This project aims to investigate the possibilities for a passive infrastructure-less self-localization by vehicles for navigation purposes. Such localization is an essential precondition for the half-autonomous control of convoys, which was raised by the military user, the Austrian Federal Ministry of Defence. Up to here autonomous car navigation in nearly all variants is supported by global navigation satellite systems (GNSS) and / or active sensors (radar, lidar). For reasons of defilade active means must be waived. Off-road driving conditions do not offer any infrastructure. And GNSS must be waived, too, since it aggressively may be jammed or spoofed. The remaining means are mainly camera based localization, inertial navigation and the evaluation of the vehicle’s sensor data. It is now necessary to investigate by which algorithms and methods a reliable localization is possible, based on only these reduced means.
These are now the goals of the project, to reach an automated vehicle’s localization under the above mentioned conditions. Firstly it is necessary to obtain realistic data sets from cameras, inertial measurement unit (IMU) and from the vehicle’s sensors and to emulate autonomous driving conditions. For this purpose drive-by-wire capabilities of the test vehicles will be implemented, which up to here is prohibited in road traffic. Using drive-by-wire omits those characteristics of the vehicle’s control, which are generated by the driver’s subjective impressions. Especially in off- road conditions reflexive reactions by the driver may be significant. In a totally new approach the data from the cameras, from the IMU and from the vehicle’s sensors will be used to generate a map of the driven path and to estimate the vehicle’s location with respect to the map. Obstacles will be recognized and the vehicle will be stopped. Technology readiness level 1 (TRL 1), the proof of the basic principle, is already given and is the starting point of the project. With the successful experimental proof finally TRL 3 will be prepared and reached. According to the low TRL, some simplifying conditions will be established, like a speed limit of 30 kilometers per hour or constant vegetation conditions.
Target goals and findings are the successful implementation of a demonstrator of principle. For validation purposes a reference trajectory with all available means (GNSS, etc.) will be acquired, too. The found trajectory in comparison with the reference trajectory finally shall reach an accuracy, which is acceptable for the targeted application. Challenges in the methods shall be recognized and shall be accordingly be addressed. Visual localization, inertial navigation and also the evaluation of the vehicle’s sensor data are affected by systematic inaccuracies, which is to be eliminated by data fusion techniques as much as possible. The concrete project results are the algorithms for determining the trajectory, which are openly explained and documented in the project’s reports. Further, the acquired data sets and their analyses are part of the project results and the documentation of the experimental setup.