Abstract
This paper introduces a novel solution to the manual control challenge for indoor blimps. The problem’s complexity arises from the conflicting demands of executing human commands while maintaining stability through automatic control for underactuated robots. To tackle this challenge, we introduced an assisted piloting hybrid controller with a preemptive mechanism, that seamlessly switches between executing human commands and activating automatic stabilization control. Our algorithm ensures that the automatic stabilization controller operates within the time delay between human observation and perception, providing assistance to the driver in a way that remains imperceptible.
Introduction
Indoor miniature blimps represent a safe, cost-effective platform for applications such as indoor surveillance and environmental monitoring. However, their physical characteristics make them notoriously difficult to pilot manually. Most indoor blimps are underactuated, meaning they possess fewer independent control inputs than degrees of freedom. This physical limitation leads to significant "swing" oscillations; for instance, a command for forward motion often induces unwanted pitch or side-to-side sway.
While automatic stabilization is a common solution, it often introduces a psychological conflict for the human operator. If an autopilot system constantly applies high-frequency corrections that oppose the pilot's inputs, the pilot's "sense of agency" — the feeling of being the cause of the robot's actions — is diminished. According to Wegner’s principles of human-computer interaction, a loss of agency can lead to pilot frustration, overcorrection, and eventual system failure.
Methodology: The Preemptive Mechanism
To resolve the tension between automatic stability and human agency, researchers at the Hong Kong University of Science and Technology (HKUST) developed a hybrid controller architecture. The core innovation is a preemptive mechanism that exploits the inherent latency in human perception.
The system integrates three primary components:
- Balancing Controller: A PID-based controller that manages the blimp's pitch angle to ensure stable horizontal flight during commanded movements.
- Stabilizing Controller: A "bang-bang" control strategy designed to rapidly nullify rotational velocity ($\omega_z$), returning the blimp to a stable idle state.
- The Preemptive Mechanism: Neuroscience indicates a delay of approximately 200ms between a human observing a physical event and consciously perceiving it. The hybrid controller divides this 200ms window into two distinct 100ms slots. In the first slot, the balancing controller executes the human's command. In the second slot, the stabilizing controller activates to dampen residual oscillations.
By interleaving stabilization within the pilot's perceptual "blind spot," the assistance remains imperceptible. The pilot perceives the blimp as naturally stable and responsive, maintaining a high sense of agency.
Experimental Setup
The proposed system was validated through live experimental trials involving three human participants. The researchers utilized an Opti-track motion capture system to provide high-precision ground truth data regarding the blimp's position and angular velocity. This setup allowed for a rigorous comparison between unassisted manual flight and the proposed human-assistive piloting mode.
Results and Discussion
The effectiveness of the hybrid controller was evaluated based on its ability to suppress unwanted oscillations during a target-ring navigation task.
- Swing Reduction: Experimental data demonstrated that the assisted blimp maintained an angular velocity ($\omega_z$) near 0 rad/s. Conversely, unassisted flights showed significant, cumulative oscillations that made precise maneuvering difficult.
- Sway Suppression: The controller successfully reduced the frequency and amplitude of "sway" along the flight path, resulting in smoother, more predictable trajectories.
- Preservation of Agency: Qualitative feedback indicated that the hybrid design successfully decoupled complex motions (such as the pitch-up tendency during acceleration) without the pilot feeling the "tug-of-war" typical of standard autopilots.
Conclusion
This research demonstrates that human perceptual limits can be leveraged to improve robot control. By treating human reaction time as a window for background stabilization, the hybrid controller provides necessary assistance without compromising the operator's sense of control.
Future work will focus on transitioning from external motion capture to fully on-board sensing and testing the scalability of the 200ms timing window across various blimp sizes and mass properties. This approach paves the way for more intuitive human-machine interfaces in complex, underactuated robotic systems.
Summary updated