Fully automated or autonomous vehicles (AVs) hold great promise for the future of transportation, with Google and other auto manufacturers intending on introducing self-driving cars to the public by 2020. New automation functionalities will produce dramatic transportation system changes, across safety, mobility, travel behavior, and the built environment. This work’s results indicate that AVs may save the U.S. economy up to $37.7 billion from safety, mobility and parking improvements at the 10% market penetration level (in terms of system-wide vehicle-miles traveled [VMT]), and up to $447.1 billion with 90% market penetration. With only 10% market share, over 1,000 lives could be saved annually. However, realizing these potential benefits while avoiding pitfalls requires overcoming significant barriers including AV costs, liability, security, privacy, and missing research. Additionally, once fully self-driving vehicles can safely and legally drive unoccupied, a new personal travel transportation mode looks set to arrive. This new mode is the shared automated vehicle (SAV), combining on-demand service features with self-driving capabilities. This work simulates a fleet of SAVs operating within Austin, Texas, first using an idealized grid-based representation, and next using Austin’s actual transportation network and travel demand flows. This second model incorporates dynamic ride-sharing (DRS), allowing two or more travelers with similar origins, destinations and departure times to share a ride. Model results indicate that each SAV could replace around 10 conventionally-owned household vehicles while serving over 56,000 person-trips. SAVs’ ability to relocate unoccupied between serving one traveler and the next may cause an increase of 7-10% more travel; however, DRS can result in reduced overall VMT, given enough SAV-using travelers willing to ride-share. Furthermore, using DRS results in overall lower wait and service times for travelers, particularly from pooling rides during peak demand. SAVs should produce favorable emissions outcomes, with an estimated 16% less energy use and 48% lower volatile organic compound (VOC) emissions, per person-trip compared to conventional vehicles. Finally, assuming SAVs cost $70,000 each, an SAV fleet in Austin could provide a 19% return on investment, when charging $1 per trip-mile served. In summary, this new paradigm holds much promise that technological advances may soon realized. / text

No description available.

"Autonomous vehicle is an engineering technology that can improve transportation safety, alleviate traffic congestion and reduce carbon emissions. Research on autonomous vehicles can be categorized by functionality, for example, object detection or recognition, path planning, navigation, lane keeping, speed control and driver status monitoring. The research topics can also be categorized by the equipment or techniques used, for example, image processing, computer vision, machine learning, and localization. This dissertation primarily reports on computer vision and machine learning algorithms and their implementations for autonomous vehicles. The vision-based system can effectively detect and accurately recognize multiple objects on the road, such as traffic signs, traffic lights, and pedestrians. In addition, an autonomous lane keeping system has been proposed using end-to-end learning. In this dissertation, a road simulator is built using data collection and augmentation, which can be used for training and evaluating autonomous driving algorithms. The Graphic Processing Unit (GPU) based traffic sign detection and recognition system can detect and recognize 48 traffic signs. The implementation has three stages: pre-processing, feature extraction, and classification. A highly optimized and parallelized version of Histogram of Oriented Gradients (HOG) and Support Vector Machine (SVM) is used. The system can process 27.9 frames per second with the active pixels of a 1,628 by 1,236 resolution, and with the minimal loss of accuracy. In an evaluation using the BelgiumTS dataset, the experimental results indicate that the detection rate is about 91.69% with false positives per window of 3.39e-5, and the recognition rate is about 93.77%. We report on two traffic light detection and recognition systems. The first system detects and recognizes red circular lights only, using image processing and SVM. Its performance is better than that of traditional detectors and it achieves the best performance with 96.97% precision and 99.43% recall. The second system is more complicated. It detects and classifies different types of traffic lights, including green and red lights in both circular and arrow forms. In addition, it employs image processing techniques, such as color extraction and blob detection to locate the candidates. Subsequently, a pre-trained PCA network is used as a multi-class classifier for obtaining frame-by-frame results. Furthermore, an online multi-object tracking technique is applied to overcome occasional misses and a forecasting method is used to filter out false positives. Several additional optimization techniques are employed to improve the detector performance and to handle the traffic light transitions. A multi-spectral data collection system is implemented for pedestrian detection, which includes a thermal camera and a pair of stereo color cameras. The three cameras are first aligned using trifocal tensor, and the aligned data are processed by using computer vision and machine learning techniques. Convolutional channel features (CCF) and the traditional HOG+SVM approach are evaluated over the data captured from the three cameras. Through the use of trifocal tensor and CCF, training becomes more efficient. The proposed system achieves only a 9% log-average miss rate on our dataset. Autonomous lane keeping system employs an end- to-end learning approach for obtaining the proper steering angle for maintaining a car in a lane. The convolutional neural network (CNN) model uses raw image frames as input, and it outputs the steering angles corresponding to the input frames. Unlike the traditional approach, which manually decomposes the problem into several parts, such as lane detection, path planning, and steering control, the model learns to extract useful features on its own and learns to steer from human behavior. More importantly, we find that having a simulator for data augmentation and evaluation is important. We then build the simulator using image projection, vehicle dynamics, and vehicle trajectory tracking. The test results reveal that the model trained with augmented data using the simulator has better performance and achieves about a 98% autonomous driving time on our dataset. Furthermore, a vehicle data collection system is developed for building our own datasets from recorded videos. These datasets are used in the above studies and have been released to the public for autonomous vehicle research. The experimental datasets are available at http://computing.wpi.edu/Dataset.html."

The development, implementation, and testing of a leader-follower based robust nonlinear formation controller is discussed in this thesis. This controller uses sliding mode control on the length and angle between the leader and follower vessels to produce the desired formation. A boat model, assuming planar motion (three degrees of freedom), is used as the bases for the controller. Open loop testing is performed to determine parameter values to match the simulation model to the physical one and, upon tuning of the controller to match, closed loop testing of the controller with a virtual leader is also performed. From these tests it is found that the controller is unstable, thus improvements to the controller, through changes made to the model and to the parameter identification process, are undertaken. Simulations comparing the initial and updated models of the vehicle to open loop data show an improvement in the new model.

It is known in the literature on Automated Highway Systems that information flow can significantly affect the propagation of errors in spacing in a collection of vehicles. This thesis investigates this issue further for a homogeneous collection of vehicles. Specifically, we consider the effect of information flow on the propagation of errors in spacing and velocity in a collection of vehicles trying to maintain a rigid formation. The motion of each vehicle is modeled using a Linear Time Invariant (LTI) system. We consider undirected and connected information flow graphs, and assume that that each vehicle can communicate with a maximum of q(n) vehicles, where q(n) may vary with the size n of the collection. The feedback controller of each vehicle takes into account the aggregate errors in position and velocity of the vehicles, with which it is in direct communication. The controller is chosen in such a way that the resulting closed loop system is a Type-2 system. This implies that the loop transfer function must have at least two poles at the origin. We then show that if the loop transfer function has three or more poles at the origin, and if the size of the formation is sufficiently large, then the motion of the collection is unstable. Suppose l is the number of poles of the transfer function relating the position of a vehicle with the control input at the origin of the complex plane, and if the number (q(n)l+1)/(nl) -> 0 as n -> (Infinity), then we show that there is a low frequency sinusoidal disturbance with unity maximum amplitude acting on each vehicle such that the maximum errors in spacing response increase at least as much as O (square_root(n^l/(q(n)^(l+1)) ) consequence of the results presented in this paper is that the maximum of the error in spacing and velocity of any vehicle can be made insensitive to the size of the collection only if there is at least one vehicle in the collection that communicates with at least O(square_root(n)) other vehicles in the collection.

The DARPA Urban Challenge in 2007 showed that fully autonomous vehicles, driven by computers without human intervention on public roads, are technologically feasible with current intelligent vehicle technology [6]. Some researchers predict that within 5-20 years there will be autonomous vehicles for sale on the automobile market. Therefore, the time is right to rethink our current transportation infrastructure, which is primarily designed for human drivers, not autonomous vehicles. The Autonomous Intersection Management (AIM) project at UT Austin aims to propose a large-scale, real-time framework to be a substitute for current traffic light and stop signs. Automobiles in modern urban settings spend a lot of time idling at intersections. In 2007, US drivers wasted 4.16 billion hours of their time and 2.81 billion gallons of gas in congestion, costing a total of 87.2 billion dollars nationwide [18]. A big portion of this waste takes place at intersections. The AIM project is able to utilize the capacity of intersections to minimize time waste and fuel consumption. The fundamental idea of Autonomous Intersection management (AIM) [13] is a reservation system in which cells in space-time will be reserved by the au- tonomous vehicles based on their trajectories. An intersection manager takes care of the reservation as well as communication with the vehicles. This mechanism tries to maximize the usage of the intersection area. It ensures a collision free intersection as well. The main question of this project is what intersection control mechanism is appropriate for reducing an autonomous vehicle's waiting time and improving the throughput of the intersection. Previous work proposed the first-come-first-served (FCFS) policy in which the reservation requests are served as soon as they are received. The results of simulation show that FCFS outperforms the current traffic systems, traffic light and stop sign, by orders of magnitude. We, however, observe that FCFS performs suboptimal in certain traffic patterns that are pretty common in urban settings. In this project, first we study the limitations of FCFS, then develop a more efficient policy to alleviate these limitations. The idea that we examined is a systematic policy of granting reservations that have the objective of minimizing the cost of delaying vehicles. In an attempt to build the system in reality, we used miniature robots called Eco-be. Due to their cost and size, Eco-bes are good candidates for testing a multi-agent system with a large number of agents. In spite of the fact that the physical challenges of Eco-bes do not perfectly match those of full size autonomous vehicles, they are still useful for demonstration and education purposes as well as for the study of collisions for which experiments with full size vehicles are costly and dangerous. / text

No description available.

In a variety of military operations, a convoy of autonomous followers may need to traverse the leader's path without using Global Positioning System (GPS), lane markers/magnets and/or a vision-based vehicle-following system. This can be achieved by using Visual Teach and Repeat (VT and R), which provides an effective method for autonomous repeating of a previously driven path. This thesis describes the design of a distributed control system that uses the idea behind the VT and R method to allow a convoy of inter-communicable autonomous vehicles to follow a manually-driven lead vehicle's path with a desired inter-vehicle spacing, even when the leader is not in the camera view of the followers. The longitudinal controller is designed for addressing a 1D spacing problem and then combined with a path tracker for tracking a path in a 2D environment. The designed control model is tested in simulations.

In a variety of military operations, a convoy of autonomous followers may need to traverse the leader's path without using Global Positioning System (GPS), lane markers/magnets and/or a vision-based vehicle-following system. This can be achieved by using Visual Teach and Repeat (VT and R), which provides an effective method for autonomous repeating of a previously driven path. This thesis describes the design of a distributed control system that uses the idea behind the VT and R method to allow a convoy of inter-communicable autonomous vehicles to follow a manually-driven lead vehicle's path with a desired inter-vehicle spacing, even when the leader is not in the camera view of the followers. The longitudinal controller is designed for addressing a 1D spacing problem and then combined with a path tracker for tracking a path in a 2D environment. The designed control model is tested in simulations.

Light Detection and Ranging (LiDAR) systems are expected to become the de facto sensors of choice for autonomous vehicles and robotics systems due to their long range and high resolution, allowing them to map the environment accurately. Current available LiDAR systems are based on mechanical apparatus and discrete components that result in large, bulky, and expensive systems with yet-to-be-proven reliability. The advent of Silicon Photonics technology, advanced CMOS foundries allow us to fabricate miniaturized optical components such as phased arrays that combined enable reliable, solid-state, and cost-effective chip-scale LiDAR systems. Furthermore, Silicon Photonics based platform has the advantage of integrating many complex optical components in to a single chip. It is possible to realize an optical phased array based on waveguides with gratings for emitters. These emitters allow to steer the beam by tuning the source's wavelength exploiting the grating's sensitivity to wavelength in one axis and standard phase tuning on the other axis. Such a steering scheme requires only N phase shifters for an N-channel system thus leading to high power efficiency. Another example that could leverage the Silicon Photonics platform is a full coherent LiDAR system utilizing Frequency-Modulated Continuous-Wave (FMCW) detection scheme that was recently reported. However, miniaturizing a LiDAR system to chip-scale has many challenges. The work in this dissertation presents solutions to some of the key challenges we face in order to demonstrate high performance LiDAR based on phased array. One key challenge is the trade-off between beam divergence and field of view. Here, we show a platform based on silicon-nitride/silicon that achieves simultaneously minimal beam divergence and maximum field of view while maintaining performance that is robust to fabrication variations. In addition, in order to maximize the emission from the entire length of the grating, we design the grating’s strength by varying its duty cycle (apodization) to emit uniformly. We fabricate a millimeter long grating emitter with diffraction-limited beam divergence of 0.089°. Another challenge that is intertwined with the aperture length mention before is how maximizing the steering range in an optical phased array. The array's field of view that is perpendicular to the light propagation is governed by the spacing between emitters. In contrast to Radio Frequency based devices, achieving maximum field of view by placing the emitters at half wavelength pitch to avoid side lobes, is challenging for optical phased arrays as the size of the mode is comparable to the wavelength that give rise to cross-talk issues. Emitter pitch that is larger than half the wavelength induce grating lobes in the steered range, effectively limiting the field of view. The closer together the waveguides, the shorter emitters must be to avoid cross-talk, fundamentally limiting the spot size at the farfield. Cross-talk between waveguides induces wavefront aberrations in the beam, thereby increasing beam divergence and limiting the system resolution and range. Here, we improve the mode confinement in the waveguide by increasing the index along the waveguide axis. We use thin Silicon rods, known as metamaterials, between the emitters to tightly confine the mode in the waveguide. Concentrating the mode in the waveguide reduces cross-talk between emitters and maximizes the optical phased array field of view. By embedding an array in a Mach–Zehnder interferometer we demonstrate a sensitive method of measuring cross-talk between the waveguide. We also measure in the nearfield the width of an array of waveguides over a millimeter long emitters. We show that by using the metamaterials we can realize a dense array with a pitch of 1.2 µm over a millimeter long waveguides with gratings at negligible cross-talk. This short pitch allows for 83° steering angle range (Field of View). Combining this the work of Silicon Nitride based long gratings, will allow for a LiDAR system with minimal beam divergence while achieving record large Field of View. Finally, the last chapter discusses Subwavelength Grating structures that due to their sub-wavelength dimensions guide light without diffraction. These structures allow us to tailor the required effective index by varying their duty cycle. We evaluate their robustness to fabrication variations by embedding them inside a sensitive race track. Using this resonator we measured the sensitivity of Subwavelength Grating structures to an off-set in the element's location, elements' width, duty cycle variation, and width change of a single element. Lastly, we show that due to their periodic structure, they are also robust to as many as three consecutive missing elements. This protection property opens the possibility of realizing a plethora of new devices not possible using wire waveguides. One such example is a T-splitter in which an incoming Transverse Magnetic polarized mode could be split to two separate branches at a 90° angle. The demonstrated platform we show here paves the way for on-chip LiDAR systems for autonomous automotive, robotics, wireless communications, and particle trapping.