Robô humanoide de membros inferiores proprioceptivos com controle adaptativo em inteligência artificial

Abstract

This thesis presents the design and development of a humanoid robot’s lower limbs with adaptive control performed through proprioception, implemented using limbs made from intelligent composite materials. The project details the development of two feet, two tibias, two fêmures, and a hip, along with their respective joints. The limbs are structured using fiber optic sensors based on Fiber Bragg Gratings (FBGs) embedded in Carbon Fiber–Reinforced Polymer (CFRP). Due to the FBGs’ dimensions, approximately 125 µm in diameter and 3 to 5 mm in length, along with the multiplexed sensor arrangement and the strong adhesion with the composite material, the development of intelligent limbs sensitive to deformation was enabled. To fabricate the CFRP limbs, 3D-printed parts inspired by human anatomy were used as both molds for manufacturing and for incorporating the internal structure of the bone-like limbs. Subsequently, FBG sensors were integrated to monitor the forces acting on the humanoid robot’s limbs. The instrumentation in the feet includes the monitoring of forces on the calcaneus, metatarsal bones, and proximal and middle phalanges. In the tibias, it covers the area around the nutrient foramen, and in the fêmures, the region of the linea aspera. The FBG instrumentation method tracks these forces by measuring the composite’s deformation, which results in changes in the Bragg wavelength. These signals provide proprioceptive feedback for decision-making in controlling the movements of the robotic limbs. A supervisory system monitors the forces on each of the intelligent limbs using a color-coded map, as well as a video monitoring camera and a configured position simulator. As part of the actuator structure, servo motors were developed for the bilateral joints of the ankle, knee, and hip. A hybrid adaptive controller is presented, consisting of a set of Fuzzy Proportional-Integral-Derivative (FPID) controllers to execute the necessary movements of the humanoid lower limbs. A proprioceptive abstraction layer with fuzzy logic is implemented due to its suitability for abstract applications, adjusting the servo motor gains and the speed of planned movements according to the perceived effort on the robot’s limbs. Experimental results demonstrate that the use of intelligent limbs and robotic proprioception leads to performance improvements when employing the fuzzy logic controller and a centralized movement planner.