Robots, Autonomous Vehicles and Deep Learning

Soft actuators have been getting increased attention with developing of medical fields etc. A miniature pneumatic bending rubber actuator is one of the soft actuators. The actuator has the bellows shape and are made of silicone rubbers. Due to the bellows shape, the actuator can do two-way large bending by supplying positive or negative air-pressure. However, to control the actuator and make its model accurately are difficult because the actuator has nonlinearity. Moreover, the actuator should be controlled without sensor because its expected application are medical fields, especially, in operations. On the other hand, a control system based on operator theory can apply nonlinear systems with uncertainties. The relationship between operator theory and passivity or adaptive control which is an important idea in control engineering has discussed by some researchers. Meanwhile, support vector regression (SVR) has been utilized for classification and regression analysis, where the design parameters are selected by using particle swarm optimization (PSO). Therefore, operator-based control system is discussed. In order to realize sensorless control, PSO-SVR-based moving estimation with generalized Gaussian distribution (GGD) kernel is employed. That is, operator-based sensorless adaptive nonlinear control system considering passivity for the actuator and PSO-SVR-based moving estimation with GGD kernel are shown. Finally, some simulations and experimental results are introduced.