Testing Conditions for Communicating Stream X-machine Systems

X-machines were proposed by Holcombe as a possible specification language and since then a number of further investigations have demonstrated that the model is intuitive and easy to use. In particular, stream X-machines (SXM), a particular class of X-machines, have been found to be extremely useful in practice. Furthermore, a method of testing systems specified as SXMs exists and is proved to detect all faults of the implementation provided that the system meets certain “design for test conditions”. Recently, a system of communicating SXMs was introduced as a means of modelling parallel processing. This paper proves that each communicating machine component can be transformed in a straightforward manner so that the entire system will behave like a single stream X-machine - the equivalent SXM of the system. The paper goes on to investigate the applicability of the SXM testing method to a system of communicating SXMs and identifies a class of communicating SXMs for which the equivalent SXM of the system meets the “design for test conditions”. Received November 1999 / Accepted in revised form June 2001     

Relationships Between LRI Process Parameters and Impact and Post-Impact Behaviour of Stitched and Unstitched NCF Laminates

The general context of the development of out-of-autoclave processes in the aeronautics industry raises the question of the possible links between these new processes and impact behaviour. In this study, a Taguchi table was used in a design of experiment approach to establish possible links. The study focused on the liquid resin infusion process applied to laminates made with stitched or unstitched quadri-axial carbon Non-Crimp Fabric (NCF). On the basis of previous studies and an analysis of the literature, five process parameters were selected (stitching, curing temperature, preform position, number of highly porous media, vacuum level). The impact energy was set at 35?J in order to obtain enough residual dent depth. The parameters analysed during and after impact were: maximum displacement of the impactor, energy absorbed, permanent indentation depth, and delaminated surface. Then, compression after impact tests were performed and the corresponding average stress was measured. The interactions found by statistical analysis show a very high sensitivity to stitching, which was, of course, expected. A very significant influence of curing temperature and a significant influence of preform position were also found on the permanent indentation depth and a physical explanation is provided. Globally, it was demonstrated that the resin infusion process itself did not influence the impact behaviour.     

A unified integration and component testing approach from deterministic stream X-machine specifications

One of the great benefits of using a stream X-machine to specify a system is its associated testing method. Under certain design for test conditions, this method produces a test suite that can determine the correctness of the implementation under test (IUT), provided that the basic components of the stream X-machine model have been correctly implemented. However, such an approach implies that each component can be tested in isolation from the rest of the system. This is a limitation that, in practice, can be resolved by developing stubs and drivers. However, this adds complexity to the testing process and, furthermore, these new pieces of software can introduce faults that can invalidate the theoretical results of the aforementioned testing method. This paper extends the approach by allowing component testing to be performed in parallel with integration testing, while still guaranteeing the IUT correctness under the given design for test conditions. It also shows how the integration test suite, produced in previous publications, can be reduced.     

Learning to pour with a robot arm combining goal and shape learning for dynamic movement primitives

When describing robot motion with dynamic movement primitives (DMPs), goal (trajectory endpoint), shape and temporal scaling parameters are used. In reinforcement learning with DMPs, usually goals and temporal scaling parameters are predefined and only the weights for shaping a DMP are learned. Many tasks, however, exist where the best goal position is not a priori known, requiring to learn it. Thus, here we specifically address the question of how to simultaneously combine goal and shape parameter learning. This is a difficult problem because learning of both parameters could easily interfere in a destructive way. We apply value function approximation techniques for goal learning and direct policy search methods for shape learning. Specifically, we use “policy improvement with path integrals” and “natural actor critic” for the policy search. We solve a learning-to-pour-liquid task in simulations as well as using a Pa10 robot arm. Results for learning from scratch, learning initialized by human demonstration, as well as for modifying the tool for the learned DMPs are presented. We observe that the combination of goal and shape learning is stable and robust within large parameter regimes. Learning converges quickly even in the presence of disturbances, which makes this combined method suitable for robotic applications.     

Using efference copy and a forward internal model for adaptive biped walking

To behave properly in an unknown environment, animals or robots must distinguish external from self-generated stimuli on their sensors. The biologically inspired concepts of efference copy and internal model have been successfully applied to a number of robot control problems. Here we present an application of this for our dynamic walking robot RunBot. We use efference copies of the motor commands with a simple forward internal model to predict the expected self-generated acceleration during walking. The difference to the actually measured acceleration is then used to stabilize the walking on terrains with changing slopes through its upper body component controller. As a consequence, the controller drives the upper body component (UBC) to lean forwards/backwards as soon as an error occurs resulting in dynamical stable walking. We have evaluated the performance of the system on four different track configurations. Furthermore we believe that the experimental studies pursued here will sharpen our understanding of how the efference copies influence dynamic locomotion control to the benefit of modern neural control strategies in robots.     

Identification of the parameters of an elastic material model using the constitutive equation gap method

Today, the identification of material model parameters is based more and more on full-field measurements. This article explains how an appropriate use of the constitutive equation gap method (CEGM) can help in this context. The CEGM is a well-known concept which, until now, has been used mainly for the verification of finite element simulations. This has led to many developments, especially concerning the techniques for constructing statically admissible stress fields. The originality of the present study resides in the application of these recent developments to the identification problem. The proposed CEGM is described in detail, then evaluated through the identification of heterogeneous isotropic elastic properties. The results obtained are systematically compared with those of the equilibrium gap method, which is a well-known technique for the resolution of such identification problems. We prove that the use of the enhanced CEGM significantly improves the quality of the results.     

Temporal sequence learning, prediction, and control: a review of different models and their relation to biological mechanisms

In this review, we compare methods for temporal sequence learning (TSL) across the disciplines machine-control, classical conditioning, neuronal models for TSL as well as spike-timing-dependent plasticity (STDP). This review introduces the most influential models and focuses on two questions: To what degree are reward-based (e.g., TD learning) and correlation-based (Hebbian) learning related? and How do the different models correspond to possibly underlying biological mechanisms of synaptic plasticity? We first compare the different models in an open-loop condition, where behavioral feedback does not alter the learning. Here we observe that reward-based and correlation-based learning are indeed very similar. Machine control is then used to introduce the problem of closed-loop control (e.g., actor-critic architectures). Here the problem of evaluative (rewards) versus nonevaluative (correlations) feedback from the environment will be discussed, showing that both learning approaches are fundamentally different in the closed-loop condition. In trying to answer the second question, we compare neuronal versions of the different learning architectures to the anatomy of the involved brain structures (basal-ganglia, thalamus, and cortex) and the molecular biophysics of glutamatergic and dopaminergic synapses. Finally, we discuss the different algorithms used to model STDP and compare them to reward-based learning rules. Certain similarities are found in spite of the strongly different timescales. Here we focus on the biophysics of the different calcium-release mechanisms known to be involved in STDP.