Efficient multitask learning with an embodied predictive model for door opening and entry with whole-body control

Robots need robust models to effectively perform tasks that humans do on a daily basis. These models often require substantial developmental costs to maintain because they need to be adjusted and adapted over time. Deep reinforcement learning is a powerful approach for acquiring complex real-world models because there is no need for a human to design the model manually. Furthermore, a robot can establish new motions and optimal trajectories that may not have been considered by a human. However, the cost of learning is an issue because it requires a huge amount of trial and error in the real world. Here, we report a method for realizing complicated tasks in the real world with low design and teaching costs based on the principle of prediction error minimization. We devised a module integration method by introducing a mechanism that switches modules based on the prediction error of multiple modules. The robot generates appropriate motions according to the door's position, color, and pattern with a low teaching cost. We also show that by calculating the prediction error of each module in real time, it is possible to execute a sequence of tasks (opening door outward and passing through) by linking multiple modules and responding to sudden changes in the situation and operating procedures. The experimental results show that the method is effective at enabling a robot to operate autonomously in the real world in response to changes in the environment.

Imitative Neural Mechanism-Based Behavior Intention Recognition System in Human–Robot Interaction

This paper is concerned with an imitative neural mechanism for recognizing behavior intention in human–robot interaction system. The intention recognition process is inspired by the neural mechanism of the mirror neurons in macaque monkey brain. We try to renovate a standard neural network with parametric biases as a reference model to imitate between sensory-motor data pair. The imitation process is primarily directed toward reproducing the goals of observed actions rather than the exact action trajectories. Several experiments and their results show that the proposed model allows to develop useful robotic application for human–robot interaction system application.