Self-supervised learning for using overhead imagery as maps in outdoor range sensor localization

Traditional approaches to outdoor vehicle localization assume a reliable, prior map is available, typically built using the same sensor suite as the on-board sensors used during localization. This work makes a different assumption. It assumes that an overhead image of the workspace is available and utilizes that as a map for use for range-based sensor localization by a vehicle. Here, range-based sensors are radars and lidars. Our motivation is simple, off-the-shelf, publicly available overhead imagery such as Google satellite images can be a ubiquitous, cheap, and powerful tool for vehicle localization when a usable prior sensor map is unavailable, inconvenient, or expensive. The challenge to be addressed is that overhead images are clearly not directly comparable to data from ground range sensors because of their starkly different modalities. We present a learned metric localization method that not only handles the modality difference, but is also cheap to train, learning in a self-supervised fashion without requiring metrically accurate ground truth. By evaluating across multiple real-world datasets, we demonstrate the robustness and versatility of our method for various sensor configurations in cross-modality localization, achieving localization errors on-par with a prior supervised approach while requiring no pixel-wise aligned ground truth for supervision at training. We pay particular attention to the use of millimeter-wave radar, which, owing to its complex interaction with the scene and its immunity to weather and lighting conditions, makes for a compelling and valuable use case.

Modality Difference in N2 and P3 Components Between Visual and Auditory Go/No-Go Paradigms

We investigated the modality difference in the N2 and P3 components of event-related potentials (ERPs) between visual and auditory Go/No-go paradigms. We evaluated the relationship between RT and the amplitudes and latencies of N2 and P3 in visual and auditory Go/No-go paradigms. No significant differences were observed in the latencies of N2 and P3 between visual and auditory paradigms. Significant correlations were observed between RT and the latency of P3 in the visual and auditory paradigms. In contrast, the amplitudes of N2 and P3 were significantly larger in the visual paradigm than in the auditory paradigm. A significant correlation was observed between RT and the amplitude of P3 only in the auditory paradigm. These results suggested that there were two neural networks for the response execution and inhibitory function, common and uncommon, that depended on the stimulus modality.

Lexical development in Mandarin–English bilingual children

Two groups of Mandarin–English bilingual children (3–5-year-olds, 6–8-year-olds) participated in a picture identification task and a picture naming task in both languages. Results revealed age-related growth in English, but not Mandarin vocabulary. Composite vocabulary was larger than either single-language vocabulary in the younger children but was similar to English vocabulary in the older children. Furthermore, children showed a larger receptive–expressive modality difference in their weaker language (Mandarin) than in their stronger language (English). These patterns indicate rapid growth in English vocabulary along with early stabilization of Mandarin vocabulary despite considerable Mandarin input in the home setting.

Activation of Right Inferior Frontal Gyrus during Response Inhibition across Response Modalities

The go/no-go task, which effectively taps the ability to inhibit prepotent response tendency, has consistently activated the lateral prefrontal cortex, particularly the right inferior frontal gyrus (rIFG). On the other hand, rIFG activation has rarely been reported in the antisaccade task, seemingly an oculomotor version of the manual go/no-go task. One possible explanation for the variable IFG activation is the modality difference of the two tasks: The go/no-go task is performed manually, whereas the antisaccade task is performed in the oculomotor modality. Another explanation is that these two tasks have different task structures that require different cognitive processes: The traditional antisaccade task requires (i) configuration of a preparatory set prior to antisaccade execution and (ii) response inhibition at the time of antisaccade execution, whereas the go/no-go task requires heightened response inhibition under a minimal preparatory set. To test these possibilities, the traditional antisaccade task was modified in the present functional magnetic resonance imaging study such that it required heightened response inhibition at the time of antisaccade execution under a minimal preparatory set. Prominent activation related to response inhibition was observed in multiple frontoparietal regions, including the rIFG. Moreover, meta-analyses revealed that the rIFG activation in the present study was observed in the go/no-go tasks but not in the traditional antisaccade task, indicating that the rIFG activation was sensitive to the task structure difference, but not to the response modality difference. These results suggest that the rIFG is part of a network active during response inhibition across different response modalities.

Stimulus-response compatibility in intensity-force relations

Romaiguère, Hasbroucq, Possamaï, and Seal (1993) reported a new compatibility effect from a task that required responses of two different target force levels to stimuli of two different intensities. Reaction times were shorter when high and low stimulus intensities were mapped to strong and weak force presses respectively than when this mapping was reversed. We conducted six experiments to refine the interpretation of this effect. Experiments 1 to 4 demonstrated that the compatibility effect is clearly larger for auditory than for visual stimuli. Experiments 5 and 6 generalized this finding to a task where stimulus intensity was irrelevant. This modality difference refines Romaiguère et al.'s (1993) symbolic coding interpretation by showing that modality-specific codes underlie the intensity-force compatibility effect. Possible accounts in terms of differences in the representational mode and action effects are discussed.

Comparing communicative competence in child and chimp: the pragmatics of repetition

ABSTRACTThrough an analysis of chimpanzee–human discourse, we show that two Pan troglodytes chimpanzees and two Pan paniscus chimpanzees (bonobos) exposed to a humanly devised symbol system use partial or complete repetition of others' symbols, as children do: they do not produce rote imitations, but instead use repetition to fulfil a variety of pragmatic functions in discourse. These functions include agreement, request, promise, excitement, and selection from alternatives. In so doing, the chimpanzees demonstrate contingent turn-taking and the use of simple devices for lexical cohesion. In short, they demonstrate conversational competence. Because of the presence of this conversational competence in three sibling species, chimpanzees, bonobos, and humans, it is concluded that the potential to express pragmatic functions through repetition was part of the evolutionary history of human language, present in our common ancestor before the phylogenetic divergence of hominids and chimpanzees. In the context of these similarities, two interesting differences appeared: (1) Human children sometimes used repetition to stimulate more talk in their conversational partner; the chimpanzees, in contrast, use repetition exclusively to forward the non-verbal action. This difference may illuminate a unique feature of human linguistic communication, or it may simply reflect a modality difference (visual symbols used by the chimpanzees, speech used by the children) in the symbol systems considered in this research. A second difference seems likely to reflect a true species difference: utterance length. The one- and two-symbol repetitions used by the chimpanzees to fulfil a variety of pragmatic functions were less than half the maximum length found in either the visual symbol combinations addressed to them by their adult human caregivers or the oral repetitions of two-year-old children. This species difference probably reflects the evolution of increased brain size and consequent increased memory capacity that has occurred since the phylogenetic divergence of hominids and chimpanzees four to seven million years ago.

Intra- and Inter-Modal Effects of Prior Stimulation on Cardiac Responsiveness to Repeated Stimulation in the Human Newborn

The effect of prior equivalent auditory and somesthetic input on cardiac acceleration responses to repeated auditory and somesthetic stimulation was studied in 56 healthy 2- or 3-day-old infants. Responsiveness to initial presentation of the reiterative auditory stimulus was significantly greater than to the reiterative somesthetic stimulus regardless of the modality stimulated during the preceding series. In contrast the course of response decrement to subsequent repeated presentations of both stimuli was significantly more rapid when the reiterative and prior stimuli were in different modalities (inter-modal) than when they were in the same modality (intra-modal). The findings demonstrate a modality difference in effect of antecedent input on initial responsiveness and a greater efficacy of inter- than of intra-modal auditory and somesthetic stimulation for the habituation of neonatal cardiac responses to repeated stimulation.