This study introduces expedite the complete transfer of distributed gravel piles with an automated wheel loader. The wheel loader scoops the gravel and unloads it onto the bed of a truck. The total mileage for the repeated scooping and unloading work is reduced. The complicated optimal task is divided into three simple, organized subjects and different appropriate algorithms are used for each subject. A method is proposed that interactively selects the appropriate scooping points, unloading points, and appropriate path for the varying shape of the gravel pile until all gravel piles have been scooped up. To expedite the complete transfer of distributed gravel piles, the deep reinforcement learning model, trained via simulation, can be used in practice by implementing as part of the proposed methods. This paper describes these issues and the proposed methods. The performance of the actual application is demonstrated in terms of the calculation time and the feasibility of the simulations.

This paper proposes methods for unsupervised lexical acquisition for relative spatial concepts using spoken user utterances. A robot with a flexible spoken dialog system must be able to acquire linguistic representation and its meaning specific to an environment through interactions with humans as children do. Specifically, relative spatial concepts (e.g. front and right) are widely used in our daily lives, however, it is not obvious which object is a reference object when a robot learns relative spatial concepts. Therefore, we propose methods by which a robot without prior knowledge of words can learn relative spatial concepts. The methods are formulated using a probabilistic model to estimate the proper reference objects and distributions representing concepts simultaneously. The experimental results show that relative spatial concepts and a phoneme sequence representing each concept can be learned under the condition that the robot does not know which located object is the reference object. Additionally, we show that two processes in the proposed method improve the estimation accuracy of the concepts: generating candidate word sequences by class n-gram and selecting word sequences using location information. Furthermore, we show that clues to reference objects improve accuracy even though the number of candidate reference objects increases.

In future Mars exploration scenarios, astronauts orbiting the planet will control robots on the surface with supervised autonomy to construct infrastructure necessary for human habitation. Symbol-based planning enables intuitive supervised teleoperation by presenting relevant action possibilities to the astronaut. While our initial analog experiments aboard the International Space Station (ISS) proved this scenario to be very effective, the complexity of the problem puts high demands on domain models. However, the symbols used in symbolic planning are error-prone as they are often hand-crafted and lack a mapping to actual sensor information. While this may lead to biased action definitions, the lack of feedback is even more critical. To overcome these issues, this paper explores the possibility of learning the mapping between multi-modal sensor information and high-level preconditions and effects of robot actions. To achieve this, we propose to utilize a Multi-modal Latent Dirichlet Allocation (MLDA) for unsupervised symbol emergence. The learned representation is used to identify domain-specific design flaws and assist in supervised autonomy robot operation by predicting action feasibility and assessing the execution outcome. The approach is evaluated in a realistic telerobotics experiment conducted with the humanoid robot Rollin' Justin.

Humans create internal models of an environment (i.e. cognitive maps) through subjective sensorimotor experiences and can also understand spatial locations by looking at an external map as a symbol of an environment. We simulate the development of the cognitive map from sensorimotor experiences and grounding of the external map in a single deep neural network model. Our proposed network has a shared module that processes the features of multiple modalities (i.e. vision, hearing, and touch) and even external maps in the same manner. The multiple modalities are encoded into feature vectors by modality-specific encoders, and the encoded features are processed by the same shared module. The proposed network was trained to predict the sensory inputs of a simulated mobile robot. After the predictive learning, the spatial representation was developed in the internal states of the shared module, and the same spatial representation was used for predicting multiple modalities, including the external map. The network can also perform spatial navigation by associating the external map with the cognitive map. This implies that the external maps are grounded in subjective sensorimotor experiences, being bridged through the developed internal spatial representation in the shared module.

This paper reports the effectiveness of a tactile stimulus in a virtual environment to increase people's motivations during a monotonous task by comparing a touch with only visual stimuli and another with both visual and tactile stimuli. Although touch interaction with robots showed various positive effects such as improved motivation, visual and haptic stimuli were not separated due to the form of the touch in physical environments. Virtual environments enable us to investigate such effects by separating the modality of touch: a visual-only-touch and a visual-tactile touch. We experimented in a virtual environment where participants did a monotonous task after experiencing either a visual-only-touch or a visual-tactile-touch by an agent to compare them in the context of motivation improvements and the perceived impressions. The experimental results showed that a visual-tactile touch significantly increases the motivations of the participants for monotonous tasks. On the other hand, their likability of the agents is not significantly changed by the touch modalities.

In this study, we examine robotic proximity perception for both reactive control and 3D point-cloud acquisition using a multi-fingered robot hand. A reactive controller adjusts the relative position and orientation between the fingertips and the target object, while the robot-hand system gathers point-cloud data related to the shape of the object. An experimental result showed that the accuracy of the point-cloud generated using proximity perception was equivalent to or better than that of the point-cloud generated using a vision sensor. Based on the results, we propose proximity-based point-cloud acquisition using ‘non-contact perception’ motion to search the object's surface while preventing unintended collisions. The motion is then combined with the approaching and twisting motions of the robot hand. This enables multi-angles object detection complementarily with reactive fingertip control. Our experiments reveal that the proposed method can generate data-rich point-clouds for various general objects.

Planetary rovers are becoming indispensable for exploration activities and science missions. The rover used in such mission is often limited in its operation time owing to power and computational resources of the rover. In this paper, a trajectory planning method for a planetary rover is proposed that considers vehicle dynamics, and energy management of the rover. The vehicle dynamics is approximated from a dynamic simulation of the rover, which can estimate the power consumption in accordance with terrain traversability of the rover. The power generation of the solar array panel mounted on the rover is also taken into account. The simulation study confirmed the usefulness of the proposed method, especially in scenarios where slopes could be observed, and one result indicated that the energy margin could be improved by 4.1 kJ, 13.9 at maximum.

NASA and ESA are proposing to retrieve samples from Mars with a rover that will need to drive long distances, beyond what is seen on images from previous downlinks, to complete its mission in less than 200 Martian days. This paper proposes a planning method to support Mars Sample Return by integrating terrain information into the planning phase. The first step consists of predicting rover slip and drive speed using orbital data. When tested with Mars data, slip ranges are predicted and an expected velocity map is successfully drawn. The second step involves planning the traverse for the driving objective of a sol, based on the map of expected slip and drive speed. While driving, the rover monitors the terrain by converting its wheel slip and speed into terrain information, projected forward to confirm or refute predicted behavior, update the map and replan if necessary. Simulation showed successful map updates given new information and replanning when needed. If monitoring is insufficient to confirm traversability, a third step describes an instrument would be deployed to compute intrinsic terrain parameters, information then propagated ahead to update and replan if necessary. Experiments showed results close to the expected parameters of the tested soils.

This article proposes a novel two-staged trajectory planning algorithm toward the cooperative visual coverage of multiple asteroid sites with the utilization of multiple spacecraft. The objective of the novel established planning scheme is to determine an observation tour for each spacecraft and the associated fuel optimal trajectories, such that each site of interest is observed only once during the entire mission. The completion of the observation task of an asteroid site is limited to the observation time window, referred as the period for which a site is illuminated by the Sun. The proposed planning algorithm divides the overall multi-site coverage problem into a nonlinear and integer optimization problem as: (i) a single target optimization and (ii) a multi-target multiple spacecraft optimal sequencing problem. The first part aims to generate fuel optimal trajectories, for all the initial-final imaging site location pairs. In the second part, the problem of distributing the observation task, among the fleet of spacecraft, is addressed by designing a feasible tour to observe a subset of the desired asteroid sites, for each spacecraft, while considering their individual fuel capacity. The efficacy of the proposed algorithm is evaluated in multiple simulation scenarios and while considering different asteroids.

The installation of remotely-operated service robots in the environments of our daily life (including offices, homes, and hospitals) can improve work-from-home policies and enhance the quality of the so-called new normal. However, it is evident that remotely-operated robots must have partial autonomy and the capability to learn and use local semiotic knowledge. In this paper, we argue that the development of semiotically adaptive cognitive systems is key to the installation of service robotics technologies in our service environments. To achieve this goal, we describe three challenges: the learning of local knowledge, the acceleration of onsite and online learning, and the augmentation of human–robot interactions.