Design and Evaluation of an Interactive Micro Stage Control System Based on ESP32 S3 for AI Theatre Applications
TL;DR
The results show that simple filtering and deadzone strategies can support stable and predictable control behaviour in a low-cost embedded prototype that can be extended with mobile control, speech recognition, gesture interaction, rhythm-aware lighting, script-driven automation, or other intelligent interaction functions.
Abstract
Aims: This study aimed to design and evaluate a low-cost ESP32-S3-based interactive micro stage control system for future AI Theatre applications. The objective was to develop a compact embedded prototype that could support stage demonstrations, theatre education, interactive exhibitions, and later AI-assisted performance interaction. Study Design: An applied engineering prototype design was used. The study combined physical prototype implementation, firmware development, functional testing, and software-based simulation evaluation. Place and Duration of Study: The prototype design, hardware assembly, firmware development, simulation, and manuscript preparation were conducted collaboratively by authors affiliated with Youngsan University, Republic of Korea, and contributors in China during 2026. Methodology: The physical prototype integrated a dual-axis analogue joystick, an SG90 servo motor, a DC motor with a driver, and an I2C LCD1602 display. The joystick X-axis controlled the servo angle to represent stage lighting rotation, while the Y-axis controlled motor direction and speed to represent curtain movement. The firmware applied one dummy analogue reading, a short ADC stabilisation delay, 15-sample averaging, and wide deadzone mapping. The prototype was assembled and functionally tested in a classroom setting. A synthetic software simulation was then used to evaluate analogue input filtering and input-to-actuator mapping under controlled conditions. Results: In the synthetic ADC simulation, the error standard deviation decreased from 85.80 to 16.66 after the 15-sample averaging filter was applied. This corresponded to an estimated reduction of 80.58% under the simulated noise conditions. Deadzone mapping maintained neutral actuator behaviour by keeping the servo at 90 degrees and the motor stopped when joystick input remained within the predefined neutral ranges. The results show that simple filtering and deadzone strategies can support stable and predictable control behaviour in a low-cost embedded prototype. Conclusion: The ESP32-S3-based prototype provides a practical embedded control foundation for interactive micro stage systems and future AI Theatre applications. The current system does not implement autonomous AI decision-making and has not been validated in a real theatre environment. Its main contribution is a stable and reproducible physical control layer that can later be extended with mobile control, speech recognition, gesture interaction, rhythm-aware lighting, script-driven automation, or other intelligent interaction functions.
