Modelagem e Controle Neuromimético Aplicados a Garra Robótica Inteligente
Modelagem e Controle Neuromimético Aplicados a Garra Robótica Inteligente
-
DOI: https://doi.org/10.22533/at.ed.3941226080112
-
Palavras-chave: Controlador, Escorregamento, Modelo de Izhikevich, Sensores de pressão.
-
Keywords: Controller, Slip, Izhikevich Model, Pressure Sensors.
-
Abstract: For a robotic prosthesis to hold objects firmly, it must sense when slipping begins and react quickly. In this work, we developed a system inspired by the human nervous system to detect and correct slip. To do so, we combined a pressure sensor array (4x4 taxels) with a model that simulates neuronal behavior, known as the Izhikevich model. In practice, pressure variations at the robotic fingertip are transformed into electrical pulses, as if they were neural signals. These signals feed a controller (Monotonic Proportional-Integral) that continuously adjusts grip force to prevent object drop. The entire system was implemented directly on a simple microcontroller (ARM Cortex-M3), operating at 500 Hz. In the tests, we induced sudden slips using a 20 g weight. The robotic hand was able to react and stabilize the object in an average of only 68 ms, allowing a minimal slip of 2.8 ± 0.7 mm. Our results show that this approach not only outperforms traditional methods but also matches the speed and accuracy of much more complex techniques, with the major advantage of requiring far less processing and energy. This demonstrates that brain-inspired processing is a highly efficient path for creating robots capable of intelligently manipulating objects in the real world.
- Vinicius Teixeira da Costa
- Alcimar Barbosa Soares