Autonomous Vehicles are complex modular systems with various inter-dependent safety critical
modules, the failure of which leads to failure of the overall system. The Localization
system, which estimates the pose of the vehicle in the global coordinate frame with respect
to a map, has a drift in error, when operated only based on data from proprioceptive sensors.
Current solutions to the problem are computationally heavy SLAM algorithms. An alternate
system is proposed in the thesis which eliminates the drift by resetting the global coordinate
frame to the local frame at every motion planning update. The system replaces the mission
planner with a user interface(UI) onto which the User provides local navigation inputs, thus
eliminating the need for maintenance of a Global frame. The User Input is considered in the
decision framework of the behavioral planner, which selects a safe and legal maneuver for the
vehicle to follow. The path and trajectory planners generate a trajectory to accomplish the
maneuver and the controller follows the trajectory until the next motion planning update.
A prototype of the system has been built on a wheeled robot and tested for the feasibility
of continuous operation in Autonomous Vehicles. / Master of Science / Autonomous Vehicles are complex modular systems with various inter-dependent safety critical
modules, the failure of which leads to failure of the overall system. One such module
is the Localization system, that is responsible for estimating the pose of the vehicle in the
global coordinate frame, with respect to a map. Based on the pose, the vehicle navigates
to the goal waypoints, which are points in the global coordinate frame specified in the map
by the route or mission planner of the planning module. The Localization system, however,
consists of a drift in position error, due to poor GPS signals and high noise in the inertial sensors.
This has been tackled by applying computationally heavy Simultaneous Localization
and Mapping based methods, which identify landmarks in the environment at every time
step and correct the vehicle position, based on the relative change in position of landmarks.
An alternate solution is proposed in this thesis, which delegates navigation to the passenger.
This system replaces the mission planner from the planning module with a User Interface
onto which the passenger provides local Navigation input, which is followed by the vehicle.
The system resets the global coordinate frame to the vehicle frame at every motion planning
update, thus eliminating the error accumulated between the two updates. The system is also
designed to perform default actions in the absence of user Navigation commands, to reduce
the number of commands to be provided by the passenger in the journey towards the goal.
A prototype of the system is built and tested for feasibility.
Identifer | oai:union.ndltd.org:VTETD/oai:vtechworks.lib.vt.edu:10919/108401 |
Date | 26 August 2020 |
Creators | Keshavamurthi, Karthik Balaji |
Contributors | Mechanical Engineering, Taheri, Saied, Ferris, John B., Sandu, Corina |
Publisher | Virginia Tech |
Source Sets | Virginia Tech Theses and Dissertation |
Detected Language | English |
Type | Thesis |
Format | ETD, application/pdf |
Rights | In Copyright, http://rightsstatements.org/vocab/InC/1.0/ |
Page generated in 0.0113 seconds