Browsing by Subject "Autonomous"
Now showing 1 - 4 of 4
- Results Per Page
- Sort Options
Item Adapting to unseen driving conditions using context-aware neural networksAbdulquddos, Suhaib; Miikkulainen, Risto; Tutum, Cem COne of the primary inhibitors to successful deployment of autonomous agents in real-world tasks such as driving is their poor ability to adapt to unseen conditions. Whereas a human might be able to deduce the best course of action when confronted with an unfamiliar set of conditions based on past experiences, artificial agents have difficulty performing in conditions that are significantly different from those in which they were trained. This thesis explores an approach in which the addition of a context module to a neural network is used to overcome the challenge of adapting to unseen conditions during evaluation. The approach is tested in the CARLA simulator wherein the torque and steering curves of a vehicle are modified during training and evaluation. Furthermore the agent is trained only on a track with a relatively large radius of curvature but is evaluated on a track with much sharper turns and the agent must learn to adapt its speed and steering during evaluation. Three different neural network architectures are used for these experiments, and their respective performances are compared: Context+Skill, Context only, Skill only. It is observed that when both performance and safety of agents behavior are considered, the context+skill network consistently outperforms both the skill only and the context only architectures. The results presented in this thesis indicate that the context aware approach is a promising step towards solving the generalization problem in the autonomous vehicle domain. Furthermore, this research presents a framework for comparing the generalization capabilities of various network architectures and approaches. It is posited that the context+skill neural network has the potential to advance the field of machine learning with regards to generalization in domains beyond just autonomous driving; that is, any domain where awareness of changing environment parameters can have a positive impact on performance.Item Analysis and order reduction of an autonomous lunar lander navigation system(2009-08) Newman, Clark Patrick; Bishop, Robert H., 1957-; Akella, Maruthi R.A navigation system for precision lunar descent and landing is presented and analyzed. The navigation algorithm is based upon the extended Kalman Filter and employs measurements from an inertial measurement unit to propagate the vehicle position, velocity, and attitude forward in time. External measurements from an altimeter, star camera, terrain camera, and velocimeter are utilized in state estimate updates. The navigation algorithm also attempts to estimate the values of uncertain parameters associated with the sensors. The navigation algorithm also estimates the map-tie angle of the landing site which is a measure of the misalignment of the actual landing site location on the surface of the Moon versus the estimated position of the landing site. The navigation algorithm is subject to a sensitivity analysis which investigates the contribution of each error source to the total estimation performance of the navigation system. Per the results of the sensitivity analysis, it is found that certain error sources need not be actively estimated to achieve similar estimation performance at a reduced computational burden. A new, reduced-order system is presented and tested through covariance analysis and a monte carlo analysis. The new system is shown to have comparable estimation performance at a fraction of the computer run-time, making it more suitable for a real-time implementation.Item Computationally efficient algorithms for spacecraft relative navigation and rendezvous(2023-08-02) Kaki, Siddarth Bhargava; Akella, Maruthi Ram, 1972-; Jones, Brandon; Zanetti, Renato; Russell, Ryan; D’Souza, ChristopherThere is tremendous interest in the development of spacecraft guidance, navigation, and control technologies that enable on-orbit servicing, assembly, and manufacturing missions with limited human supervisory support. These applications provide strong motivations for the design of computationally lightweight algorithms that enable autonomous operations, especially for missions constrained by size, weight, power, and cost (SWaP-C) such as cubesats. However, the more democratic access to space has also further contributed to the space debris problem, with worryingly many defunct and derelict spacecraft crowding up precious space. Dealing with such noncooperative objects poses significant challenges, with large uncertainties in mass properties and applied forces by the environment. This dissertation addresses such problems related to SWaP-C constraints and also noncooperative targets within the fields of relative spacecraft navigation and rendezvous. The first topic addresses relative angular velocity and associated uncertainty estimation from a geometrical perspective via batch processes. The second topic addresses real-time relative pose estimation and tracking using monocular imagery for cubesat applications. The third topic addresses analytical guidance solutions for radial-thrust-based far-field rendezvous and orbit rotations. Finally, the fourth topic addresses semi-analytical techniques to achieve circular orbit changes with radial and velocity-normal thrust.Item Potential impacts of connected-autonomous vehicles on congestion and safety : a look at Austin, Texas(2017-05-02) Archer, Jackson Longstreet; Zhang, Ming, 1963 April 22-; Jiao, JunfengData is a central component of Connected-Autonomous Vehicle (CAV) systems: the advantages and potential challenges of both vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) CAV data underlie the question of wide scale CAV implementation. This report looks at the potential congestion and safety benefits of a vehicle system highly saturated with CAVs in Austin, Texas. Traffic factors such as capacity, intersection delay, and crash rate are examined with respect to their effect on an urban corridor in Austin. The case study relies almost entirely on collected field data to be used as a comparison against potential CAV advantages. In addition to a presentation of the quantitative benefits of CAVs, an infrastructure placement scheme that maximizes data transmission efficiency is also proposed. The results find that vehicle systems can see large improvements in capacity, intersection delay, and number of crashes, and at a relatively inexpensive cost.