The perception ability of automated systems such as autonomous cars plays an outstanding role for safe and reliable functionality. With the continuously growing accuracy of deep neural networks for object detection on one side and the investigation of appropriate space representations for object tracking on the other side both essential perception parts received special research attention within the last years. However, early fusion of multiple sensors turns the determination of suitable measurement spaces into a complex and not trivial task. In this paper, we propose the use of a deep multi-modal object detection network for the early fusion of LiDAR and camera data to serve as a measurement source for an extended object tracking algorithm on Lie groups. We develop an extended Kalman filter and model the state space as the direct product Aff(2) × ℝ6 incorporating second- and third-order dynamics. We compare the tracking performance of different measurement space representations-SO(2) × ℝ4, SO(2)2 × ℝ3 and Aff(2)-to evaluate, how our object detection network encapsulates the measurement parameters and the associated uncertainties. With our results, we show that the lowest tracking errors in the case of single object tracking are obtained by representing the measurement space by the affine group. Thus, we assume that our proposed object detection network captures the intrinsic relationships between the measurement parameters, especially between position and orientation.
The role of autonomous cooperative vehicles will undoubtedly be important in Intelligent Transportation Systems (ITS) to increase both the safety and the overall efficiency of a high traffic network system. An autonomous platooning provides one promising strategy for decreasing total fuel consumption of a fleet of vehicles and potential risk of accidents, especially during long-distance transportation. In this work, we provide a proof-of-concept for a simulation framework in which it is possible to simulate platoon and other multi-vehicle systems using realistic vehicle models within different traffic scenarios, which is based on ROS, Gazebo and SUMO. The framework enables an easy-to-use perception and control modules of the autonomous driving stack for a realistic vehicle models, while preserving a convenient setup of different high traffic platooning scenarios. Consequently, it provides a platooning design step for conducting reliable development analyses and a platform for comparisons of different platooning strategies. We illustrate the effectiveness of the proposed platooning framework through three typical scenarios using a distributed model predictive control scheme with a platoon consisted of Toyota Prius car models.
Estimation of vehicle motion is a pivotal requirement for autonomous vehicles. This paper proposes a robust ego-vehicle motion estimation to achieve precise localization and tracking, especially in the case of highly dynamic driving. An extended H∞ filter, based on a kinematic motion model assuming constant turn-rate and acceleration is used to fuse LiDAR, IMU, and vehicle dynamic sensors’ measurements. Measurements from a real high-performance autonomous race car, the so-called DevBot 2.0, have been used to validate the fusion approach in a Roborace competition and compared to a standard Kalman-filter approach.The proposed estimation concept adapts the H∞ robustness bound based on the innovation sequence of the filter. This provides very fast tracking when it comes to highly dynamic movement, but still achieves minimal estimation uncertainty in case of stationary conditions with lower innovation. Furthermore, a pure kinematic model is used, which is robust against vehicle parameters, changes in the tire-road conditions, and changes in driving maneuvers. The resulting estimation concept shows outstanding performance for considered autonomous race scenario and can be used for a wide range of different applications, such as highway driving, urban driving, platooning, etc.
Proactive caching is a promising approach to achieve efficient content delivery, reduce content retrieval latency, and improve user experience in vehicular content networks. This paper proposes a mobility prediction based proactive caching scheme utilizing a sequence prediction algorithm, namely Sequence Prediction-based Proactive Caching, to predict the next possible RSU along a vehicle’s path and pre-locate relevant content. Four systems’ performance is evaluated in two areas of Las Vegas and Manchester. The obtained results in Las Vegas have shown that the proposed system outperforms the other three systems i.e., Baseline Proactive Caching system, non-proactive caching system and no-caching system. It is shown to be up to over three times and twice better than the non-proactive caching system and Baseline Proactive Caching system respectively in terms of cache performance and on average, network delay of SPPC is reduced by 18% and 24% compared with non-proactive caching system and no-caching system respectively. Performance benchmark in Manchester generalized the application of SPPC system and asserted its superiority. The paper also gives insight into solving prediction issues with data mining techniques.
Emerging Vehicle-to-Everything (V2X) technologies promise to improve the perception of streets by enabling data sharing like camera views between multiple vehicles. However, to ensure accuracy of such enhanced perception, the problem of vehicle matching becomes important; the goal of a vehicle matching system is to identify if images of vehicles seen by different cameras correspond to the same vehicle. Such a system is necessary to avoid duplicate detections for a vehicle seen by multiple cameras and to avoid detections being discarded due to a false match being made. One of the most challenging scenarios in vehicle matching is when the camera positions have very large viewpoint differences, as will commonly be the case when the cameras are in geographically separate locations like in vehicles and street infrastructure. In these scenarios, traditional handcrafted features will not be sufficient to create these correspondences due to the lack of common visual features. In this paper we will examine the performance of random forests and neural networks as classifiers for both learned features and high level visual features when used for this vehicles matching problem. Additionally, a novel dataset of vehicles from cameras with very large viewpoint differences was recorded to validate our method; our preliminary results achieve high classification accuracy with low inference time which shows the feasibility of a real time vehicle matching system.
