Unexpected Input Research Articles

Approach to the analysis of user requirements in assistive technology

Assistive technology plays an important role for independent living of people with special needs. Therefore it is a key issue to identify the problems that come up in the everyday use of assistive technology. The analysis of user requirements is one of the most important factors for effective special needs developments. The user involvement in every step of the development procedure is meaningful. Especially the experiences of long-term users are very helpful. In case of new technology prototype testing by users during the development can correct the aims by giving unexpected inputs. Although some research activities with user involvement are going on, it still needs more concern in industry and appropriate methodologies have to be established. In this paper two different examples of user involvement are described with respect to the different users' experience. As an example of emerging technologies the use of a robotic aid is investigated. In this case the analysis of user requirements is based on tests of available prototypes with users. As an example of already existing technical aids an investigation of practical user experiences with wheelchairs is described. In both cases the experiences of the users are fed back to the development to achieve real improvement. Our experiences may stimulate researchers and industrialists to initiate user involvement within their activities.

Ｔ‐２ ＣＣＶのＰｉｌｏｔ‐Ｉｎｄｕｃｅｄ Ｏｓｃｉｌｌａｔｉｏｎ（ＰＩＯ）特性とその改善

Pilot-Induced Oscillation (PIO) was encountered in the company flight test of the T-2CCV research vehicle. The main cause of the roll PIO was due to unexpected large pilot input and to increased delay of the aircraft response by its large pilot input. The delay of roll response has much difference between with large input and with small input. That is why the flight control system falls into nonlinear system by actuator rate limit in case of controlling the aircraft with large pilot input. This paper shows analysis of the characteristics of the PIO and its improvement.

Perceptual inhibition of expected inputs: The key that opens closed minds

The mind appears to be biased simultaneously toward both expected and unexpected inputs. For example, familiar scenes are usually perceived more readily than novel scenes, indicating the former bias, but a single novel object sometimes pops out from a familiar field, indicating the latter bias. A diverse literature and a computational model converge on the following resolution to this paradox: The former bias is conceptually driven and actually suppresses data-driven processing of expected inputs; in turn, this suppression disinhibits data-driven processing of unexpected inputs, yielding the latter bias. Evidence for suppressed data-driven processing of expected inputs is drawn from studies of perceptual habituation, semantic satiation, memory inhibition, inhibition of return, repetition blindness, primed inhibition, the word-inferiority effect, registration without learning, and both expert- and schema-based inhibitory effects. Evidence for enhanced data-driven processing of unexpected inputs is drawn from studies of the orienting response, mismatch negativity, memory facilitation, both expert- and schema-based facilitatory effects, and perceptual popout. The model, calledmismatch theory, incorporates inhibitory and facilitatory perceptual dynamics and is found to simulate the opposing biases. Implications of mismatch theory for perceptual phenomenology, dynamic systems theory, mental health, and individual differences are also discussed.

Learning Complex, Extended Sequences Using the Principle of History Compression

Previous neural network learning algorithms for sequence processing are computationally expensive and perform poorly when it comes to long time lags. This paper first introduces a simple principle for reducing the descriptions of event sequences without loss of information. A consequence of this principle is that only unexpected inputs can be relevant. This insight leads to the construction of neural architectures that learn to “divide and conquer” by recursively decomposing sequences. I describe two architectures. The first functions as a self-organizing multilevel hierarchy of recurrent networks. The second, involving only two recurrent networks, tries to collapse a multilevel predictor hierarchy into a single recurrent net. Experiments show that the system can require less computation per time step and many fewer training sequences than conventional training algorithms for recurrent nets.

Stability analysis of independent joint PID feedback control for trajectory tracking of robot motion

Abstract In this paper, the effectiveness of a local PID feedback control is discussed via a Lyapunov method. In particular, an energy‐like Lyapunov function is contrived to facilitate the analysis. It is shown that with suitable choice of feedback gains the exponential convergence tracking can be achieved using on‐line state‐dependent or off‐line target‐dependent feedforward compensation, where the latter alleviates excessive computational burden. In the presence of parametrized or non‐parametrized uncertainty of the dynamic model of the system, bounded tracking error is also ensured under similar conditions. This implies the robustness of this control scheme to payload variation and/or unexpected input disturbances. Moreover, the controller is implemented in a decentralized manner despite the highly non‐linear couplings among the joints.

Gating in the spino‐olivocerebellar pathways to the c1 zone of the cerebellar cortex during locomotion in the cat.

1. The field potentials evoked in the cerebellar cortical c1 zone by single-pulse, non-noxious stimulation of the superficial radial nerve have been recorded with tungsten-in-glass microelectrodes in awake cats. Responses that were due to transmission in the spino-olivocerebellar pathways (SOCPs), which terminate in the cortex as climbing fibres, were identified and studied while the cat walked on a moving belt. 2. The size of the climbing fibre-evoked potentials varied systematically during the step cycle. They were invariably largest in mid- to late swing of the ipsilateral forelimb and, at most recording sites (5/6), they were smallest during the first half of stance. 3. With low stimulus strength, the probability of evoking a measurable response also varied. The probability was greatest in mid- to late swing and least in early stance. 4. Similar variations were shown to occur when the analysis was restricted to responses evoked by a single functionally homogenous SOCP, the dorsal funiculus SOCP. 5. It is proposed that these variations reflect the operation of a gating mechanism which modulates the excitability of the SOCPs and prevents them transmitting self-generated tactile inputs to the cerebellum while facilitating transmission when unexpected inputs are most likely to arise. 6. The present data are compared with those from a similar study of the c2 zone SOCPs (Apps, Lidierth & Armstrong, 1990) and are discussed in relation to a study of the effects of unexpected mechanical perturbations of stepping (Andersson & Armstrong, 1987).

Ronald Reagan and Humor: A politician's velvet weapon

The use of humor was an important aspect of Ronald Reagan's rhetorical appeal. Three major theories of humor have been suggested in the development of the literature: the relief theory, involving emotional release of tension; the incongruity theory, where the cognitive functioning reacts to unexpected input; and the superiority theory, which holds that humor results from feeling superior to others, who may be considered disobedient or somehow “wrong.” The application of the theories to Reagan's humor shows how he used humor to criticize authority without being perceived as too harsh or negative, as well as to promote positive feelings from his audiences.

ADDING LEARNED EXPECTATION INTO THE LEARNING PROCEDURE OF SELF-ORGANIZING MAPS

The self-organizing topological map is generalized by adding a learned expectation to its learning procedure, in order to improve its stability in nonstationary environments with unexpected inputs and abnormal noise. Such a learned expectation is realized through a compatibility test which checks whether an input is compatible with the earlier learned patterns on a map unit before the unit starts to adapt to the input. The generalized map consists of multi-maps with a pipeline architecture, equipped with a parallel search strategy which allows for the implementation of the learned expectation to be fulfilled without extra computing costs. Computer experiments on several examples are given to show the characteristics of the generalized map in comparison with the original map.

Neural adaptability: a biological determinant of behavioral intelligence.

Despite a history of attempts, neuroscience has thus far failed to identify a validated biological determinant of behavioral intelligence. Cognitive variables such as selective attention, expectancy and information processing workload modulate the amplitude of evoked cortical potentials (EPs), thereby demonstrating the operation of cognitive neural adaptability. It seems reasonable to postulate that individual differences in this aspect of neural function could relate to individual differences in behavioral intelligence: the electrophysiologically adaptable brain should be the behaviorally bright brain.To test this hypothesis I gathered auditory EPs from 109 normal and 52 mentally retarded adults under three stimulation conditions (periodic, self, and random) designed to manipulate temporal expectancy.The normal adults showed a strong temporal expectancy effect on their EPs, giving smaller than average EPs to expected inputs and larger than average brain responses to unexpected stimuli. In contrast, the retarded adults failed to show a temporal expectancy effect on their EPs, indicating a deficit in neural adaptability. A measure of neural adaptability derived from EP amplitude ratios correlated. 66 with IQ scores obtained on 74 normal adults, indicating a definite association between neural adaptability and behavioral intelligence. People who gave larger than average EPs to unexpected inputs and smaller than average EPs to stimuli whose timing they knew tended to have higher IQs.Results suggest that the brain which efficiently inhibits its response to insignificant inputs and orients vigorously to unexpected, potentially dangerous stimuli is also the brain which manifests high behavioral intelligence. Neural adaptability as indexed by the temporal expectancy effect on evoked cortical potentials provides a biological determinant of behavioral intelligence.

On GLR detection and estimation of unexpected inputs in linear discrete systems

In this paper, we present a recursive generalized likelihood ratio (GLR) test algorithm for detecting sudden changes in linear discrete systems. We demonstrate the application of linear filtering techniques to obtain a recursive GLR algorithm so that the requirement for matrix inversions in the previously known GLR algorithms can be reduced or avoided. Furthermore, the GLR algorithm is extended to the case when the sudden change follows known linear dynamics. An adaptive filtering scheme which uses the input estimate to correct the state estimate is also presented for the time-varying input case.