Online Gain Adaptation of Whole-Body Control for Legged Robots with Unknown Disturbances

This paper proposes an online gain adaptation approach to enhance the robustness of whole-body control (WBC) framework for legged robots under unknown external force disturbances. Without properly accounting for external forces, the closed-loop control system incorporating WBC may become unstable, and therefore the desired task goals may not be achievable. To study the effects of external disturbances, we analyze the behavior of our current WBC framework via the use of both full-body and centroidal dynamics. In turn, we propose a way to adapt feedback gains for stabilizing the controlled system automatically. Based on model approximations and stability theory, we propose three conditions to ensure that the adjusted gains are suitable for stabilizing a robot under WBC. The proposed approach has four contributions. We make it possible to estimate the unknown disturbances without force/torque sensors. We then compute adaptive gains based on theoretic stability analysis incorporating the unknown forces at the joint actuation level. We demonstrate that the proposed method reduces task tracking errors under the effect of external forces on the robot. In addition, the proposed method is easy-to-use without further modifications of the controllers and task specifications. The resulting gain adaptation process is able to run in real-time. Finally, we verify the effectiveness of our method both in simulations and experiments using the bipedal robot Draco2 and the humanoid robot Valkyrie.

Post-interval EEG activity is related to task-goals in temporal discrimination

Studies investigating the neural mechanisms of time perception often measure brain activity while participants perform a temporal task. However, several of these studies are based exclusively on tasks in which time is relevant, making it hard to dissociate activity related to decisions about time from other task-related patterns. In the present study, human participants performed a temporal or color discrimination task of visual stimuli. Participants were informed which magnitude they would have to judge before or after presenting the two stimuli (S1 and S2) in different blocks. Our behavioral results showed, as expected, that performance was better when participants knew beforehand which magnitude they would judge. Electrophysiological data (EEG) was analysed using Linear Discriminant Contrasts (LDC) and a Representational Similarity Analysis (RSA) approach to investigate whether and when information about time and color was encoded. During the presentation of S1, we did not find consistent differences in EEG activity as a function of the task. On the other hand, during S2, we found that temporal and color information was encoded in a task-relevant manner. Taken together, our results suggest that task goals strongly modulate decision-related information in EEG activity.

Humans reconfigure target and distractor processing to address distinct task demands

When faced with distraction, we can focus more on goal-relevant information (targets) or focus less goal-conflicting information (distractors). How people decide to distribute cognitive control across targets and distractors remains unclear. To help address this question, we developed a parametric attentional control task with a graded manipulation to both target discriminability and distractor interference. We find that participants exert independent control over target and distractor processing. We measured control adjustments through the influence of incentives and previous conflict on target and distractor sensitivity, finding that these have dissociable influences on control. Whereas incentives preferentially led to target enhancement, conflict on the previous trial preferentially led to distractor suppression. These distinct drivers of control altered sensitivity to targets and distractors early in the trial, and were promptly followed by reactive reconfiguration towards task-appropriate feature sensitivity. Finally, we provide a process-level account of these findings by show that these control adjustments are well-captured by an evidence accumulation model with attractor dynamics over feature weights. These results help establish a process-level account of control configuration that provides new insights into how multivariate attentional signals are optimized to achieve task goals.

Bayesian Experience Reuse for Learning from Multiple Demonstrators

Learning from Demonstrations (LfD) is a powerful approach for incorporating advice from experts in the form of demonstrations. However, demonstrations often come from multiple sub-optimal experts with conflicting goals, rendering them difficult to incorporate effectively in online settings. To address this, we formulate a quadratic program whose solution yields an adaptive weighting over experts, that can be used to sample experts with relevant goals. In order to compare different source and target task goals safely, we model their uncertainty using normal-inverse-gamma priors, whose posteriors are learned from demonstrations using Bayesian neural networks with a shared encoder. Our resulting approach, which we call Bayesian Experience Reuse, can be applied for LfD in static and dynamic decision-making settings. We demonstrate its effectiveness for minimizing multi-modal functions, and optimizing a high-dimensional supply chain with cost uncertainty, where it is also shown to improve upon the performance of the demonstrators' policies.

Conveyor of a new assortment of peaches and nectarines for Crimea and southern Russia

Aim and task: Goals and objectives: to form a conveyor of industrial assortment of peach and nectarine for the Crimea and southern Russia. To do this, study the new varieties of peach and nectarine bred by the Nikitsky Botanical Gardens for a complex of economic and biological characteristics. The study of varieties was carried out according to the "Program and methodology for the study of varieties of fruit, berry and nut crops" (Sedov, 1995), as well as in accordance with the classifier (Khloptseva, Sharova, Korneichuk, 1988). 34 new varieties of peach and 7 - nectarine of NBG selection of different ripening periods were studied by a complex of economically valuable traits. Genotypes with a complex of outstanding properties in terms of yield, frost resistance of flower buds, drought resistance, resistance to fungal diseases, and commercial qualities of fruits with an increased content of biologically active substances have been identified. A conveyor belt of a new industrial assortment of peaches and nectarines of the Nikitsky Botanical Garden selection for gardening in Crimea and southern Russia has been proposed.

Working memory resources can be efficiently deallocated from items that become obsolete

Visual working memory (VWM) resources are limited, placing constraints on how much visual information can be simultaneously retained. During visually-guided activity, stored information can quickly become outdated, so updating mechanisms are needed to ensure the contents of memory remain relevant to current task goals. In particular, successful deallocation of resources from items that become obsolete is likely to be critical for maintaining the precision of those representations still in memory. The experiments in this study involved presenting two memory arrays of coloured disks in sequence. The appearance of the second array was a cue to replace either one, or all, of the colours in memory from the first array. When compared to conditions that were matched in the total number of items presented, we predicted that successful updating would result in fewer actively maintained items, owing to the removal of pre-replacement features. Comparison of recall precision between these two conditions should therefore provide a test of successful updating. We identified a very small proportion (~5%) of trials in which participants incorrectly reported a feature from the first array in place of its replacement in the second, which we interpreted as a failure to incorporate the information from the second display into memory. Once these trials were discounted, precision estimates were consistent with complete deallocation of resources from pre-replacement features. We conclude that working memory can be efficiently updated when previous information becomes obsolete, but that this is a demanding active process that occasionally fails.

Body–Space Interactions: Same Spatial Encoding but Different Influence of Valence for Reaching and Defensive Purposes

The space around our body, the so-called “peripersonal space,” is where interactions with nearby objects may occur. “Defensive space” and “Reaching space” respectively refer to two opposite poles of interaction between our body and the external environment: protecting the body and performing a goal-directed action. Here, we hypothesized that mechanisms underlying these two action spaces are differentially modulated by the valence of visual stimuli, as stimuli with negative valence are more likely to activate protective actions whereas stimuli with positive valence may activate approaching actions. To test whether such distinction in cognitive/evaluative processing exists between Reaching and Defensive spaces, we measured behavioral responses as well as neural activations over sensorimotor cortex using EEG while participants performed several tasks designed to tap into mechanisms underlying either Defensive (e.g., respond to touch) or Reaching space (e.g., estimate whether object is within reaching distance). During each task, pictures of objects with either positive or negative valence were presented at different distances from the participants' body. We found that Defensive space was smaller for positively compared with negatively valenced visual stimuli. Furthermore, sensorimotor cortex activation (reflected in modulation of beta power) during tactile processing was enhanced when coupled with negatively rather than positively valenced visual stimuli regarding Defensive space. On the contrary, both the EEG and behavioral measures capturing the mechanisms underlying Reaching space did not reveal any modulation by valence. Thus, although valence encoding had differential effects on Reaching and Defensive spaces, the distance of the visual stimulus modulated behavioral measures as well as activity over sensorimotor cortex (reflected in modulations of mu power) in a similar way for both types of spaces. Our results are compatible with the idea that Reaching and Defensive spaces involve the same distance-dependent neural representations of sensory input, whereas task goals and stimulus valence (i.e., contextual information) are implemented at a later processing stage and exert an influence on motor output rather than sensory/space encoding.

Automatic object-based spatial selection depends on the distribution of sustained attention

Several space-based and object-based attention studies suggest these selection mechanisms may be voluntarily deployed, depending on task parameters and the attentional scope of the observer. Here, we sought to elucidate factors related to involuntary deployment of object-mediated space-based attention through two experiments. Experiment 1 used a modified flanker task where a target and nearby distractor were presented within the same or different object frames, such that an object-based attentional spread should be detrimental to performance. Results showed the presence of a flanker effect with no significant difference in magnitude between grouping conditions, indicating participants may have uniformly used a diffused attentional spotlight regardless of object segmentation. In a second experiment, we manipulated the extent of the observer’s sustained attentional scope via an inducer task to determine whether object-based selection depends on the initial spotlight size. The results revealed object-based effects solely when attention narrowly encompassed the target, but not when it was widened to include the distracting flanker. This suggests the deployment of object-based attention may occur when spatial attention is initially focused narrowly. Because selecting the whole object frame directly interfered with task goals, we conclude that object-based attention may not always fully conform to relevant task goals or operate in a goal-oriented manner. We discuss these results in the context of existing literature while proposing a reconciliation of previously inconsistent findings of object-based selection.

Research on the Professional Development Attitude of Minority English Teachers

A total of 180 minority English teachers of different ages in middle schools are involved as quantitative research participants in this study. SPSS is applied to analyze the status quo and differences of their professional development attitude among different ages. The results show that they have the strong positive attitude toward their professional development, but teachers of different ages have significant differences in the dimensions of Professional Teaching Knowledge, Flexible Task Goals Consciousness as well as Rigid Task Target Consciousness.