Real-time side-slip angle measurements using digital image correlation
Introduction
Modern passenger vehicles rely on complex control systems to assist drivers in avoiding accidents or at least to limit damages in the event of a collision. These systems, known as Advanced Driver Assist Systems (ADAS) use features such as Anti-lock Brake Systems (ABS) and Electronic Stability Control schemes (ESC) to enhance the safety of the occupants. Typically, these systems require a large combination of sensors to determine the state of the vehicle. A vehicle model then uses the information to predict the behaviour of the vehicle within normal driving conditions. The predicted model is compared to the actual vehicle behaviour and then corrects any deviation using ADAS (Rajamani, 2005). One parameter that can greatly aid the performance of ADAS is the vehicle side-slip angle, . Vehicle side-slip angle can be used as a measure of the vehicle‘s handling and stability and is, therefore, a valuable parameter in vehicle dynamics. Inagaki et al. (1995) demonstrated that and it’s derivative, , offer better insight to the vehicle‘s lateral stability as compared to yaw-rate. However, due to the difficulty in measuring the side-slip angle most control systems rely on the yaw rate. Chung and Yi (2006) however found that using a stability control scheme that was based on side-slip angle resulted in an overall improved vehicle performance on a virtual test track.
The tyre-road interface is one of the most important research areas in the field of vehicle dynamics, due to all vehicle excitation forces acting at this interface (excluding aerodynamic forces). Many parameters that govern the forces at this interface are required to fully understand the dynamics, one of which is the side-slip angle. Bakker et al. (1987) showed that there is a strong relationship between the lateral force generated by the tyre and the tyre side-slip angle. Most tyre models require accurate side-slip angle measurements as a necessity for the characterisation of tyres, especially during dynamic manoeuvres or when performing tests on tyre test tracks.
Measuring side-slip angle on off-road terrain presents unique challenges such as measuring at low speeds and over rough, uneven terrain where current measuring solutions fail to accurately measure in these conditions.
Side-slip angle is notoriously difficult to measure, however, such sensors do exist. The Kistler Correvit S-HR (Kistler, 2016) is a commercially available side-slip angle sensor, developed for mainly smooth, hard roads where the motion is predominately planar with little body motion. This sensor uses a combination of the Doppler effect and an absolute measuring method for determining the side-slip angle. The sensor is limited to a maximum side-slip angle of ±15 deg, does not perform well below 20 km/h and experiences difficulties when moving over uneven surfaces. Therefore, this sensor is not suitable for terramechanic applications which generally occurs on uneven terrain and low speeds. The sensor has a maximum sampling frequency of 250 Hz. Due to prohibitive costs of such sensors, the field of vehicle dynamics has opted to instead estimate the vehicle side-slip angle rather than directly measure it. These estimation methods use sensor fusion techniques that combine sensors such as accelerometers, GPS, and rate gyroscopes to estimate the side-slip angle. These techniques proved successful and correlated well with the values of simulated vehicle side-slip angle (Bevly et al., 2006, Botha and Els, 2012, Hac and Simpson, 2000). These sensors are inherently noisy and require large sensor excitations, such as experienced high-speed dynamic manoeuvres, to provide accurate results and are also highly prone to integration drift. As a result, these estimation methods are unsuitable for off-road scenarios, where tests occur at low speeds and experience high levels of noise due to terrain roughness. Tyre side-slip angle is not typically measured in field tests, however, are measured during tyre characterisation. The tyres are typically mounted to a rig where the side-slip angle is accurately controlled (Dora et al., 2006).
Botha and Els (2015) proposed an alternative to current methods whereby the side-slip angle could be accurately measured using inexpensive, off-the-shelf cameras and digital image correlation. This technique was developed to overcome the hurdles faced with measuring the side-slip angle in off-road conditions. It measures side-slip angle robustly at low speed and over varied terrain and does not require large dynamic excitation. This technique was successfully tested on smooth concrete, rough cobblestone paving, snow, ice and various other mixed off-road conditions at various speeds and on several different vehicles. Various methods were proposed, either using a single camera or a calibrated stereo-vision rig, comprising of two cameras at a fixed distance apart. The sensor could be mounted on the vehicle for vehicle side-slip angle or on the tyre for tyre side-slip angle. For both cases, the sensors must be facing downwards towards the terrain at all times. The extraction of side-slip angle from the digital images was performed in post-processing from pre-recorded footage due to the large computational expense of image processing. This technique proved to give excellent results that are valuable for vehicle and tyre testing under typical rough, low speed, off-road conditions where terramechanic aspects are important.
This study builds on Botha and Els (2015) by implementing the single camera technique in real-time. The sensor can be used for vehicle and tyre testing but also as a direct input to ADAS systems. The result is a camera-based sensor that uses off-the-shelf cameras and lenses to accurately measure the side-slip angle in real-time and at low cost.
Section snippets
Side-slip angle with digital image correlation
Three algorithms were developed by Botha and Els (2015). Algorithm 1 uses a simple and efficient planar motion algorithm to track the motion of a single camera. Identifiable points on the image, known as features, are tracked across sequential images, meaning the location of that point is found on the corresponding image. Feature tracking will be discussed in Section 3.2. The locations of the features are represented in pixel coordinates, and not in real-world coordinates. The direction that
Computer vision techniques
In this paper an open source computer vision library called OpenCV (OpenCV, 2017) was used and all computer vision techniques were implemented in C++. It was chosen due to the large library of optimised algorithms that allows for fast execution, as required for real-time implementation.
Algorithm
The algorithm starts by obtaining images from a single camera that is pointed downwards towards the terrain. Before features are identified, the distortion is removed from the images. Features are identified and then subsequently tracked using the Lucas-Kanade algorithm. Since the tracked features may have outliers present, the Random SAmple Consensus (RANSAC) (Fischler and Bolles, 1981) algorithm is used. RANSAC is an iterative algorithm that is used to estimate parameters of a model from data
Testing
The side-slip angle measurement was tested using the same test setup as Botha and Els (2015) described in Section 2 and is shown in Fig. 4. The camera was mounted 450 mm from the testing surface, resulting in a field-of-view (FOV) of 480 mm × 350 mm. Ideally, the camera should be mounted on or close to the centre of gravity of the vehicle to minimise the effect of vehicle motion, however, for testing purposes, it was mounted on the rear of the vehicle. The vehicle was driven in a straight line
Results
The performance of the sensor is evaluated by considering the processing time per measurement and the accuracy of the measurement.
Conclusion
From the results obtained it is clear that the real-time implementation of the camera-based side-slip angle sensor was successful. A maximum sampling frequency of 250 Hz was obtained. The mean error across measurements was 0.25 deg with a mean STD of 0.56 deg. Although no direct comparison can be made between the camera based sensor and the commercial sensor, a comparison can be made to previous results obtained that used the same experimental setup. This proved that the camera-based sensor
Acknowledgements
The research leading to these results has received funding from the European Union Horizon 2020 Framework Program, Marie Sklodowska-Curie actions, under grant agreement No. 645736.
References (22)
- et al.
Speeded-Up Robust Features (SURF)
Comput. Vis. Image Underst.
(2008) - et al.
Digital image correlation techniques for measuring tyre-road interface parameters: Part 1 - Side-slip angle measurement on rough terrain
J. Terrramech.
(2015) - et al.
Tyre modelling for use in vehicle dynamics
- et al.
Integrating INS sensors with GPS measurements for continuous estimation of vehicle sideslip, roll, and tire cornering stiffness
IEEE Trans. Intell. Transp. Syst.
(2006) - Botha, T.R., Els, P.S., 2012. Vehicle sideslip estimation using unscented Kalman filter, AHRS and GPS. In: ASME 2012...
Pyramidal Implementation of the Lucas Kanade Feature Tracker
(2000)- et al.
Design and evaluation of side slip angle-based vehicle stability control scheme on a virtual test track
IEEE Trans. Control Syst. Technol.
(2006) - et al.
Design and development of innovative tyre test facilities for measuring tyre characteristics
- et al.
Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography
Commun. ACM
(1981) - et al.
Estimation of vehicle side slip angle and yaw rate
Analysis on vehicle stability in critical cornering using phase-plane method
JSAE Rev.
Cited by (11)
Mechanical behavior of transverse fillet welded joints of high strength steel using digital image correlation techniques
2019, Journal of Constructional Steel ResearchCitation Excerpt :Those traditional measurements have several weaknesses: i) auxiliary devices were complicated, including wheels, bracket or cart [36,37]; ii) the transducers were easily broken with the specimen fracture; iii) measuring accuracy was not enough due to the tiny total deformation; iv) the effective test data was rare since the traditional displacement measurements were contacted to the specimen directly. With the development of technology, digital image correlation system (DIC), a non-contact technique that is independent of the material being tested or the length-scale of interest, was developed to measure the deformation in recent years [8,38–40]. DIC is an optically-based technique used to measure the evolving full-field 2D or 3D coordinates on the surface of a test piece throughout a mechanical test.
Experimental study on the behavior of mismatched butt welded joints of high strength steel
2019, Journal of Constructional Steel ResearchCitation Excerpt :The detailed working mechanism is presented in the user manual [41]. Even though DIC technique measurement has been used in mechanics and industry fields [42–44] in recent years, the accuracy for measuring the deformation of our specimens required verification. A benchmark specimen was tested and the deformation and strain were recorded by traditional measurement and DIC measurement as shown in Fig. 3, and then the comparison of the deformation development recorded by those two methods are displayed in Fig. 4.
Smart System Side Slip Tester Results Accuracy Improvement Using Exponential Filter
2023, Indonesian Journal of Electrical Engineering and InformaticsRegularities of Interaction between a Caterpillar Drive and a Supporting Base
2022, AIP Conference ProceedingsA Geometric approach for estimating sideslip angle for Powered Two- Wheeled Vehicles
2022, 2022 17th International Conference on Control, Automation, Robotics and Vision, ICARCV 2022Tractor-trailer-train braking time sequence detection based on monocular vision
2021, Advances in Mechanical Engineering