Wheeled mobile robots (WMRs) are very interesting regarding different applications from in-house activities in assisting elderly people and patients to space exploration. While the design concept and the application of the WMRs determine specifications of the robot, the positional errors occur during the WMR motion. The positional errors are inevitable, as they are caused by imperfections in design to fabrication; therefore, there is a need to rectify them using calibration techniques such as odometry, camera-based error detection, or using gyroscope and compasses. This chapter focuses on the use of odometry as it provides improved short-term accuracy with high sampling rates while it is more economical and requires fewer landmarks to localize the WMR. The context provides an overview of WMRs mechanisms, differential and omnidirectional drive, and then introduces an odometry-based method to correct the motion of both types of WMRs. Experimental results on four robots exhibited that positional error was significantly improved. Using analysis of variance (ANOVA) test, the authors could not detect any change in error improvement when the robot changed.
Top1. Introduction
WMRs are employed in many areas such as logistics, inspection and maintenance, security and defense, agriculture, medical care, home care, urban transport, planetary exploration and surveillance operations. Due to the precision required during the performance of a WMR, the robot should properly be calibrated, and the positional error should be improved, before implementation in real field. The accuracy is of importance when complicated mechanisms with tiny structures are proposed. Examples are biologically inspired robots that help scientists to develop intelligent robots to serve in our home, hospital, office, and outdoors (Shi, Habib, Xiao & Hu, 2015; Habib, 2011; Kalvandi & Maddahi, 2012a).
Calibration is defined as a set of operations that establishes the relationship between the values of quantities indicated by a measuring instrument and the corresponding values realized by standards (Balazs, 2008). Developing effective calibration techniques have recently been interested in robotic systems (de Wit, 1998). They include odometry (Y. Maddahi & A. Maddahi, 2004b; Y. Maddahi, A. Maddahi, Sepehri, 2013; Y. Maddahi, Sepehri, A. Maddahi & Abdolmohammadi, 2012a), 3D camera error detection (Ko, Hwang, Jung & Kim, 2005), active beacons (Piaggio, Sgorbissa & Zaccaria, 2001), gyroscope (Bury & Hope, 1995) and magnetic compasses (Rahok & Koichi, 2009). Odometry uses the information of positional sensors attached to each actuator to estimate change in position over time (Maddahi et al, 2013). In WMRs, odometry builds an incremental model of the motion using measurements of the wheel angular displacements. The odometry method is applied to correct errors of different types of WMRs including differential drive and omnidirectional robots (Maddahi et al, 2013; Maddahi et al, 2012a).
The calibration of WMRs using odometry approach has been performed by researchers and engineers. Correcting systematic errors with localization based on magnetic fields has been discussed in Rahok and Koichi (2009). A model-based method was developed to formulate odometry of WMRs (Song & Wang, 2008), while the wheel slippage was considered as a significant factor that affects the accuracy of computations. Nistér et al. (2006) presented a platform to estimate the motion of a stereo head based on video input that operated in real time with low delay. University of Michigan Benchmark (UMBmark) method was introduced to measure odometry errors in differential drive mobile robots (Borenstein & Feng, 1994; Borenstein & Feng, 1995; Borenstein & Feng, 1996). The UMBmark was then employed to reduce positional errors of several WMRs including differential drive and omni-mate robots. An integrated system was also proposed for WMRs that rely on a wireless transceiver infrastructure (Marantos, Koveos, Stergiopoulos, Panousopoulou & Tzes, 2008). Maddahi et al. (2004a; 2004b; 2005, 2006, 2012b, 2013) applied the UMBmark benchmark test on different types of wheeled mobile robots. They also proposed a new technique to reduce both systematic (Maddahi et al., 2004a, 2012a, 2013; Maddahi, Maddahi & Monsef, 2012b) and non-systematic (Maddahi et al., 2013; Maddahi et al., 2012a; Maddahi, Monsef & Mastorakis, 2014) positional errors.