposed in [9]. The desired trajectory is tracked even in the presence of external disturbances and model uncertainties.
In this work, an NN-based TID controller is developed for trajectory tracking of WMRs and the performance is evaluated. PID controllers are widely used in industrial process control owing to its simplicity and easy implementation. TID controllers are popular due to better capability for disturbance rejection, simpler tuning method, and lesser control effort as compared to PID controllers. This paper compares the performance of the proposed controller with these two popular controllers. WMR kinematics is explained in Sect. 2. The proposed controller is described in Sect. 3 followed by the results in Sect. 4 and finally ends up with conclusion in Sect. 5.
2 Kinematic Model of a Robot
A differential drive mobile robot (DDMR) consisting of two driving wheels and one castor wheel is considered. Let \(v_l\) and \(v_r\) be the left and right wheel velocities, respectively. Then the robot velocity is the average of the two wheel velocities and the angular velocity of the DDMR is
where L is the distance between two wheels. From Fig. 1 and from the definition of angular velocity, the kinematic equations of a DDMR can be obtained as
Consider a DDMR with angular velocity \(\omega \) and the linear velocity v, respectively. The point \((x_r(t), y_r(t))\) on the reference trajectory is the desired point. The robot which is currently at (x(t), y(t)) should track this reference point. The error in position and orientation is defined as
3 Proposed NN-Based TID Controller
The proposed methodology uses a neural network (NN)-based TID controller for the effective trajectory tracking of a WMR. The optimal parameters for the TID are generated using a neural network. The proposed structure of the neural network-based non-linear TID is shown in Fig. 2. The objective of trajectory tracking problem is to make the errors in distance and the deviation angle zero. The error in distance and deviation in azimuth angle are obtained from (4) and (5), respectively
The errors in position and orientation are fed back to two TID controllers (one for x and other for y) and to a proportional controller, respectively.
The errors in x- and y-positions are provided as the input signals to the proposed NN controllers. The deviation in azimuth is adjusted to zero by a proportional controller. The tracking errors are described by \(e=\begin{bmatrix} x_e&y_e&\theta _e \end{bmatrix}\). A TID controller whose parameters are optimized using the neural networks is used to track the desired trajectory with minimum control input and tracking error. The TID is similar to that of a PID [10], but the proportional gain is substituted with a tilted mode consisting of a transfer function \(1/s^{1/n}\).
The neural networks trained using backpropagation algorithm are used to find the optimal \(K_i,K_d\) and \(K_t \) of a TID controller. The inputs of the neural networks are the reference trajectory, control input, and current position of the robot. There are 15 neurons in the hidden layer and the output layer consists of three neurons which corresponds to \(K_t, K_i \) and \(K_d\). Inputs to the TID controller are the errors in x- and y-positions given by \(x_e(t) \) and \(y_e(t)\), respectively. The control inputs are the linear velocity v and angular velocity \(\omega \). The accuracy of the proposed NN-based TID controller is evaluated using different performance analysis indices such as integral squared error and maximum of absolute tracking error. The integral squared error is calculated by
The integral of absolute error denotes the accumulated error and is computed as
4 Results and Discussions
The effectiveness and the feasibility of the NN-based TID controller for trajectory tracking of a WMR are evaluated by MATLAB/SIMULINK simulations. This section describes the key findings. The model of the differential WMR described in Sect. 2 and the NN-based controller presented in Sect. 3 are developed using MATLAB 2020a. Neural networks are trained using backpropagation algorithm. Simulations are carried out by considering various trajectories such as circular trajectory, lemniscate trajectory, B spline trajectory, and straight line trajectory as the reference trajectories.
4.1 Case 1: Circular Trajectory
A circle of radius 1m is considered as the reference trajectory. The trajectories are tracked using PID-, TID-, and NN-based TID controllers are shown in Fig. 3a. It is evident from this figure that the NN-based TID controller tracks the desired trajectory exactly whereas the tracked trajectories by other controllers slightly deviate from the desired trajectory. Figure 3b shows the control history of each controllers. From Table 1, it is very clear that the maximum control, total control effort, ISE, and IAE are very small as compared to PID and TID controllers. Thus, for a circular trajectory, the proposed NN-based TID requires lesser control and achieves negligible small displacement error and so is superior to PID and TID controllers.
4.2 Case 2: Lemniscate Trajectory
The trajectories tracked by PID-, TID-, and NN-based TID controllers are shown in Fig. 4a. The displacement errors made by PID and TID controllers are high as compared with the proposed NN-based TID controller. As shown in Table 2, controls required for PID and TID are 3.1 and 2.8, respectively, as compared to a smaller value of 1.12 for the proposed method. In Fig. 4b, the control histories for each of these controllers are shown. It is obvious from Table 2 that the total control effort, ISE, and IAE also are very small for the proposed method.
4.3 Case 3: B Spline Trajectory
A DDMR cannot take sharp turn. The performance of DDMR with a sharp turn can be analyzed with a B spline trajectory. The B spline trajectory is tracked by using the three control algorithms. As can be observed from Fig. 5a, the displacement error with the NN-based TID controller is less as compared to other controllers. The initial part of the trajectory is a straight line and all the three control algorithms are able to track the desired trajectory. However, due to the sharp turn the PID and TID controllers deviate from the desired trajectory as it reaches the goal position. The NN-based TID controller reports a highest tracking accuracy. From Table 3, it can be concluded that the proposed method possesses superior performance in terms of control input and displacement error.
4.4 Case 4: Straight Line Trajectory
This section considers a straight line trajectory. The trajectories plotted by the NN-based TID controller, the PID controller, and the TID controller are shown in Fig. 6a. The tracking error is least for the NN-based TID controller. The error in tracking a straight line is small compared to the error occurred while tracking circular trajectory, B spline trajectory, and lemniscate trajectory. The control history shown in Fig. 6b and the performance comparison as per Table 4 clearly illustrate that the proposed NN-based TID comes with a better performance. The tracking error is reduced by nearly 50\(\%\) by proposed controller when compared to the PID and TID controllers. The maximum control input is only 1.12 in the proposed controller which is very small as compared to TID (2.8) and PID (3.1). Analyzing the performance on simulation, NN-based TID controller is more advantageous in terms of accuracy, control input, and total control effort.
5 Conclusion
In this paper, a neural network-based TID controller for the trajectory tracking of a differential drive mobile robot is presented. A neural network is used to tune the TID parameters. Simulations are done for circular, lemniscate, straight line, and B spline trajectories. The performance of the proposed controller is evaluated with the conventional TID and PID controllers. The simulation results illustrated that the proposed method exhibited superior performances with respect to tracking error, total control effort, and maximum control input.
References
Asif M, Memon AY, Khan MJ (2016) Output feedback control for trajectory tracking of wheeled mobile robot. Intell Autom Soft Comput 22(1):75–87
Zohar I, Ailon A, Rabinovici R (2015) Trajectory tracking and related application robot. Auton Syst 59(6):343–353
Al Khatib EI, Wasim MF (2015) A comparison of adaptive controllers for wheeled mobile robots. In: ISMA
Zhao P, Chen J, Song Y, Tao X, Xu T, Mei T (2016) Design of a control system for an autonomous vehicle based on adaptive-PID. Int J Adv Robot Syst 9(2):44
Al-Dahhan MRH, Ali MM (2016) Path tracking control of a mobile robot using fuzzy logic. In: Proceedings of the 13th international multi-conference on systems, signals and devices (SSD), pp 82–88, Mar 2016
Tiep DK, Lee K, Im D-Y, Kwak B, Ryoo Y-J (2018) Design of fuzzy-PID controller for path tracking of mobile robot with differential drive. Int J Fuzzy Logic Intell Syst 18(3):220–228
Ma L, Sun H, Song J (2020) Fractional-order adaptive integral hierarchical sliding mode control method for high-speed linear motion of spherical robot. IEEE Access PP(99):1–1
Domingos RMM, da Cruz G (2018) Neuroevolved controller for an omnidirectional mobile robot. In: IEEE Latin American conference (LACCI)
Wu X, Jin P, Zou T, Qi Z, Xiao H, Lou P (2019) Backstepping trajectory tracking based on fuzzy sliding mode control for WMR. J Intell Robot Syst 96(1):109–121
Saleh AL, Hussain MA, Klim SM (2018) Optimal trajectory tracking control of WMR using FOPID controller. J Univ Babylon Eng Sci 26(4)
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2022 The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.
About this paper
Cite this paper
Hassan, N., Saleem, A. (2022). Neural Network-Based TID Controller for Wheeled Mobile Robot Trajectory Tracking. In: Yang, XS., Sherratt, S., Dey, N., Joshi, A. (eds) Proceedings of Sixth International Congress on Information and Communication Technology. Lecture Notes in Networks and Systems, vol 235. Springer, Singapore. https://doi.org/10.1007/978-981-16-2377-6_21
Download citation
DOI: https://doi.org/10.1007/978-981-16-2377-6_21
Published:
Publisher Name: Springer, Singapore
Print ISBN: 978-981-16-2376-9
Online ISBN: 978-981-16-2377-6
eBook Packages: Intelligent Technologies and RoboticsIntelligent Technologies and Robotics (R0)