Hazard prediction is an important element for intelligent robotic transporters to detect potential hazards like road roughness, drivability, and positive/negative obstacles from features obtained by sensor measurements. Analysis results by means of variable importance are presented for a hazard prediction model learned by random forests. Mean decrease accuracy (MDA) provides a quantitative feature importance estimation that explains which features are influential for the prediction model to make predictions. Partial dependence plot provides a qualitative explanation about how values of an important feature are used by the model. A data-driven feature selection method to find a threshold of important features by exploiting MDA is introduced. Those properties give an insight into the domain knowledge learned by the hazard prediction model as well as a reason why a prediction is returned. Explanation of inside mechanism of intelligent robotic systems is a key factor for the acceptance by societies.
In Human-Robot Communication, emotional expressions of robots almost resembled people and animals. However, a robot should have an ideal form and action according to its purpose, and it is not necessarily a rational design policy to resemble human beings and animals. Therefore, a promotion of communication by methods other than resembling humans and animals should be considered, but little research has been done from such a viewpoint. In this research, we considered a method to express emotion with a robot that is far away physically from a person. We used a flying robot as a robot physically far from the person. Instead of linking actions and emotions directly, we adopted a method of indirectly linking actions and emotions by placing indexes for emotional expression, modeling the relationship between actions and indexes, emotions and indexes. Using the Russell annulus model as indexes, we created motion using the indexes and confirmed that the motion created makes the person feel emotion by a subject experiment.
This paper presents a map-merging method which enables the robot to continue working for months and years. The proposed method merges a reference map and the current map to create a new reference map for the next use. Consistency check for map merging is performed using pseudo χ2 test. After map merging, the method reduces the map size by removing redundant data in the pose graph and point cloud for long-term map maintenance. Experiments show the proposed method successfully merges maps and reduces the map size.
This paper proposes an integration of two balancing controllers; so called, Hip Strategy and Ankle Strategy for humanoid model. The former achieves the balance equilibrium utilizing the reaction moment from the high-speed upper body motion. The latter also achieves the balancing, but does so in the way that won't interrupt the hip strategy. The controller integration is done by computing the optimal inputs; the horizontal ground reaction force and ankle torque, by linear quadratic regulator, then transformed to the joint torques. We also utilize a special structure of the system having the relative degree of three, first found by Stojic, to stabilize the upright posture. Simulation results shows that the humanoid model robustly achieves dynamic upright balance from the initial posture far from static equilibria, while simultaneously avoiding the negative vertical ground reaction force that often happens with Hip Strategy alone.