Iterative-learning error compensation for autonomous parking of mobile manipulator in harsh industrial environment
Introduction
Beneficial from the locomotive flexibility and operability, mobile manipulators are generally accepted in wide-area and multi-sites manufacturing scenarios to perform transportation and manipulation [1], [2], [3], [4]. This brings notable potentials for retrofitting and renovating of the typical manufacturing workshops shown in Fig. 1. For practical manufacturing, a mobile manipulator has to navigate among multiple workbenches under specific task-scheduled requirements [5], [6]. It implies that the mobile manipulator needs to achieve autonomous parking to reach each desired site reliably and repeatedly [7], [8], [9]. However, the parking precision will be affected by the inevitable systematic errors due to the mechanical clearance, sensor errors and harsh environment, such as unsatisfactory ground conditions and dynamic obstacles [10], [11]. Consequently, to guarantee successful manufacturing with high accuracy and efficiency, it is crucial to explore an applicable method to eliminate the autonomous parking errors under fluctuating environment as the mobile manipulators go further into various applications.
For the typical manufacturing scenario demonstrated in Fig. 1, the mobile manipulators can be applied for transferring the workpiece between different sites on the pitted and greasy ground under dynamic obstacles. To make the whole workshop to be automated, we have developed a series of commercial mobile manipulators shown as Fig. 2. It consists of a four-wheel-steering and four-wheel-driving (4WS4WD) mobile platform, a manipulator, a vision system, a storage rack, eight motor encoders and related sensing devices, including, lidars, ultrasonic transducers and several anti-collision strips. As a typical modular robotic system [12,13], this robot can flexibly configure the onboard operating platform (e.g., manipulator, roller or belt conveyer and inspection equipment, etc.) as required. Besides, it has some remarkable characteristics, such as automatic charging, multiple collision avoidance, trackless autonomous navigation and vision-based operating the workpiece. With four actively steerable wheels actuated by hub motors and gearboxes, this mobile manipulator has remarkable maneuverability to perform autonomous parking and manipulation among multi-sites. Based on the above advantages, it is applicable for loading, unloading and transferring the workpieces in manufacturing workshops, which greatly improves the manufacturing flexibility and production efficiency.
Comparing to conventional differential mobile platforms running in a favorable environment [14], the difficulties in compensating the systematic parking error of the proposed system lie in: 1) The structure of the 4WS4WD mobile manipulator is more complicated and interconnected, and the coupling features between the moving platform and onboard manipulator make the calculation of the mechanical clearance and precise control hard to achieve; 2) It is cumbersome to calibrate the applied sensors accurately and uniformly; 3) In a harsh industrial environment, the error model is difficult to be built or identified accurately due to the unmodeled dynamics, uneven ground and system uncertainties.
Bearing the above-mentioned challenges in mind, an iterative-learning error compensation method is proposed in this paper to achieve precise autonomous parking of a mobile manipulator. This method simplifies the compensation process and rejects the environmental fluctuations. Without cumbersome sensor calibration and time-consuming modeling, it has three distinguish improvements: 1) With improved Monte-Carlo localization and vision-based technique, a high- precision and low-cost measurement method is designed to obtain the parking errors, which is applicable for wide-area and multi-sites industrial scenarios; 2) Using a small amount of testing data with removed outliers, the offline pre- regulation provides an optimized initial value to enhance the compensation convergence and efficiency; 3) The iterative-learning law constructed by fuzzy logic rules can be adjusted adaptively to avoid oscillation caused by error fluctuation changes and mitigate the environmental influence on the parking accuracy. Implemented in various scenarios, the practicability of our approach has been verified by autonomous parking assessments with an industrial mobile manipulator. It is concluded that the proposed method can effectively reduce the parking errors and accommodate the dynamic environment without violent oscillation.
The remainder of this paper is structured as follows. In Section 2, we provide related works. The error measurement method is given in Section 3, followed by the iterative-learning error compensation method in Section 4. Section 5 and Section 6 offer experimental validations and conclusions, respectively.
Section snippets
Related works
By virtue of easy implementation and robustness, error compensation methods have been recognized as an effective solution for autonomous parking issues of a mobile robot. As one of the primary choices for error compensation, sensor calibration has attracted considerable attention [10,15,16]. Note that it can only reduce the systematic errors caused by inherent characteristics or installation positions, and the calibration of a mobile manipulator with multi-sensors is cumbersome and
Error measurement
We measure the parking errors with the mounted equipment of the mobile manipulator, such as the lidar, eye-in-hand vision system, and the Quick Response (QR) code landmarks on the workbenches. the coordinate system of the measurement system is shown in Fig. 3. The error measurement consists of the following steps: 1) Acquire the absolute mobile nanipulator pose, i.e., target pose; 2) Get the relative pose between the target and actual pose when the robot parks at the site repetitively; 3)
Error compensation methodology
By modifying the reference parking pose of motion control, the parking errors are mitigated in a feedforward manner. Without a precise system model, only partial knowledge of the considered system is required for the compensation implementation. The proposed error compensation process contains two steps, i.e., offline pre-regulation and online iterative-learning compensation. By designing offline pre-regulation, we obtain an accurate initial pose value for iterative-learning and speed up the
Experimental environment
We use a developed mobile manipulator to verify the effectiveness of the proposed method. The considered harsh manufacturing environment is shown in Fig. 8. We carry out the parking experiments continuously in three manufacturing environments, i.e., low dynamic (perturbed by harsh ground with sprayed oil and a 3.5 cm speed bump), medium dynamic (perturbed by harsh ground and several obstacles) and high dynamic environment (perturbed by harsh ground and plentiful obstacles). As shown in Fig. 9
Conclusion
This paper proposes a practical iterative-learning error compensation method for autonomous parking of mobile manipulators, which can effectively eliminate the parking errors in the presence of environmental fluctuations. Compared with traditional methods, our method does not require complex error modeling or cumbersome sensor calibration. To obtain the parking error data in a wide-area and multi-sites industrial scene, a parking error measurement system is developed by using improve MCL and
CRediT authorship contribution statement
Jie Meng: Conceptualization, Methodology, Writing - original draft. Shuting Wang: Methodology, Supervision, Resources, Funding acquisition. Gen Li: Visualization, Investigation, Software. Liquan Jiang: Data curation, Methodology. Xiaolong Zhang: Data curation, Validation. Chao Liu: Visualization, Validation. Yuanlong Xie: Methodology, Writing - review & editing, Supervision, Funding acquisition.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
The authors declare the following financial interests/personal relationships which may be considered as potential competing interests:
Acknowledgments
The work was supported in part by National Key R&D Program of China under Grant no. SQ2020YFB170258, in part by Hubei technical innovation project under Grant no. 2018AAA027, in part by China Postdoctoral Science Foundation under Grant no. 2019M650179, in part by Launch fund of huazhong university of science and technology (02) and in part by Guangdong major science and technology project under Grant no. 2019B090919003.
References (45)
- et al.
Design of cooperative mobile robots for co-manipulation and transportation tasks
Robot. Comput. Integr. Manuf.
(2019) - et al.
An outlook on future assembly systems introducing robotic mobile dual arm workers
Procedia CIRP
(2018) - et al.
Base position optimization for mobile painting robot manipulators with multiple constraints
Robot. Comput. Integr. Manuf.
(2018) - et al.
Online calibration and compensation of total odometer error in an integrated system
Measurement
(2018) - et al.
Model based reconfiguration of flexible production systems
Procedia CIRP
(2019) - et al.
A multi-mode real-time terrain parameter estimation method for wheeled motion control of mobile robots
Mech. Syst. Signal Process.
(2018) - et al.
An algorithm for safe navigation of mobile robots by a sensor network in dynamic cluttered industrial environments
Robot. Comput. Integr. Manuf.
(2018) - et al.
Iterative learning based trajectory generation for machine tool feed drive systems
Robot. Comput. Integr. Manuf.
(2018) - et al.
A sensor-to-segment calibration method for motion capture system based on low cost MIMU
Measurement
(2019) - et al.
A Mobile Robotic Measurement System for Large-scale Complex Components Based on Optical Scanning and Visual Tracking
Robot. Comput. Integr. Manuf.
(2021)
High-accuracy vehicle localization for autonomous warehousing
Robot. Comput. Integr. Manuf.
A novel efficient algorithm for mobile robot localization
Robot. Auton. Syst.
Outliers detection method of multiple measuring points of parameters in power plant units
Appl. Therm. Eng.
A methodology for implementation of mobile robot in adaptive manufacturing environments
J. Intell. Manuf.
Innovative Mobile Manipulator Solution for Modern Flexible Manufacturing Processes
Sensors
Scheduling of smart intra–factory material supply operations using mobile robots
Int. J. Prod. Res.
Deployment of an autonomous mobile manipulator at MBZIRC
J. Field Robot
Automatic Parallel Parking Algorithm for a Carlike Robot using Fuzzy PD+ I Control
Eng. Lett.
Switched modeling and task–priority motion planning of wheeled mobile robots subject to slipping
J. Intell. Robot. Syst.
B-theta*: an efficient online coverage algorithm for autonomous cleaning robots
J. Intell. Robot. Syst.
Cited by (42)
A general mobile manipulator automation framework for flexible tasks in controlled environments
2023, Advanced Engineering InformaticsAn inverse kinematics framework of mobile manipulator based on unique domain constraint
2023, Mechanism and Machine TheoryA precise scan matching based localization method for an autonomously guided vehicle in smart factories
2022, Robotics and Computer-Integrated ManufacturingCitation Excerpt :3D scanner based techniques are also discussed in [15] but this system is not suitable to utilize in the entire factory. A hybrid vision-approach using QR codes and iterative learning approaches for autonomous parking of mobile platforms has been developed by Meng et al. [16] Vision-based systems are not preferred in industrial projects and most AGVs have range sensors. In a factory environment, light conditions may change over time, such as day and night, and optical disturbances may occur due to flashing, smoke, steam and dust.
Fractional robust finite time control of four-wheel-steering mobile robots subject to serious time-varying perturbations
2022, Mechanism and Machine TheoryCitation Excerpt :Different from differential wheels, the mobile robot actuated by Mecanum wheels has the potential in constrained or confined environments. This is because it can achieve an arbitrary effective movement [6,7]. The disadvantage of the Mecanum wheel lies in its poor locomotive efficiency because not all the wheels rotate in the direction of movement [8].
AGV robot for laser-SLAM based method testing in automated container terminal
2023, Industrial RobotScalable Motion Planning and Task-Oriented Coordination Scheme for Mobile Manipulators in Smart Manufacturing
2023, Journal of Intelligent and Robotic Systems: Theory and Applications