User Interface Design for Robotic Systems: A Study on Usability and Human-Machine Interaction

Context

The Research Question

As robotics systems move from industrial floors into everyday environments — maintenance, infrastructure, healthcare — the interfaces that humans use to control and monitor them become critical. Poor interface design in high-stakes robotic contexts doesn't just create frustration: it creates risk.

This research explored GUI design for service robotics systems, focusing on usability and human-machine interaction in everyday contexts. The central question: what design guidelines produce interfaces that are simultaneously learnable for novice operators and efficient for experienced ones?

Method

Research Approach

The research was conducted in two main contexts: EMMA (a robotic system for turbine maintenance at 13 Robotics) and ROSA (a hydroelectric plant monitoring system at LEAD Coppe/UFRJ). Both involved real operational environments with non-expert users.

• Task analysis: mapping operator workflows to identify cognitive load peaks and decision points
• Usability studies: structured lab testing with operators across skill levels
• Field research: observation in operational environments to capture real usage patterns vs. designed usage patterns
• Quantitative methods: surveys and A/B testing to validate design decisions

Findings

Design Guidelines for HMI

The research produced a set of design guidelines for robotic system interfaces, organized around three principles:

• Transparency of system state: operators need unambiguous feedback about what the robot is doing, why, and what it will do next. Ambiguity in state feedback was the leading cause of operator errors.
• Graceful degradation: interfaces must communicate failure states clearly without triggering panic responses. Alert design required careful calibration of urgency cues.
• Minimal mode switching: complex robotic systems often require operators to switch between monitoring and control modes. Reducing mode transitions significantly improved task completion time and error rate.

Framework

The Supervisory Control Paradigm

The theoretical backbone of this research was the Supervisory Control paradigm, part of Parasuraman and Sheridan's Human-Automation Interaction (HAI) model. In it, the human stops being the task executor and becomes a strategist: planning, teaching the system, monitoring execution, intervening when needed, and learning from the outcome. The machine runs the operation; the person stays in control of the decisions.

The operator's cognitive cycle was modeled through the OODA Loop (Observe, Orient, Decide, Act), and supervision quality measured across three human factors: situational awareness, mental workload, and trust in automation. These guidelines were applied directly to EMMA's operator interface and validated through subsequent usability rounds.

Today

From HMI to Agentic Flows

The same framework that structured the supervision of industrial robots describes, almost without translation, the design of agentic flows in AI products. An agent is an autonomous system that runs the operational steps and delegates the strategist role to the human: the supervisory control paradigm, now applied to software.

The guidelines I tested between 2016 and 2019 are the same decisions that define a good agentic flow today:

• Transparency of system state → making visible what the agent is doing, why, and what it will do next, avoiding the black box that breaks trust.
• Intervention points (human-in-the-loop) → designing where the person approves, corrects, or takes control before an irreversible action.
• Graceful degradation → communicating the agent's failures and uncertainty with calibrated clarity, without alarm or false confidence.
• Calibrated trust → matching the level of autonomy to the task's risk, one of the core human factors of the original research.

It is this rigor — task analysis, cognitive load assessment, and the vocabulary of supervisory control — that I bring today to designing AI products and high-stakes agentic experiences in fintech and crypto.