Portail HAL Ensta
Not a member yet
11080 research outputs found
Sort by
A compact and recursive Riemannian motion descriptor for untrimmed activity recognition
No full textInternational audienceA very low dimension frame-level motion descriptor is herein proposed with the capability to represent incomplete dynamics, thus allowing online action prediction. At each frame, a set of local trajectory kinematic cues are spatially pooled using a covariance matrix. The set of frame-level covariance matrices forms a Riemannian manifold that describes motion patterns. A set of statistic measures are computed over this manifold to characterize the sequence dynamics, either globally, or instantaneously from a motion history. Regarding the Riemannian metrics, two different versions are proposed: (1) by considering tangent projections with respect to updated recursive statistics, and (2) by mapping the covariance onto a linear matrix using as reference the identity matrix. The proposed approach was evaluated for two different tasks: (1) for action classification on complete video sequences and (2) for online action recognition, in which the activity is predicted at each frame. The method was evaluated using two public datasets: KTH and UT-interaction. For action classification, the method achieved an average accuracy of 92.27 and 81.67%, for KTH and UT-interaction, respectively. In partial recognition task, the proposed method achieved similar classification rate as for the whole sequence using only the 40 and 70% on KTH and UT sequences, respectively
La mobilisation des SHS dans la formation des ingénieurs au développement durable
No full textArticle de synthèse et point de vueArticle de synthèse et point de vu
EPAS Fail-Safe Control using Differential Braking
No full textInternational audienceOne of the main systems for ground vehicles is the Electric Power Assisted Steering (EPAS). Due to its importance, a redundant system is usually implemented. This solution remain however expensive, especially for small urban cars. This paper proposes an alternative fail-safe control algorithm using this time differential braking. Controlling each brake apart generates a yaw moment enabling turning the vehicle. Optimal coordination is ensured due to the importance of this maneuver. Results show that differential braking, that already exist in most of passenger cars, is able to generate a sufficient additional yaw moment to help the driver steer his/her vehicle. This can reduce the space and the cost of the vehicle
Managing Single or Multi-Users Channel Allocation for the Priority Cognitive Access
No full textInternational audienceThis manuscript investigates the problem of the Multi-Armed Bandit (MAB) in the context of the Opportunistic Spectrum Access (OSA) case with priority management (e.g. military applications). The main aim of a Secondary User (SU) in OSA is to increase his transmission throughput by seeking the best channel with the highest vacancy probability. In this manuscript, we propose a novel MAB algorithm called ?-UCB in order to enhance the spectrum learning of a SU and decrease the regret, i.e. the loss of reward due to the selection of worst channels. We analytically prove, and corroborate with simulations, that the regret of the proposed algorithm has a logarithmic behavior. So, after a finite number of time slots, the SU can estimate the vacancy probability of channels in order to target the best one for transmitting. Hereinafter, we extend ?- UCB to consider multiple priority users, where a SU can selfishly estimate and access the channels according to his prior rank. The simulation results show the superiority of the proposed algorithm for a single or multi-user cases compared to the well-known MAB algorithms
Intelligent Behavior Depends on the Ecological Niche
No full textInternational audienceIn this interview, Pierre-Yves Oudeyer outlines his perspective on the future of AI, and discusses directions for machines models of human-like intelligence. He explains how neuro-cognitive and evolutionary theories of human cognition should further inform artificial intelligence. He emphasizes the role of ecological niches in sculpting intelligent behavior, and in particular that human intelligence was fundamentally shaped to adapt to a constantly changing socio-cultural environment. He argues that a major limit of current work in AI is that it is missing this perspective, both theoretically and experimentally. Finally, he discusses the promising approach of developmental machine learning, modeling infant development through multi-scale interaction between intrinsically motivated learning, embodiment and a fastly changing socio-cultural environment
Intrinsically Motivated Open-Ended Multi-Task Learning Using Transfer Learning to Discover Task Hierarchy
No full textInternational audienceIn open-ended continuous environments, robots need to learn multiple parameterised control tasks in hierarchical reinforcement learning. We hypothesise that the most complex tasks can be learned more easily by transferring knowledge from simpler tasks, and faster by adapting the complexity of the actions to the task. We propose a task-oriented representation of complex actions, called procedures, to learn online task relationships and unbounded sequences of action primitives to control the different observables of the environment. Combining both goal-babbling with imitation learning, and active learning with transfer of knowledge based on intrinsic motivation, our algorithm self-organises its learning process. It chooses at any given time a task to focus on; and what, how, when and from whom to transfer knowledge. We show with a simulation and a real industrial robot arm, in cross-task and cross-learner transfer settings, that task composition is key to tackle highly complex tasks. Task decomposition is also efficiently transferred across different embodied learners and by active imitation, where the robot requests just a small amount of demonstrations and the adequate type of information. The robot learns and exploits task dependencies so as to learn tasks of every complexity
Grounding Artificial Intelligence in the Origins of Human Behavior
No full textRecent advances in Artificial Intelligence (AI) have revived the quest for agents able to acquire an open-ended repertoire of skills. However, although this ability is fundamentally related to the characteristics of human intelligence, research in this field rarely considers the processes that may have guided the emergence of complex cognitive capacities during the evolution of the species. Research in Human Behavioral Ecology (HBE) seeks to understand how the behaviors characterizing human nature can be conceived as adaptive responses to major changes in the structure of our ecological niche. In this paper, we propose a framework highlighting the role of environmental complexity in open-ended skill acquisition, grounded in major hypotheses from HBE and recent contributions in Reinforcement learning (RL). We use this framework to highlight fundamental links between the two disciplines, as well as to identify feedback loops that bootstrap ecological complexity and create promising research directions for AI researchers
BSDEs with no driving martingale, Markov processes and associated Pseudo Partial Differential Equations. Part II: Decoupled mild solutions and Examples.
No full textInternational audienceLet be a family of probability measures, where is a Polish space,defined on the canonical probability space of -valued cadlag functions. We suppose that a martingale problem with respect to a time-inhomogeneous generator is well-posed. We consider also an associated semilinear {\it Pseudo-PDE} with generator for which we introduce a notion of so called {\it decoupled mild} solution and study the equivalence with the notion of martingale solution introduced in a companion paper. We also investigate well-posedness for decoupled mild solutions and their relations with a special class of BSDEs without driving martingale. The notion of decoupled mild solution is a good candidate to replace the notion of viscosity solution which is not always suitable when the map is not a PDE operator
Interval Inspired Approach Based on Temporal Sequence Constraints to Place Recognition
No full textInternational audiencePlace recognition is an essential task in many robotics applications. Recognizing if the robot is crossing an already visited place may be used to improve its localization and map estimation. A place recognition strategy must be as accurate as possible, despite the challenges related to environment dynamicity. It should avoid generating false positives since even a few erroneous matches may be enough to cause the degradation of the Simultaneous Localization and Mapping (SLAM) process. We propose a novel approach for place recognition inspired by interval analysis theory. Our approach models the known world as a set of intervals based on the robot’s observations. The search to determine whether the current robot location is new or known begins as the robot explores its surroundings. Our approach has three main steps. First, it selects a set of nearest neighbors based on the similarity between the current robot observation and the intervals composing the known world. In the second step, our approach uses temporal constraints to select one element of the set. And finally, the third step is to sweep the selected interval looking for the query best match. We evaluate our proposal by dealing with visual place recognition using only image information and demonstrate its effectiveness using some challenging public datasets
Anatomy of Strike Slip Fault Tsunami-genesis
No full textTsunami generation from earthquake induced seafloor deformations has long been recognized as a major hazard to coastal areas. Strike—slip faulting has generally been believed as insufficient for triggering large tsunamis, except through the generation of submarine landslides. Herein, we demonstrate that ground motions due to strike–slip earthquakes can contribute to the emergence of large tsunamis (>1m) underrather generic conditions. To this end, we have developed a computational framework that integrates models for earthquake rupture dynamics with models of tsunami generation and propagation. The three-dimensional time-dependent vertical and horizontal ground motions from spontaneous dynamic rupture models are used to drive boundary motions in the tsunami model. Our results suggest that super shearruptures propagating along strike–slip faults, traversing narrow and shallow bays are prime candidates for tsunami generation. We show that dynamic focusing and the large horizontal displacements, characteristic of strike-slip earthquakes on long faults, are critical drivers for the tsunami hazard. These findings point to intrinsic mechanisms for sizeable tsunami generation by strike–slip faulting, which do not require complex seismic sources, landslides, or complicated bathymetry. Furthermore, our model identifies three distinct phases in the tsunamic motion; an instantaneous dynamic phase, a lagging coseismic and a classical postseismic phase, each of which may affect coastal areas differently. We conclude that near-source tsunami hazards and risk from strike-slip faulting need to be re–evaluated