Robotics Commons^™

Neuron Clustering For Mitigating Catastrophic Forgetting In Supervised And Reinforcement Learning, Benjamin Frederick Goodrich

Doctoral Dissertations

Neural networks have had many great successes in recent years, particularly with the advent of deep learning and many novel training techniques. One issue that has affected neural networks and prevented them from performing well in more realistic online environments is that of catastrophic forgetting. Catastrophic forgetting affects supervised learning systems when input samples are temporally correlated or are non-stationary. However, most real-world problems are non-stationary in nature, resulting in prolonged periods of time separating inputs drawn from different regions of the input space.

Reinforcement learning represents a worst-case scenario when it comes to precipitating catastrophic forgetting in neural networks. …

Using Eye And Head Movements As A Control Mechanism For Tele-Operating A Ground-Based Robot And Its Payload, Kathryn C. Hicks

Computational Modeling & Simulation Engineering Theses & Dissertations

To date, eye and head tracking has been used to indicate users' attention patterns while performing a task or as an aid for disabled persons, to allow hands-free interaction with a computer. The increasing accuracy and the reduced cost of eye- and head-tracking equipment make utilizing this technology feasible for explicit control tasks, especially in cases where there is confluence between the visual task and control.

The goal of this research was to investigate the use of eye-tracking as a more natural interface for the control of a camera-equipped, remotely operated robot in tasks that require the operator to simultaneously …

An Intelligent Attitude Determination And Control System Concept For A Cubesat Class Spacecraft, Jeremy Straub

Jeremy Straub

An attitude determination and control system (ADCS) is used to orient a spacecraft for a wide variety of purposes (e.g., to keep a camera facing Earth or orient the spacecraft for propulsion system use). The proposed intelligent ADCS has several key features: first, it can be used in multiple modes, spanning from passive stabilization of two axes and unconstrained spin on a third to three-axis full active stabilization. It also includes electromagnetic components to ‘dump’ spin from the reaction wheels. Second, the ADCS utilizes an incorporated autonomous control algorithm to characterize the effect of actuation of the system components and, …

An Onboard Distributed Multiprocessing System For A Cubesat Spacecraft Created From Gumstix Computer-On-Module Units, Michael Wegerson, Jeremy Straub, Ronald Marsh

Jeremy Straub

The OpenOrbiter Small Spacecraft Development Initiative at the University of North Dakota [1] aims to make ac-cess to space for research and educational purposes easier by enabling the creation of low-cost CubeSats. It is creating the Open Prototype for Educational Nanosats (OPEN), a framework for developing a 1-U CubeSat space-craft with a parts cost of less than $5,000 [2]. The designs [3], documentation and computer code from this will be made publically available to enable the development of programs at other institutions.

Considering Scheduling Algorithms For An Open Source Software Spacecraft, Calvin Bina, Jeremy Straub, Ronald Marsh

Jeremy Straub

The OpenOrbiter Small Satellite Development Initiative at the University of North Dakota [1] is working make space research and education more accessible world-wide [2], through the design and public release of a complete set of plans, software and other documents (see [3]) for a 1-U CubeSat. This design targets a parts cost of no more than $5,000 [4]. These lowered costs, combined with the efficiencies of the CubeSat form fac-tor [5] and free-to-qualified-developer launch services [6, 7] should facilitate greater access to space for the ed-ucational, research and other communities.

Learning Parameterized Skills, Bruno Castro Da Silva

Doctoral Dissertations

One of the defining characteristics of human intelligence is the ability to acquire and refine skills. Skills are behaviors for solving problems that an agent encounters often—sometimes in different contexts and situations—throughout its lifetime. Identifying important problems that recur and retaining their solutions as skills allows agents to more rapidly solve novel problems by adjusting and combining their existing skills. In this thesis we introduce a general framework for learning reusable parameterized skills. Reusable skills are parameterized procedures that—given a description of a problem to be solved—produce appropriate behaviors or policies. They can be sequentially and hierarchically combined with other …

Swarm Intelligence, A Blackboard Architecture And Local Decision Making For Spacecraft Command, Jeremy Straub

Jeremy Straub

Control of a multi-spacecraft constellation is a topic of significant inquiry, at present. This paper presents and evaluates a command architecture for a multi-spacecraft mission. It combines swarm techniques with a decentralized / local decision making architecture (which uses a set of shared blackboards for coordination) and demonstrates the efficacy of this approach. Under this approach, the Blackboard software architecture is used to facilitate data sharing between craft as part of a resilient hierarchy and the swarm techniques are used to coordinate activity. The paper begins with an overview of prior work on the precursor command technologies and then presents …