|Abstract:||Cooperative Intelligent Transport Systems (C-ITS) aim to deliver innovative services and systems relevant to different modes of transport and traffic management, that will enable savings in journey times, greater safety, and reduced traffic congestion and air pollution. Some examples of advanced C-ITS applications include autonomous (self-driving) cars, automated guideway transit, automated trains, and management of transport during emergencies. A popular example of C-ITS is the advanced driver assistance systems (ADAS) with features such as lane change assist and automatic braking, which have experienced a rapid growth over the past few years. Precise vehicle positioning is a critical capability for C-ITS and ADAS operations. To ensure reliable and safer positioning, the system needs to have full integrity monitoring (IM) capability and to guarantee real-time positioning continuity and accuracy. In IM, the vehicles’ on-board computer system automatically performs checks to detect faults, isolates faulty information or anomalous measurements, and raises an alarm in the event of the positioning system operation being considered unsafe for use. To date, most research on integrity monitoring focuses on aviation, and very little on C-ITS applications. The two though have significant differences in terms of their operational environment, application-dependent threats and vulnerabilities, and positioning accuracy and reliability requirements, which all have to be considered when modelling and designing the IM approach. In this contribution we present a new integrity monitoring approach for a low-cost automotive positioning system suitable for ADAS applications. The system integrates Global Navigation Satellite Systems (GNSS) Real-Time Kinematic (RTK) system, low-cost Micro-electro-mechanical systems (MEMS) IMU and speed sensors (SS). The system is capable of maintaining positioning during periods of GNSS signal blockage, such as in urban environments or in tunnels. Doppler measurements are also used to aid positioning and for correcting direction of the vehicle derived from the IMU. Positioning is performed in the following priority of modes depending on surrounding environment and work conditions: i) RTK; ii) integration of MEMS IMU and speed sensor data, correcting for IMU heading by GNSS Doppler-derived heading when available. The objective is to assure positioning continuity by providing alternate solutions such that when one positioning mode is not available another mode can be activated. A brief overview of the system is given. The main focus of this work is on presenting a new integrity monitoring approach for the integrated RTK/IMU/Speed-sensor system for ADAS applications. A threat model of the integrated system is firstly discussed, which includes both the nominal performance and possible fault modes. Next, the fault detection and exclusion step (FDE) of the IM is presented, where the uniformly most powerful invariant (UMPI) test is used for detection of faults. When faults are detected, faulty observations are identified. We apply a bundle of tests for verification and to maximize the success rate of the correct identification of faults. We first apply the w test, next the Chi-square statistic is computed for the subsets that exclude each suspected candidate faulty observation(s) and the UMPI test is reapplied. The possibility of a single and multiple-faults are considered. Moreover, a procedure is applied to avoid the masking effect when identifying faulty observations. Finally, a confirmation test is applied where the difference between the solution with and without the excluded faulty observations should be constrained by a certain threshold derived from the expected uncertainty information of the differenced solutions, computed independent from the observations, and scaled by a factor computed based on the probability of false alert. We next discuss computations of the horizontal and vertical protection levels (HPL and VPL) for the proposed positioning system. These PLs are bounds to the error in the estimation of the position solution up to a given confidence level. Prior studies in IM for precise positioning built the PL models taking into consideration only a limited fault-free case and assuming that all faults are perfectly isolated. To improve the assurance level, we introduce a new model that considers both the fault-free case and the likelihood of undetected faults. Both accuracy and continuity requirements as well as the target miss-detection probability of faults are considered. For the observations that pass the FDE stage, a position error bound is created for each possible fault mode that might be miss-detected. This is done by computing a position solution unaffected by the fault, computing an error bound around this solution and accounting for the difference between the all-observations position solution and the fault tolerant position. In building the HPL, existing methods only consider standard deviations along the Easting (E) and Northing (N) directions with a uniform confidence level. However, in practice, the maximum error can be in a direction different from E and N. Therefore, we introduce PL based on a confidence error ellipsoid, where the semi-major axis is aligned with the direction of the maximum error. We propose two PL metrics for horizontal positioning. The first PL bounds the maximum horizontal positioning error component, and a second PL representing an overall HPL. When modelling the effect of biases in the PL, the bias value in RTK can practically be assumed zero; thus, it is not included in the computation of the protection levels. However, when IMU measurements are used or when GNSS Doppler-based velocities are employed, the impact of a possible heading bias of the IMU or the GNSS accumulated velocity biases have to be overbounded and thus included in the PL. These biases were modeled as a function of the time intervals between measurements. For evaluation of the proposed approach, a kinematic test was conducted in an urban area in Japan with a focus on horizontal positioning. The integrated system comprises a Trimble SPS855 receiver, an Analog Devices Co. MEMS (IMU) and a speed-sensor, all mounted on a moving vehicle. GNSS provided a sampling rate at 10 Hz, the IMU and speed sensor data were obtained at 20 Hz and 100 Hz respectively. The RTK was performed using a reference Trimble NetR9 receiver. To validate meeting the integrity monitoring requirements the position error was computed and we check the percentage of bounding this error by the HPL. To compute the position error, the RTK results were referenced to post-processing positioning. When RTK was not available, the IMU/speed sensor positioning results were referenced to a POS-LV system (manufactured by Applanix Inc.) installed in the vehicle, which has 20-30 cm positioning accuracy. Challenges of the used system were discussed, such as controlling the effect of the heading error of the IMU and calibration of the system. Test results show that by integrating the RTK with IMU/speed sensor, we had positioning availability of 100%. The integrity monitoring availability was assessed and found to meet a target value of 99%. The HPL bounds the position error during the whole test period indicating the effectiveness of the proposed approach. Future work includes conducting more tests under various work conditions for a comprehensive validation of the proposed method.|
Proceedings of the 29th International Technical Meeting of The Satellite Division of the Institute of Navigation (ION GNSS+ 2016)
September 12 - 16, 2016
Oregon Convention Center
|Pages:||2733 - 2753|
|Cite this article:||
El-Mowafy, Ahmed, Kubo, Nobuaki, "Integrity Monitoring for Advanced Driver Assistance Systems," Proceedings of the 29th International Technical Meeting of The Satellite Division of the Institute of Navigation (ION GNSS+ 2016), Portland, Oregon, September 2016, pp. 2733-2753.
ION Members/Non-Members: 1 Download Credit