soconnor/honors-thesis
1
0
Fork 0
mirror of https://github.com/soconnor0919/honors-thesis.git synced 2026-05-08 07:08:55 -04:00
Code Issues Actions Packages Projects Releases 1 Activity

Refactor implementation and evaluation chapters for clarity and detail

Browse Source 
- Revised the implementation chapter to emphasize HRIStudio as a reference implementation of design principles, detailing architectural choices and mechanisms.
- Enhanced descriptions of platform architecture, experiment storage, execution engine, and access control.
- Updated evaluation chapter to reflect the study as a pilot validation study, clarifying research questions, study design, participant roles, and measures.
- Improved consistency in language and structure throughout both chapters.
- Added details on participant recruitment and task specifications to better contextualize the study.
- Adjusted measurement instruments table to align with the new chapter title.
- Updated LaTeX document to include additional TikZ library for improved diagram capabilities.
This commit is contained in:
Sean O'Connor
2026-03-05 23:28:59 -05:00
parent 4d960b0ca9
commit 7757046eec
7 changed files with 160 additions and 117 deletions
Download Patch File Download Diff File Expand all files Collapse all files
thesis/chapters/05_implementation.tex
+34 -21
View File
@@ -1,19 +1,21 @@
\chapter{Implementation}
\label{ch:implementation}
This chapter explains how HRIStudio implements the design from Chapter~\ref{ch:design}. It covers the architectural choices and mechanisms behind how the platform stores experiments, executes trials, integrates robot hardware, and controls access. Technology stack specifics are in Appendix~\ref{app:tech_docs}.
HRIStudio is a reference implementation of the design principles established in Chapter~\ref{ch:design}. The central contribution of this work is not the tool itself but the design concepts that underpin it: the hierarchical specification model, the event-driven execution model, and the integrated data flow. Any system built on those concepts would satisfy the same requirements. This chapter explains how HRIStudio realizes them, covering the architectural choices and mechanisms behind how the platform stores experiments, executes trials, integrates robot hardware, and controls access. Technology stack specifics are presented in Appendix~\ref{app:tech_docs}.
\section{Platform Architecture}
HRIStudio runs as a web application. Researchers access it through a standard browser without installing specialized software, and the entire study team, including researchers, wizards, and observers, connect to the same shared system. This eliminates installation complexity and ensures the platform works identically on any operating system, directly addressing the low-technical-barrier requirement (R2, from Chapter~\ref{ch:background}). It also enables natural collaboration (R6): multiple team members can access experiment data and observe live trials simultaneously from different machines without any additional configuration.
HRIStudio follows the model of a web application. Researchers access it through a standard browser without installing specialized software, and the entire study team, including researchers, wizards, and observers, connect to the same shared system. This eliminates the need for a local installation and ensures the platform works identically on any operating system, directly addressing the low-technical-barrier requirement (R2, from Chapter~\ref{ch:background}). It also enables easy collaboration (R6): multiple team members can access experiment data and observe live trials simultaneously from different machines without any additional configuration.
The system is organized into three layers: a browser-based user interface, an application server that manages execution, authentication, and logging, and a data and robot control layer covering storage and hardware communication. These layers are described architecturally in Chapter~\ref{ch:design}; what matters for implementation is that the server runs on the same local network as the robot hardware. This keeps communication latency low during live trials, where a delay between the wizard's input and the robot's response would disrupt the interaction. All three layers are implemented in the same language--TypeScript \cite{TypeScript2014}, a statically-typed superset of JavaScript. When the structure of experiment data changes, the type checker surfaces inconsistencies across the entire codebase at compile time rather than allowing them to surface as runtime failures during a live trial.
I organized the system into three layers: a browser-based user interface, an application server that manages execution, authentication, and logging, and a data and robot control layer covering storage and hardware communication. This layered structure is shown in Figure~\ref{fig:three-tier}. A key deployment constraint is that the application server runs on the same local network as the robot hardware. This keeps communication latency low during live trials: a noticeable delay between the wizard's input and the robot's response would break the interaction.
I implemented all three layers in the same language — TypeScript~\cite{TypeScript2014}, a statically-typed superset of JavaScript. The single-language decision keeps the type system consistent across the full stack. When the structure of experiment data changes, the type checker surfaces inconsistencies across the entire codebase at compile time rather than allowing them to appear as runtime failures during a live trial.
\section{Experiment Storage and Trial Logging}
Experiments are saved to persistent storage when a researcher completes them in the Design interface. A saved experiment is a complete, reusable specification that can be run across any number of trials without modification.
The system saves experiments to persistent storage when a researcher completes them in the Design interface. A saved experiment is a complete, reusable specification that a researcher can run across any number of trials without modification.
When a trial begins, the system creates a new trial record linked to that experiment. Every action the wizard triggers during the trial is written to that record with a precise timestamp, whether it was scripted or not. Video, audio, and robot sensor data are recorded alongside the action log for the duration of the trial. Unscripted actions are flagged as deviations. Because the trial record and the experiment reference the same underlying specification, the Analysis interface can directly compare what was planned against what was executed for any trial, without any manual work by the researcher. Figure~\ref{fig:trial-record} shows the structure of a completed trial record.
When a trial begins, the system creates a new trial record linked to that experiment. The system writes every action the wizard triggers to that record with a precise timestamp, whether scripted or not, including any unscripted actions triggered outside the protocol. The system flags those unscripted actions as deviations. The browser records video, audio, and robot sensor data alongside the action log for the duration of the trial. The Analysis interface can directly compare what was planned against what was executed for any trial, without any manual work by the researcher, because the trial record and the experiment reference the same underlying specification. Figure~\ref{fig:trial-record} shows the structure of a completed trial record: action log entries, video, audio, and robot sensor data all share a common timestamp reference so the Analysis interface can align them without manual synchronization; dashed lines mark step boundaries; and the system flags any deviation from the experiment specification inline.
\begin{figure}[htbp]
\centering
@@ -79,27 +81,29 @@ When a trial begins, the system creates a new trial record linked to that experi
};
\end{tikzpicture}
\caption{Structure of a completed trial record. Action log entries, video, audio, and robot sensor data share a common timestamp reference so the Analysis interface can align them without manual synchronization. Deviations from the experiment specification are flagged inline. Dashed lines mark step boundaries.}
\caption{Structure of a completed trial record, showing synchronized action log, media, and sensor tracks.}
\label{fig:trial-record}
\end{figure}
Video and audio are recorded locally in the researcher's browser during the trial rather than streamed to the server in real time. This prevents network delays or server load from dropping frames or degrading audio quality during the interaction. When the trial concludes, the browser transfers the complete recordings to the server and associates them with the trial record. Because the timestamp when recording starts is logged alongside the action log, the Analysis interface can align video and audio with the logged actions without any manual synchronization.
Video and audio are recorded locally in the researcher's browser during the trial rather than streamed to the server in real time. This prevents network delays or server load from dropping frames or degrading audio quality during the interaction. When the trial concludes, the browser transfers the complete recordings to the server and associates them with the trial record. The Analysis interface can align video and audio with the logged actions without any manual synchronization, because the timestamp when recording starts is logged alongside the action log.
This reflects a deliberate split in how data is stored. Experiment specifications and trial records are kept in a structured database, which makes it efficient to query across trials, for example retrieving all trials for a specific participant or comparing action timing across conditions. Video and audio files are stored separately in a dedicated file store, since their size makes them unsuitable for a database but their content is not queried directly.
The system stores structured and media data separately. Experiment specifications and trial records live in a structured database, which makes it efficient to query across trials (for example, retrieving all trials for a specific participant or comparing action timing across conditions). Video and audio files live in a dedicated file store, since their size makes them unsuitable for a database and the system never queries their content directly.
\section{The Execution Engine}
When a trial begins, the server loads the experiment and maintains a live connection to the wizard's browser and any observer connections. The execution engine does not advance the experiment on a timer; it waits for the wizard to trigger each step. This preserves the natural pacing of the interaction: the wizard advances only when the participant is ready, while the experiment structure ensures the protocol is followed. When the wizard triggers an action, the server dispatches the robot command, writes the log entry, and pushes the updated experiment state to all connected clients in the same operation. This is what keeps the wizard's view, the observer view, and the actual robot state synchronized in real time.
The execution engine is the component that runs a live trial: it loads the experiment, manages the wizard's connection, dispatches robot commands, and keeps all connected clients in sync.
Unscripted actions go through the same path. The wizard triggers them via the manual controls in the Execution interface, the robot command runs, and the action is logged with a deviation flag. The result is a complete, unambiguous trial record regardless of how closely the interaction followed the script.
When a trial begins, the server loads the experiment and maintains a live connection to the wizard's browser and any observer connections. The execution engine does not advance the experiment on a timer; it waits for the wizard to trigger each step. This preserves the natural pacing of the interaction: the wizard advances only when the participant is ready, while the experiment structure ensures the protocol is followed. When the wizard triggers an action, the server dispatches the robot command, writes the log entry, and pushes the updated experiment state to all connected clients in the same operation — keeping the wizard's view, the observer view, and the actual robot state synchronized in real time.
No two participants respond identically. One subject gives a one-word answer; another offers a paragraph; a third asks the robot a question the script never anticipated. A fully programmed robot has no answer for that third subject: the interaction stalls, or immersion breaks. The wizard exists to fill that gap: where the program runs out of instructions, the wizard draws on their knowledge of human social interaction to keep the exchange coherent. Unscripted actions give the wizard the tools to exercise that judgment in the moment. The wizard triggers them via the manual controls in the Execution interface, the robot command runs, and the system logs the action with a deviation flag. This design preserves research value: the interaction gains the flexibility only a human can provide, and that flexibility appears explicitly in the record rather than disappearing into it.
\section{Robot Integration}
Each robot platform is described by a configuration file that lists the actions it supports and specifies how each one maps to a command the robot understands. The execution engine reads this file at startup and uses it whenever it needs to dispatch a command: it looks up the action type, assembles the appropriate message, and sends it to the robot over a bridge process running on the local network. The web server itself has no knowledge of any specific robot; all hardware-specific logic lives in the configuration file.
A configuration file describes each robot platform, listing the actions it supports and specifying how each one maps to a command the robot understands. The execution engine reads this file at startup and uses it whenever it needs to dispatch a command: it looks up the action type, assembles the appropriate message, and sends it to the robot over a bridge process running on the local network. The web server itself has no knowledge of any specific robot; all hardware-specific logic lives in the configuration file.
Control flow elements such as branches and conditionals are treated the same way as robot actions. They appear as action groups in the experiment and are resolved by the execution engine at runtime, so researchers can freely mix logical decisions and physical robot behaviors when designing an experiment without any special handling.
The execution engine treats control flow elements such as branches and conditionals the same way as robot actions. They appear as action groups in the experiment and resolve at runtime, so researchers can freely mix logical decisions and physical robot behaviors when designing an experiment without any special handling.
Figure~\ref{fig:plugin-architecture} illustrates how the same abstract actions map to different robot-specific commands through each platform's configuration, using NAO6 and TurtleBot as an example.
Figure~\ref{fig:plugin-architecture} illustrates this mapping using NAO6 and TurtleBot as an example. Actions a platform does not support (such as \texttt{raise\_arm} on TurtleBot) appear as explicitly unsupported in the configuration file rather than silently failing. The experiment itself does not change between platforms.
\begin{figure}[htbp]
\centering
@@ -142,30 +146,39 @@ Figure~\ref{fig:plugin-architecture} illustrates how the same abstract actions m
\draw[arrow] (cfg.east) -- (tb.west);
\end{tikzpicture}
\caption{The same abstract actions in an experiment are translated to platform-specific robot commands through each robot's configuration file. Actions a platform does not support are declared explicitly rather than silently failing. The experiment itself does not change between platforms.}
\caption{Abstract experiment actions translated to platform-specific robot commands through per-platform configuration files.}
\label{fig:plugin-architecture}
\end{figure}
\section{Access Control}
Each study has a membership list with assigned roles: owner, researcher, wizard, and observer. These roles determine what each team member can see and do within that study. A wizard can trigger actions during a live trial; observers can watch and annotate but cannot trigger anything. This allows studies to separate the wizard's role from the research team's observing role without any additional configuration.
I implemented access control using a role-based access control (RBAC) model. Each study has a membership list, and every member is assigned one of four roles that define a clear separation of duties: those who own the study, those who design it, those who run it, and those who observe it. This enforces need-to-know access at the study level so that each team member sees only what their role requires.
The role system also supports double-blind designs, where certain team members are restricted from seeing condition assignments or result data until the study concludes.
\begin{description}
\item[Owner.] Full control over the study: can invite or remove members, configure the study settings, and access all data.
\item[Researcher.] Can create and modify experiment designs and review all collected trial data, but cannot manage team membership.
\item[Wizard.] Can trigger actions during a live trial and view the execution interface, but cannot modify the experiment design or access other wizards' sessions.
\item[Observer.] Read-only access: can watch a live trial in real time and annotate significant moments, but cannot trigger actions or modify any data.
\end{description}
The role system also supports double-blind designs~\cite{Bartneck2024}: the Owner can restrict a Wizard's view of condition assignments, and restrict Researchers from accessing result data until the study concludes, without any changes to the underlying experiment.
\section{Architectural Challenges}
Two design problems required specific design choices during implementation.
Two problems required specific solutions during implementation.
During a live trial, the execution engine must respond quickly to wizard input. A noticeable delay between the button press and the robot's action can disrupt the interaction. The engine addresses this by maintaining a persistent connection for the duration of each trial. The connection is established once at trial start and held open, so there is no per-action setup overhead.
\begin{description}
\item[Execution latency.] During a live trial, the execution engine must respond quickly to wizard input — a noticeable delay between the button press and the robot's action can disrupt the interaction. I addressed this by maintaining a persistent connection for the duration of each trial. The connection is established once at trial start and kept open, eliminating per-action setup overhead.
Multi-source synchronization requires aligning data streams during analysis that were captured at different sampling rates by different components: video, audio, action logs, and sensor data. The solution is a shared time reference. Every data source records its timestamps relative to the same trial start time, $t_0$, so the Analysis interface can align all tracks without requiring manual calibration. This is the timestamp structure shown in Figure~\ref{fig:trial-record}.
\item[Multi-source synchronization.] Analysis requires aligning data streams captured at different sampling rates by different components: video, audio, action logs, and sensor data. The solution is a shared time reference: every data source records its timestamps relative to the same trial start time, $t_0$, so the Analysis interface can align all tracks without requiring manual calibration. This is the timestamp structure shown in Figure~\ref{fig:trial-record}.
\end{description}
\section{Implementation Status}
HRIStudio has reached minimum viable product status. The Design, Execution, and Analysis interfaces are operational. The execution engine handles scripted and unscripted actions with full timestamped logging, and robot communication has been validated with the NAO6 platform. The platform is capable of running a controlled WoZ study without modification.
HRIStudio has reached minimum viable product status. The Design, Execution, and Analysis interfaces are operational. The execution engine handles scripted and unscripted actions with full timestamped logging, and I validated robot communication on the NAO6 platform during development. The platform can run a controlled WoZ study without modification.
Work remaining for future development includes support for studies that use more than one robot at a time and validation of the configuration file approach on robot platforms beyond NAO6.
\section{Chapter Summary}
This chapter described how HRIStudio's design is realized in practice. Experiments are persistent, reusable specifications that produce complete, comparable trial records. The execution engine is event-driven rather than timer-driven, keeping the wizard in control of pacing while automatically logging every action. Robot hardware integration is handled through per-platform configuration files, keeping the execution engine itself hardware-agnostic. Access control is enforced at the study level through assigned roles. The platform is at minimum viable product status and is capable of running a controlled WoZ study.
This chapter described how HRIStudio realizes the design concepts from Chapter~\ref{ch:design} in practice. Experiments are persistent, reusable specifications that produce complete, comparable trial records. The execution engine is event-driven rather than timer-driven, keeping the wizard in control of pacing while logging every action automatically. Per-platform configuration files keep the execution engine hardware-agnostic. The role system enforces access control at the study level. The platform is at minimum viable product status and can run a controlled WoZ study today. HRIStudio is one realization of these concepts; the contribution lies in the design principles themselves, which any implementation could adopt.