Enrico Di Lello

Towards safe human-robot interaction in robotic cells: An approach based on visual tracking and intention estimation

Removing the safety fences that separate humans and robots, to allow for an effective human-robot interaction, requires innovative safety control systems. An advanced functionality of a safety controller might be to detect the presence of humans entering the robotic cell and to estimate their intention, in order to enforce an effective safety reaction. This paper proposes advanced algorithms for cognitive vision, empowered by a dynamic model of human walking, for detection and tracking of humans. Intention estimation is then addressed as the problem of predicting online the trajectory of the human, given a set of trajectories of walking people learnt offline using an unsupervised classification algorithm. Results of the application of the presented approach to a large number of experiments on volunteers are also reported.

Bayesian time-series models for continuous fault detection and recognition in industrial robotic tasks

This paper presents the application of a Bayesian nonparametric time-series model to process monitoring and fault classification for industrial robotic tasks. By means of an alignment task performed with a real robot, we show how the proposed approach allows to learn a set of sensor signature models encoding the spatial and temporal correlations among wrench measurements recorded during a number of successful task executions. Using these models, it is possible to detect continuously and on-line deviations from the expected sensor readings. Separate models are learned for a set of possible error scenarios involving a human modifying the workspace configuration. These non-nominal task executions are correctly detected and classified with an on-line algorithm, which opens the possibility for the development of error-specific recovery strategies. Our work is complementary to previous approaches in robotics, where process monitors based on probabilistic models, but limited to contact events, were developed for control purposes. Instead, in this paper we focus on capturing dynamic models of sensor signatures throughout the whole task, therefore allowing continuous monitoring and extending the system ability to interpret and react to errors.

Recognition of 6 DOF rigid body motion trajectories using a coordinate-free representation

This paper presents an approach to recognize 6 DOF rigid body motion trajectories (3D translation + rotation), such as the 6 DOF motion trajectory of an object manipulated by a human. As a first step in the recognition process, 3D measured position trajectories of arbitrary and uncalibrated points attached to the rigid body are transformed to an invariant, coordinate-free representation of the rigid body motion trajectory. This invariant representation is independent of the reference frame in which the motion is observed, the chosen marker positions, the linear scale (magnitude) of the motion, the time scale and the velocity profile along the trajectory. Two classification algorithms which use the invariant representation as input are developed and tested experimentally: one approach based on a Dynamic TimeWarping algorithm, and one based on Hidden Markov Models. Both approaches yield high recognition rates (up to 95 % and 91 %, respectively). The advantage of the invariant approach is that motion trajectories observed in different contexts (with different reference frames, marker positions, time scales, linear scales, velocity profiles) can be compared and averaged, which allows us to build models from multiple demonstrations observed in different contexts, and use these models to recognize similar motion trajectories in still different contexts.

Probabilistic gait classification in children with cerebral palsy: A Bayesian approach

Three-dimensional gait analysis (3DGA) generates a wealth of highly variable data. Gait classifications help to reduce, simplify and interpret this vast amount of 3DGA data and thereby assist and facilitate clinical decision making in the treatment of CP. CP gait is often a mix of several clinically accepted distinct gait patterns. Therefore, there is a need for a classification which characterizes each CP gait by different degrees of membership for several gait patterns, which are considered by clinical experts to be highly relevant. In this respect, this paper introduces Bayesian networks (BN) as a new approach for classification of 3DGA data of the ankle and knee in children with CP. A BN is a probabilistic graphical model that represents a set of random variables and their conditional dependencies via a directed acyclic graph. Furthermore, they provide an explicit way of introducing clinical expertise as prior knowledge to guide the BN in its analysis of the data and the underlying clinically relevant relationships. BNs also enable to classify gait on a continuum of patterns, as their outcome consists of a set of probabilistic membership values for different clinically accepted patterns. A group of 139 patients with CP was recruited and divided into a training- (n=80% of all patients) and a validation-dataset (n=20% of all patients). An average classification accuracy of 88.4% was reached. The BN of this study achieved promising accuracy rates and was found to be successful for classifying ankle and knee joint motion on a continuum of different clinically relevant gait patterns.

Augmented Switching Linear Dynamical System Model for Gas Concentration Estimation with MOX Sensors in an Open Sampling System

In this paper, we introduce a Bayesian time series model approach for gas concentration estimation using Metal Oxide (MOX) sensors in Open Sampling System (OSS). Our approach focuses on the compensation of the slow response of MOX sensors, while concurrently solving the problem of estimating the gas concentration in OSS. The proposed Augmented Switching Linear System model allows to include all the sources of uncertainty arising at each step of the problem in a single coherent probabilistic formulation. In particular, the problem of detecting on-line the current sensor dynamical regime and estimating the underlying gas concentration under environmental disturbances and noisy measurements is formulated and solved as a statistical inference problem. Our model improves, with respect to the state of the art, where system modeling approaches have been already introduced, but only provided an indirect relative measures proportional to the gas concentration and the problem of modeling uncertainty was ignored. Our approach is validated experimentally and the performances in terms of speed of and quality of the gas concentration estimation are compared with the ones obtained using a photo-ionization detector.

Robotic furniture in a smart environment : the PEIS table

According to a recent trend, robotic technologies will be included into domestic environments in the form of simple, networked robotic devices able to cooperate in the performance of tasks. These devices may take the form of smart appliances, distributed sensors, or robotic furniture. In this paper, we describe the design of an autonomous robotic table and its inclusion in a smart environment, the PEIS Ecology. The design takes into account the constraints posed by the domestic environment. The robotic table can perform autonomous point-to-point navigation, and it can collaborate with the other devices in the ecology to perform complex tasks that go beyond simple navigation.