Search Results

While machine learning in autonomous vehicles development has increased significantly in the past few years, the use of reinforcement learning (RL) methods has only recently been applied. Convolutional Neural Networks (CNNs) became common for their powerful object detection and identification and even provided end-to-end control of an autonomous vehicle. However, one of the requirements of a CNN is a large amount of labeled data to inform and train the neural network. While data is becoming more accessible, these networks are still sensitive to the format and collection environment which makes the use of others’ data more difficult. In contrast, RL develops solutions in a simulation environment through trial and error without labeled data. Our research expands upon previous research in RL and Proximal Policy Optimization (PPO) and the application of these algorithms to 1/18th scale cars by expanding the application of this control strategy to a full-sized passenger vehicle.

Prediction of vehicle velocity is important since it can realize improvements in the fuel economy/energy efficiency, drivability, and safety. Velocity prediction has been addressed in many publications. Several references considered deterministic and stochastic approaches such as Markov chain, autoregressive models, and artificial neural networks. There are numerous new sensor and signal technologies like vehicle-to-vehicle and vehicle-to-infrastructure communication that can be used to obtain inclusive datasets. Using these inclusive datasets of sensors in deep neural networks, high accuracy velocity predictions can be achieved. This research builds upon previous findings that Long Short-Term Memory (LSTM) deep neural networks provide low error velocity prediction. We developed an LSTM deep neural network that uses different groups of datasets collected in Fort Collins, Colorado.

The majority of fatigue crack growth (FCG) data associated with load interaction effects have been interpreted in terms of plasticity effects at the crack wake. This paper presents a new approach by which the FCG behavior for different R-ratios can be predicted using the master curve. In this approach, the stress intensity factor (SIF) range, ΔK, is normalized in terms of ΔK at R=0. The corresponding mean value of SIF, Km, is normalized by the SIF amplitude, Ka. A curve is plotted for this normalized SIF data, which collapses all the R-ratios into a master curve and therefore enables the prediction of other R-ratios. From the analyses of different materials taken from literature it is demonstrated that the master curve approach is very effective in predicting the R-ratio effects on fatigue crack growth behavior.

Fatigue damage sensing and monitoring of any structure is a prerequisite for reliable and effective structural health diagnosis. The designed sensor has alternate slots and strips with different strain magnification factor with respect to the nominal strain at its location. The strips experience the strains which closely resemble the actual strain distribution in the critical area of the component. One of the major advantages of this sensor is that it can be placed at any convenient location, still experiencing the same fatigue damage as a critical location. It can be used on various structures from ground civilian and military vehicles to steel bridges. This can predict the remaining useful life of a component or the number of miles (for any automobile) left for the component before it needed replacement. This paper mainly describes the design aspects of this sensor following analytical and finite element analysis (FEA) approaches.

Standard operation of autonomous vehicles on public roads results in significant exposure to high levels of risk. There is a significant need to develop metrics that evaluate safety of an automated system without reliance on the rate of vehicle accidents and fatalities compared to the number of miles driven; a proactive rather than a reactive metric is needed. Resilience engineering is a new paradigm for safety management that focuses on evaluating complex systems and their interaction with the environment. This paper presents the overall methodology of resilience engineering and the resilience assessment grid (RAG) as an evaluation tool to measure autonomous systems' resilience. This assessment tool was used to evaluate the ability to respond to the system. A Pure Pursuit controller was developed and utilized as the path tracking control algorithm, and the Carla simulator was used to implement the algorithm and develop the testing environment for this methodology.

Robust sensor fusion is a key technology for enabling the safe operation of automated vehicles. Sensor fusion typically utilizes inputs of cameras, radars, lidar, inertial measurement unit, and global navigation satellite systems, process them, and then output object detection or positioning data. This paper will focus on sensor fusion between the camera, radar, and vehicle wheel speed sensors which is a critical need for near-term realization of sensor fusion benefits. The camera is an off-the-shelf computer vision product from MobilEye and the radar is a Delphi/Aptive electronically scanning radar (ESR) both of which are connected to a drive-by-wire capable vehicle platform. We utilize the MobilEye and wheel speed sensors to create a dynamic region of interest (DROI) of the drivable region that changes as the vehicle moves through the environment.

Today’s Advanced Driver Assistance Systems (ADAS) predominantly utilize cameras to increase driver and passenger safety. Computer vision, as the enabler of this technology, extracts two key environmental features: the drivable region and surrounding objects (e.g., vehicles, pedestrians, bicycles). Lane lines are the most common characteristic extracted for drivable region detection, which is the core perception task enabling ADAS features such as lane departure warnings, lane-keeping assistance, and lane-centering. However, when subject to adverse weather conditions (e.g., occluded lane lines) the lane line detection algorithms are no longer operational. This prevents the ADAS feature from providing the benefit of increased safety to the driver. The performance of one of the leading computer vision system providers was tested in conditions of variable snow coverage and lane line occlusion during the 2020-2021 winter in Kalamazoo, Michigan.

Modern vehicles use automated driving assistance systems (ADAS) products to automate certain aspects of driving, which improves operational safety. In the U.S. in 2020, 38,824 fatalities occurred due to automotive accidents, and typically about 25% of these are associated with inclement weather. ADAS features have been shown to reduce potential collisions by up to 21%, thus reducing overall accidents. But ADAS typically utilize camera sensors that rely on lane visibility and the absence of obstructions in order to function, rendering them ineffective in inclement weather. To address this research gap, we propose a new technique to estimate snow coverage so that existing and new ADAS features can be used during inclement weather. In this study, we use a single camera sensor and historical weather data to estimate snow coverage on the road. Camera data was collected over 6 miles of arterial roadways in Kalamazoo, MI.

Practical applications of recently developed sensor fusion algorithms perform poorly in the real world due to a lack of proper evaluation during development. Existing evaluation metrics do not properly address a wide variety of testing scenarios. This issue can be addressed using proactive performance measurements such as the tools of resilience engineering theory rather than reactive performance measurements such as root mean square error. Resilience engineering is an established discipline for evaluating proactive performance on complex socio-technical systems which has been underutilized for automated vehicle development and evaluation. In this study, we use resilience engineering metrics to assess the performance of a sensor fusion algorithm for vehicle localization. A Kalman Filter is used to fuse GPS, IMU and LiDAR data for vehicle localization in the CARLA simulator.

An optimal energy management strategy (Optimal EMS) can yield significant fuel economy (FE) improvements without vehicle velocity modifications. Thus it has been the subject of numerous research studies spanning decades. One of the most challenging aspects of an Optimal EMS is that FE gains are typically directly related to high fidelity predictions of future vehicle operation. In this research, a comprehensive dataset is exploited which includes internal data (CAN bus) and external data (radar information and V2V) gathered over numerous instances of two highway drive cycles and one urban/highway mixed drive cycle. This dataset is used to derive a prediction model for vehicle velocity for the next 10 seconds, which is a range which has a significant FE improvement potential. This achieved 10 second vehicle velocity prediction is then compared to perfect full drive cycle prediction, perfect 10 second prediction.

Development of autonomous vehicle technology is expensive and perhaps more complicated than initially thought, as evidenced by the recent rollback of anticipated delivery dates from companies such as Tesla, Waymo, GM, and more. One of the most effective techniques to reduce research and development costs and speed up implementation is rigorous analysis through simulation. In this paper, we present multiple autonomous vehicle perception and control strategies that are rigorously investigated in the user friendly, free, and open-source simulation environment, CARLA. Overall, we successfully formulated potential solutions to the autonomous navigation problem and assessed their advantages and disadvantages in simulation at no cost. First, a lane finding method utilizing polynomial fitting and machine learning is proposed. Then, the waypoint navigation strategy is described, along with route planning. Object detection is then implemented using pre-trained convolutional neural networks.

The transportation sector contributes significantly to emissions and air pollution globally. Emission models of modern vehicles are important tools to estimate the impact of technologies or controls on vehicle emission reductions, but developing a simple and high-fidelity model is challenging due to the variety of vehicle classes, driving conditions, driver behaviors, and other physical and operational constraints. Recent literature indicates that neural network-based models may be able to address these concerns due to their high computation speed and high-accuracy of predicted emissions. In this study, we seek to expand upon this initial research by utilizing several deep neural networks (DNN) architectures such as a recurrent neural network (RNN) and a convolutional neural network (CNN). These DNN algorithms are developed specific to the vehicle-out emissions prediction application, and a comprehensive assessment of their performances is done.

Traditional autonomous vehicle perception subsystems that use onboard sensors have the drawbacks of high computational load and data duplication. Infrastructure-based sensors, which can provide high quality information without the computational burden and data duplication, are an alternative to traditional autonomous vehicle perception subsystems. However, these technologies are still in the early stages of development and have not been extensively evaluated for lane detection system performance. Therefore, there is a lack of quantitative data on their performance relative to traditional perception methods, especially during hazardous scenarios, such as lane line occlusion, sensor failure, and environmental obstructions.