This information is then used in sensor fusion and machine learning algorithms to help predict the brake reaction time of the driver, a key input in forward collision warning systems. We demonstrate the results in a pilot study, which highlights the benefits of the personalized, adaptive reaction time estimation in collision warning alert performance.
We conclude with improvements to further strengthen the quality of the reaction time estimation and improve alert performance. Driver input and safety measures vary for these systems, as there are no legal regulations on how the human-machine-interaction should be designed. In this paper, an interaction concept for a partially automated lane change is presented. The initiation of the lane change includes a request by the driver and a feasibility check by the system. Based on the results of expert interviews, various input variations were implemented and evaluated in a driving study on the motorway.
The experts generally preferred the activation of the indicator lever in combination with steering wheel contact for lane change requests. This input variation enables the driver to express his wish without active intervention in the vehicle guidance.
For activation of a so-called lane change abort, which results in an automated return to the starting lane, the experts preferred steering input to the starting lane. As a next step, progressing technical development could enable support in form of partly active control leading to situations of cooperative shared control between automation and driver. Priming a driver for an automated control action shares basic concepts with directional yet conventional take-over requests. Based on results from the literature concerning directional take-over requests, we designed three different modality configurations.
In contrast to previous studies and because of the background of a later cooperative take-over, we focused on perception, taskload and criticality assessment of the directional information instead of evaluating purely human take-over performance. Results indicate better perception for configurations with an auditory component, yet taskload and criticality ratings were not affected by the different modalities. Each of these takeover situations is different from the other. The shift of control can be hazardous. This study investigated whether a match between the urgency of a TOR and the criticality of the situation benefits drivers regarding their takeover performance and subjective experience.
Fifty-two participants experienced two takeover situations with varying criticality 3 s or 7 s time budget. A low urgent, auditory-tactile TOR was presented to half of the participants. The other half experienced a highly urgent, auditory-tactile TOR. This resulted in two conditions, in which the TOR-urgency and the situational criticality matched, e. Results indicate that a TOR-urgency that matches the criticality of the driving situation does not benefit drivers. However, drivers generally performed better in takeover situations with low criticality and they took over faster in takeover situations with high criticality.
Implications of the results and future work are discussed. Here, we focus on the fact that experienced drivers usually pass by a pedestrian while preparing to step on the brake at any moment when they feel danger. If driver assistant systems can estimate such experienced driver's decisions, they could early detect the pedestrian in danger of collision.
Therefore, we classify the driver's decisions into three types by referring to the accelerator operation of drivers, and propose a method to estimate the type of the driver's decision. The drivers are considered to decide their actions focusing on various behaviors and states of a pedestrian, namely pedestrian's attributes. Since the driver's decisions change along the timeline, the use of a temporal context is considered to be effective.
Thus, in this paper, we propose an estimation method using a recurrent neural network architecture with the pedestrian's attributes as input. We constructed a dataset collected by experienced drivers in control of the vehicle and evaluated the performance, and then confirmed the effectiveness of the use of pedestrian's attributes. One of the most heavily researched components of microsimulation models are car-following models, where human behavior is approximated using driver specific calibration coefficients. Inter-driver heterogeneity in calibration parameters has been observed in naturalistic data, but few studies have explored methods to characterize this heterogeneity.
This research utilizes an driver sample from the second Strategic Highway Research Program Naturalistic Driving Study dataset to calibrate over hours of car-following data to three different car-following models: Gipps, Wiedemann 99, and Intelligent Driver Model. This enables a detailed exploration of the degree to which driver attributes—such as age, gender, and income—can be leveraged to account for inter-driver heterogeneity in car-following. The sample is segmented into smaller subsamples of data as a function of driver attributes. Statistically significant differences in parameter values are observed between the subsamples of data when age, miles driven last year, and income are used as characterizing traits.
Moreover, it is shown that no single model outperforms the other models for all categories of data. With passing an obstacle before, after, right or left, there are multiple options an autonomous vehicle could choose to execute.
These combinatorial aspects need to be taken into account in the planning framework. We address this problem by proposing a novel planning approach that combines trajectory planning and maneuver reasoning. We define a classification for dynamic obstacles along a reference curve that allows us to extract tactical decision sequences.
We separate longitudinal and lateral movement to speed up the optimization-based trajectory planning. To map the set of obtained trajectories to maneuver variants, we define a semantic language to describe them. This allows us to choose an optimal trajectory while also ensuring maneuver consistency over time. We demonstrate the capabilities of our approach for a scenario that is still widely considered to be challenging.
This paper presents a situational awareness and trajectory planning framework for performing autonomous overtaking manoeuvres. A combination of a potential field like function and reachability sets of a vehicle are used to identify safe zones on a road that the vehicle can navigate towards.
Intelligent transportation system - Wikipedia
These safe zones are provided to a model predictive controller as reference to generate feasible trajectories for a vehicle. The strengths of the proposed framework are: i it is free from non-convex collision avoidance constraints, ii it ensures feasibility of trajectory, and iii it is real-time implementable. A proof of concept simulation is shown to demonstrate the ability to plan trajectories for high-speed overtaking manoeuvres.
While the existing potential field models for trajectory generation are hard to consider dynamic aspects of obstacles, the proposed potential field model overcomes such limitation by changing the risk size of the potential field. To express the risk function of moving object, the constant time gap policy is utilized. By combining the adaptive potential field and optimal trajectory generation scheme, the proposed driving framework successfully makes autonomous vehicles perform a variety of highway driving functionalities such as lane keeping, lane changing, distance keeping and collision avoidance in critical situations.
Progress maximization plays an important role not only in racing, but also in efficient and safe autonomous driving applications. The progress maximization problem is formulated as a model predictive controller, where the vehicle model is successively linearized at each time step, yielding a convex optimization problem. To ensure real-time feasibility, a kinematic vehicle model is used together with several linear approximations of the vehicle dynamics constraints. We propose a novel polytopic approximation of the 'g-g' diagram, which models the vehicle handling limits by constraining the lateral and longitudinal acceleration.
Moreover, the tire slip angles are restricted to ensure that the tires of the vehicle always operate in their linear force region by limiting the lateral acceleration. We illustrate the effectiveness of the proposed controller in simulation, where a nonlinear dynamic vehicle model is controlled to maximize the progress along a track, taking into consideration possible obstacles.
The proposed algorithm provides a path for parking, whose curvature satisfies the clothoid constraint, so that the vehicle can smoothly follow the path. In the path planning, we introduce an innovative trajectory folding scheme which does not require a high-computational cost. Trajectory folding scheme can be used to create a general vehicle control path for longitudinal and lateral movement in desired parking slots.
The folding scheme has an advantage that it can make a path satisfying the physical constraint of the vehicle while considering the surrounding obstacles.
Physics-Based Optimization of Access Point Placement for Train Communication Systems
Finally, we introduce a completely automated parking system using the lateral motion control and a simple proportional control for longitudinal motion. The proposed scheme was validated via computational simulation results showing satisfactory performances on the tested scenario. Our contributions are two fold. First, we introduce a general, flexible and complete speed planning optimisation which includes time efficiency, smoothness objectives and dynamics, friction circle, boundary condition constraints which addresses limitations of existing methods and is able to provide smooth, safety-guaranteed, dynamically-feasible, and time-efficient speed profiles.
Second, we demonstrate that our problem preserves convexity with the expanded set of constraints, and hence, global optimality of solutions is guaranteed. We show how our formulation can be used in speed planning for various autonomous driving scenarios by providing several typical case studies in both static and dynamic environments.
The results depict that the proposed method outperforms existing speed planners for autonomous driving in terms of capacity, optimality, safety and mobility. A classification of stereo camera sensor faults is initially introduced, the most common types of defects being highlighted. We next present a stereo camera failure detection method in which various additional checks are being introduced, with respect to the aforementioned error classification.
Furthermore, we propose a novel error correction method based on CNNs convolutional neural networks that is capable of generating reliable disparity maps by using prior information provided by semantic segmentation in conjunction with the last available disparity. We highlight the efficiency of our approach by evaluating its performance in various driving scenarios and show that it produces accurate disparities on images from Kitti stereo and raw datasets while running in real-time on a regular GPU.
However, they cannot detect crossing pedestrians when turning at intersections because the field of view is not wide enough. Thus, we propose to utilize a surround view camera system becoming popular by making it into stereo vision which is robust for the pedestrian recognition. However, conventional stereo camera technologies cannot be applied due to fisheye cameras and uncalibrated camera poses.
Thus we have created the new method to absorb difference of the pedestrian appearance between cameras by machine learning for the stereo vision. Good generalization performance was achieved by it compared with individual conventional algorithms. Furthermore, feature amounts of the point cloud reconstructed by the stereo pairs are utilized with Random Forest to discriminate pedestrian.
The algorithm was evaluated for the actual camera images of crossing pedestrians at various intersections, and They were compared with Faster R-CNN as the best pattern recognition technique, and our proposed method indicated better detection performance. Recent studies on depth completion have leveraged deep learning approaches applying traditional convolutional neural networks to prediction of invisible information in sparse and irregular inputs. Due to the lack of local and global structures such as object boundary cues, however, the predicted information results in unstructured and noisy depth maps.
This paper presents a supervised depth completion method using an adversarial network based only on sparse inputs.
- Intelligent transportation system?
- Transportation Research Groups | Transportation @ MIT.
- Artificial Neural Networks and Neural Information Processing — ICANN/ICONIP 2003: Joint International Conference ICANN/ICONIP 2003 Istanbul, Turkey, June 26–29, 2003 Proceedings;
- Program for Tuesday, September 5th.
- Intelligent transportation system - Wikipedia.
- Study on Development of Train Running Diagram Optimization;
- PhD on Robust train trajectory optimization — AcademicTransfer.
In the presented method, a fully convolutional depth completion network, along with the adversarial network, is designed to find and correct inconsistencies between ground truth distributions and the resulting depth maps interpolated by the depth completion network. This leads to more realistic and structured depth images without compromising runtime performance of inference.
With recent advances in long-term autonomous navigation, change detection under global viewpoint uncertainty has become a topic of considerable interest.