Matching Items (67)

133211-Thumbnail Image.png

Beyond Deep Learning: Synthesizing Navigation Programs using Neural Turing Machines

Description

This thesis aims to improve neural control policies for self-driving cars. State-of-the-art navigation software for self-driving cars is based on deep neural networks, where the network is trained on a dataset of past driving experience in various situations. With previous

This thesis aims to improve neural control policies for self-driving cars. State-of-the-art navigation software for self-driving cars is based on deep neural networks, where the network is trained on a dataset of past driving experience in various situations. With previous methods, the car can only make decisions based on short-term memory. To address this problem, we proposed that using a Neural Turing Machine (NTM) framework adds long-term memory to the system. We evaluated this approach by using it to master a palindrome task. The network was able to infer how to create a palindrome with 100% accuracy. Since the NTM structure proves useful, we aim to use it in the given scenarios to improve the navigation safety and accuracy of a simulated autonomous car.

Contributors

Agent

Created

Date Created
2018-05

132967-Thumbnail Image.png

Learning Generalized Heuristics Using Deep Neural Networks

Description

Classical planning is a field of Artificial Intelligence concerned with allowing autonomous agents to make reasonable decisions in complex environments. This work investigates
the application of deep learning and planning techniques, with the aim of constructing generalized plans capable of

Classical planning is a field of Artificial Intelligence concerned with allowing autonomous agents to make reasonable decisions in complex environments. This work investigates
the application of deep learning and planning techniques, with the aim of constructing generalized plans capable of solving multiple problem instances. We construct a Deep Neural Network that, given an abstract problem state, predicts both (i) the best action to be taken from that state and (ii) the generalized “role” of the object being manipulated. The neural network was tested on two classical planning domains: the blocks world domain and the logistic domain. Results indicate that neural networks are capable of making such
predictions with high accuracy, indicating a promising new framework for approaching generalized planning problems.

Contributors

Created

Date Created
2019-05

136275-Thumbnail Image.png

On the Semantic Equivalence of a Program and Any of its Intermediate Representations

Description

The central goal of this thesis is to develop a practical approach to validating the correctness of SSA forms. Since achieving this goal is very involved for a general program, we restrict our attention to simple programs. In particular, the

The central goal of this thesis is to develop a practical approach to validating the correctness of SSA forms. Since achieving this goal is very involved for a general program, we restrict our attention to simple programs. In particular, the programs we consider are loop-free and are comprised of simple assignments to scalar variables, as well as input and output statements. Even for such a simple program, a full formal treatment would be very involved, extending beyond the scope of an undergraduate honors thesis.

Contributors

Created

Date Created
2015-05

135810-Thumbnail Image.png

Testbed Implementation of the Meta-MAC Protocol

Description

The meta-MAC protocol is a systematic and automatic method to dynamically combine any set of existing Medium Access Control (MAC) protocols into a single higher level MAC protocol. The meta-MAC concept was proposed more than a decade ago, but until

The meta-MAC protocol is a systematic and automatic method to dynamically combine any set of existing Medium Access Control (MAC) protocols into a single higher level MAC protocol. The meta-MAC concept was proposed more than a decade ago, but until now has not been implemented in a testbed environment due to a lack of suitable hardware. This thesis presents a proof-of-concept implementation of the meta-MAC protocol by utilizing a programmable radio platform, the Wireless MAC Processor (WMP), in combination with a host-level software module. The implementation of this host module, and the requirements and challenges faced therein, is the primary subject of this thesis. This implementation can combine, with certain constraints, a set of protocols each represented as an extended finite state machine for easy programmability. To illustrate the combination principle, protocols of the same type but with varying parameters are combined in a testbed environment, in what is termed parameter optimization. Specifically, a set of TDMA protocols with differing slot assignments are experimentally combined. This experiment demonstrates that the meta-MAC implementation rapidly converges to non-conflicting TDMA slot assignments for the nodes, with similar results to those in simulation. This both validates that the presented implementation properly implements the meta-MAC protocol, and verifies that the meta-MAC protocol can be as effective on real wireless hardware as it is in simulation.

Contributors

Created

Date Created
2016-05

152324-Thumbnail Image.png

Incorporating supervisory human inputs into autonomous robot navigation

Description

With robots being used extensively in various areas, a certain degree of robot autonomy has always been found desirable. In applications like planetary exploration, autonomous path planning and navigation are considered essential. But every now and then, a need to

With robots being used extensively in various areas, a certain degree of robot autonomy has always been found desirable. In applications like planetary exploration, autonomous path planning and navigation are considered essential. But every now and then, a need to modify the robot's operation arises, a need for a human to provide it some supervisory parameters that modify the degree of autonomy or allocate extra tasks to the robot. In this regard, this thesis presents an approach to include a provision to accept and incorporate such human inputs and modify the navigation functions of the robot accordingly. Concepts such as applying kinematical constraints while planning paths, traversing of unknown areas with an intent of maximizing field of view, performing complex tasks on command etc. have been examined and implemented. The approaches have been tested in Robot Operating System (ROS), using robots such as the iRobot Create, Personal Robotics (PR2) etc. Simulations and experimental demonstrations have proved that this approach is feasible for solving some of the existing problems and that it certainly can pave way to further research for enhancing functionality.

Contributors

Agent

Created

Date Created
2013

152234-Thumbnail Image.png

Smooth path planning using splines for unmanned planetary vehicles

Description

One of the main challenges in planetary robotics is to traverse the shortest path through a set of waypoints. The shortest distance between any two waypoints is a direct linear traversal. Often times, there are physical restrictions that prevent a

One of the main challenges in planetary robotics is to traverse the shortest path through a set of waypoints. The shortest distance between any two waypoints is a direct linear traversal. Often times, there are physical restrictions that prevent a rover form traversing straight to a waypoint. Thus, knowledge of the terrain is needed prior to traversal. The Digital Terrain Model (DTM) provides information about the terrain along with waypoints for the rover to traverse. However, traversing a set of waypoints linearly is burdensome, as the rovers would constantly need to modify their orientation as they successively approach waypoints. Although there are various solutions to this problem, this research paper proposes the smooth traversability of the rover using splines as a quick and easy implementation to traverse a set of waypoints. In addition, a rover was used to compare the smoothness of the linear traversal along with the spline interpolations. The data collected illustrated that spline traversals had a less rate of change in the velocity over time, indicating that the rover performed smoother than with linear paths.

Contributors

Agent

Created

Date Created
2013

152179-Thumbnail Image.png

Model-based design, simulation and automatic code generation for embedded systems and robotic applications

Description

As the complexity of robotic systems and applications grows rapidly, development of high-performance, easy to use, and fully integrated development environments for those systems is inevitable. Model-Based Design (MBD) of dynamic systems using engineering software such as Simulink® from MathWorks®,

As the complexity of robotic systems and applications grows rapidly, development of high-performance, easy to use, and fully integrated development environments for those systems is inevitable. Model-Based Design (MBD) of dynamic systems using engineering software such as Simulink® from MathWorks®, SciCos from Metalau team and SystemModeler® from Wolfram® is quite popular nowadays. They provide tools for modeling, simulation, verification and in some cases automatic code generation for desktop applications, embedded systems and robots. For real-world implementation of models on the actual hardware, those models should be converted into compilable machine code either manually or automatically. Due to the complexity of robotic systems, manual code translation from model to code is not a feasible optimal solution so we need to move towards automated code generation for such systems. MathWorks® offers code generation facilities called Coder® products for this purpose. However in order to fully exploit the power of model-based design and code generation tools for robotic applications, we need to enhance those software systems by adding and modifying toolboxes, files and other artifacts as well as developing guidelines and procedures. In this thesis, an effort has been made to propose a guideline as well as a Simulink® library, StateFlow® interface API and a C/C++ interface API to complete this toolchain for NAO humanoid robots. Thus the model of the hierarchical control architecture can be easily and properly converted to code and built for implementation.

Contributors

Agent

Created

Date Created
2013

154026-Thumbnail Image.png

Moving obstacle avoidance for unmanned aerial vehicles

Description

There has been a vast increase in applications of Unmanned Aerial Vehicles (UAVs) in civilian domains. To operate in the civilian airspace, a UAV must be able to sense and avoid both static and moving obstacles for flight safety. While

There has been a vast increase in applications of Unmanned Aerial Vehicles (UAVs) in civilian domains. To operate in the civilian airspace, a UAV must be able to sense and avoid both static and moving obstacles for flight safety. While indoor and low-altitude environments are mainly occupied by static obstacles, risks in space of higher altitude primarily come from moving obstacles such as other aircraft or flying vehicles in the airspace. Therefore, the ability to avoid moving obstacles becomes a necessity

for Unmanned Aerial Vehicles.

Towards enabling a UAV to autonomously sense and avoid moving obstacles, this thesis makes the following contributions. Initially, an image-based reactive motion planner is developed for a quadrotor to avoid a fast approaching obstacle. Furthermore, A Dubin’s curve based geometry method is developed as a global path planner for a fixed-wing UAV to avoid collisions with aircraft. The image-based method is unable to produce an optimal path and the geometry method uses a simplified UAV model. To compensate

these two disadvantages, a series of algorithms built upon the Closed-Loop Rapid Exploratory Random Tree are developed as global path planners to generate collision avoidance paths in real time. The algorithms are validated in Software-In-the-Loop (SITL) and Hardware-In-the-Loop (HIL) simulations using a fixed-wing UAV model and in real flight experiments using quadrotors. It is observed that the algorithm enables a UAV to avoid moving obstacles approaching to it with different directions and speeds.

Contributors

Agent

Created

Date Created
2015

154497-Thumbnail Image.png

Development and analysis of stochastic boundary coverage strategies for multi-robot systems

Description

Robotic technology is advancing to the point where it will soon be feasible to deploy massive populations, or swarms, of low-cost autonomous robots to collectively perform tasks over large domains and time scales. Many of these tasks will require

Robotic technology is advancing to the point where it will soon be feasible to deploy massive populations, or swarms, of low-cost autonomous robots to collectively perform tasks over large domains and time scales. Many of these tasks will require the robots to allocate themselves around the boundaries of regions or features of interest and achieve target objectives that derive from their resulting spatial configurations, such as forming a connected communication network or acquiring sensor data around the entire boundary. We refer to this spatial allocation problem as boundary coverage. Possible swarm tasks that will involve boundary coverage include cooperative load manipulation for applications in construction, manufacturing, and disaster response.

In this work, I address the challenges of controlling a swarm of resource-constrained robots to achieve boundary coverage, which I refer to as the problem of stochastic boundary coverage. I first examined an instance of this behavior in the biological phenomenon of group food retrieval by desert ants, and developed a hybrid dynamical system model of this process from experimental data. Subsequently, with the aid of collaborators, I used a continuum abstraction of swarm population dynamics, adapted from a modeling framework used in chemical kinetics, to derive stochastic robot control policies that drive a swarm to target steady-state allocations around multiple boundaries in a way that is robust to environmental variations.

Next, I determined the statistical properties of the random graph that is formed by a group of robots, each with the same capabilities, that have attached to a boundary at random locations. I also computed the probability density functions (pdfs) of the robot positions and inter-robot distances for this case.

I then extended this analysis to cases in which the robots have heterogeneous communication/sensing radii and attach to a boundary according to non-uniform, non-identical pdfs. I proved that these more general coverage strategies generate random graphs whose probability of connectivity is Sharp-P Hard to compute. Finally, I investigated possible approaches to validating our boundary coverage strategies in multi-robot simulations with realistic Wi-fi communication.

Contributors

Agent

Created

Date Created
2016

154349-Thumbnail Image.png

Parallel optimization of polynomials for large-scale problems in stability and control

Description

In this thesis, we focus on some of the NP-hard problems in control theory. Thanks to the converse Lyapunov theory, these problems can often be modeled as optimization over polynomials. To avoid the problem of intractability, we establish a trade

In this thesis, we focus on some of the NP-hard problems in control theory. Thanks to the converse Lyapunov theory, these problems can often be modeled as optimization over polynomials. To avoid the problem of intractability, we establish a trade off between accuracy and complexity. In particular, we develop a sequence of tractable optimization problems - in the form of Linear Programs (LPs) and/or Semi-Definite Programs (SDPs) - whose solutions converge to the exact solution of the NP-hard problem. However, the computational and memory complexity of these LPs and SDPs grow exponentially with the progress of the sequence - meaning that improving the accuracy of the solutions requires solving SDPs with tens of thousands of decision variables and constraints. Setting up and solving such problems is a significant challenge. The existing optimization algorithms and software are only designed to use desktop computers or small cluster computers - machines which do not have sufficient memory for solving such large SDPs. Moreover, the speed-up of these algorithms does not scale beyond dozens of processors. This in fact is the reason we seek parallel algorithms for setting-up and solving large SDPs on large cluster- and/or super-computers.

We propose parallel algorithms for stability analysis of two classes of systems: 1) Linear systems with a large number of uncertain parameters; 2) Nonlinear systems defined by polynomial vector fields. First, we develop a distributed parallel algorithm which applies Polya's and/or Handelman's theorems to some variants of parameter-dependent Lyapunov inequalities with parameters defined over the standard simplex. The result is a sequence of SDPs which possess a block-diagonal structure. We then develop a parallel SDP solver which exploits this structure in order to map the computation, memory and communication to a distributed parallel environment. Numerical tests on a supercomputer demonstrate the ability of the algorithm to efficiently utilize hundreds and potentially thousands of processors, and analyze systems with 100+ dimensional state-space. Furthermore, we extend our algorithms to analyze robust stability over more complicated geometries such as hypercubes and arbitrary convex polytopes. Our algorithms can be readily extended to address a wide variety of problems in control such as Hinfinity synthesis for systems with parametric uncertainty and computing control Lyapunov functions.

Contributors

Agent

Created

Date Created
2016