On the current limitations of robotics research concerning the generalization of reported results to different set-ups

June 27, 2017 09:11 ,

Francesco Amigoni, Matteo Luperto, Viola Schiaffonati,Toward generalization of experimental results for autonomous robots, Robotics and Autonomous Systems, Volume 90, April 2017, Pages 4-14, ISSN 0921-8890, DOI: 10.1016/j.robot.2016.08.016.

In this paper we discuss some issues in the experimental evaluation of intelligent autonomous systems, focusing on systems, like autonomous robots, operating in physical environments. We argue that one of the weaknesses of current experimental practices is the low degree of generalization of experimental results, meaning that knowing the performance a robot system obtains in a test setting does not provide much information about the performance the same system could achieve in other settings. We claim that one of the main obstacles to achieve generalization of experimental results in autonomous robotics is the low degree of representativeness of the selected experimental settings. We survey and discuss the degree of representativeness of experimental settings used in a significant sample of current research and we propose some strategies to overcome the emerging limitations.

Posted in: Robotics , Tagged:

Robots that pre-compute a number of possible behaviours (in simulation) and then learn their performance with them (propragating that performance measures to similar behaviors through Gaussian Processes Regression) and select the best at each situation (through Bayesian Optimization), thus confronting varying environments and damages to the robot

June 27, 2017 09:02 ,

A. Cully, et al. Robots that can adapt like animals, Nature, 521 (2015), pp. 503–507, DOI: 10.1038/nature14422.

Robots have transformed many industries, most notably manufacturing, and have the power to deliver tremendous benefits to society, such as in search and rescue, disaster response, health care and transportation. They are also invaluable tools for scientific exploration in environments inaccessible to humans, from distant planets to deep oceans. A major obstacle to their widespread adoption in more complex environments outside factories is their fragility. Whereas animals can quickly adapt to injuries, current robots cannot think outside the box to find a compensatory behaviour when they are damaged: they are limited to their pre-specified self-sensing abilities, can diagnose only anticipated failure modes, and require a pre-programmed contingency plan for every type of potential damage, an impracticality for complex robots. A promising approach to reducing robot fragility involves having robots learn appropriate behaviours in response to damage, but current techniques are slow even with small, constrained search spaces. Here we introduce an intelligent trial-and-error algorithm that allows robots to adapt to damage in less than two minutes in large search spaces without requiring self-diagnosis or pre-specified contingency plans. Before the robot is deployed, it uses a novel technique to create a detailed map of the space of high-performing behaviours. This map represents the robotâ €™ s prior knowledge about what behaviours it can perform and their value. When the robot is damaged, it uses this prior knowledge to guide a trial-and-error learning algorithm that conducts intelligent experiments to rapidly discover a behaviour that compensates for the damage. Experiments reveal successful adaptations for a legged robot injured in five different ways, including damaged, broken, and missing legs, and for a robotic arm with joints broken in 14 different ways. This new algorithm will enable more robust, effective, autonomous robots, and may shed light on the principles that animals use to adapt to injury.

Posted in: Robotics , Tagged: , ,

How to improve statistical results obtained from limited set-ups through active sampling, and a nice review of possible pitfalls in conducting statistical research (and a mention to “pre-registration” of hypothesis and plans to be peer-reviewed before submitting the results)

June 27, 2017 08:42 ,

Romy Lorenz, Adam Hampshire, Robert Leech, Neuroadaptive Bayesian Optimization and Hypothesis Testing, Trends in Cognitive Sciences, Volume 21, Issue 3, March 2017, Pages 155-167, ISSN 1364-6613, DOI: 10.1016/j.tics.2017.01.006.

Cognitive neuroscientists are often interested in broad research questions, yet use overly narrow experimental designs by considering only a small subset of possible experimental conditions. This limits the generalizability and reproducibility of many research findings. Here, we propose an alternative approach that resolves these problems by taking advantage of recent developments in real-time data analysis and machine learning. Neuroadaptive Bayesian optimization is a powerful strategy to efficiently explore more experimental conditions than is currently possible with standard methodology. We argue that such an approach could broaden the hypotheses considered in cognitive science, improving the generalizability of findings. In addition, Bayesian optimization can be combined with preregistration to cover exploration, mitigating researcher bias more broadly and improving reproducibility.

Posted in: Cognitive sciences, Probability and statistics , Tagged: ,

Value iteration applied in control systems when the model of the plant is substituted by data acquired from the plant

June 27, 2017 08:33 ,

Yongqiang Li, Zhongsheng Hou, Yuanjing Feng, Ronghu Chi, Data-driven approximate value iteration with optimality error bound analysis, Automatica, Volume 78, April 2017, Pages 79-87, ISSN 0005-1098, DOI: 10.1016/j.automatica.2016.12.019.

Features of the data-driven approximate value iteration (AVI) algorithm, proposed in Li et al. (2014) for dealing with the optimal stabilization problem, include that only process data is required and that the estimate of the domain of attraction for the closed-loop is enlarged. However, the controller generated by the data-driven AVI algorithm is an approximate solution for the optimal control problem. In this work, a quantitative analysis result on the error bound between the optimal cost and the cost under the designed controller is given. This error bound is determined by the approximation error of the estimation for the optimal cost and the approximation error of the controller function estimator. The first one is concretely determined by the approximation error of the data-driven dynamic programming (DP) operator to the DP operator and the approximation error of the value function estimator. These three approximation errors are zeros when the data set of the plant is sufficient and infinitely complete, and the number of samples in the interested state space is infinite. This means that the cost under the designed controller equals to the optimal cost when the number of iterations is infinite.

NOTE: Another paper on the same issue in the same journal.

Posted in: Applications of reinforcement learning to control engineering , Tagged: ,

A study of the influence of uncertain, stochastic delays in the stability of LTI SISO systems

June 22, 2017 11:54 ,

T. Qi, J. Zhu and J. Chen, “Fundamental Limits on Uncertain Delays: When Is a Delay System Stabilizable by LTI Controllers?,” in IEEE Transactions on Automatic Control, vol. 62, no. 3, pp. 1314-1328, March 2017. DOI: 10.1109/TAC.2016.2584007.

This paper concerns the stabilization of linear time-invariant (LTI) systems subject to uncertain, possibly time-varying delays. The fundamental issue under investigation, referred to as the delay margin problem, addresses the question: What is the largest range of delay such that there exists a single LTI feedback controller capable of stabilizing all the plants for delays within that range? Drawing upon analytic interpolation and rational approximation techniques, we derive fundamental bounds on the delay margin, within which the delay plant is guaranteed to be stabilizable by a certain LTI output feedback controller. Our contribution is threefold. First, for single-input single-output (SISO) systems with an arbitrary number of plant unstable poles and nonminimum phase zeros, we provide an explicit, computationally efficient bound on the delay margin, which requires computing only the largest real eigenvalue of a constant matrix. Second, for multi-input multi-output (MIMO) systems, we show that estimates on the variation ranges of multiple delays can be obtained by solving LMI problems, and further, by finding bounds on the radius of delay variations. Third, we show that these bounds and estimates can be extended to systems subject to time-varying delays. When specialized to more specific cases, e.g., to plants with one unstable pole but possibly multiple nonminimum phase zeros, our results give rise to analytical expressions exhibiting explicit dependence of the bounds and estimates on the pole and zeros, thus demonstrating how fundamentally unstable poles and nonminimum phase zeros may limit the range of delays over which a plant can be stabilized by a LTI controller.

Posted in: Control Engineering , Tagged: ,

Very efficient global non-linear optimization for real-time robotic problems through the re-use of pre-computed solutions

June 22, 2017 11:22 ,

K. Hauser, “Learning the Problem-Optimum Map: Analysis and Application to Global Optimization in Robotics,” in IEEE Transactions on Robotics, vol. 33, no. 1, pp. 141-152, Feb. 2017. DOI: 10.1109/TRO.2016.2623345.

This paper describes a data-driven framework for approximate global optimization in which precomputed solutions to a sample of problems are retrieved and adapted during online use to solve novel problems. This approach has promise for real-time applications in robotics, since it can produce near globally optimal solutions orders of magnitude faster than standard methods. This paper establishes theoretical conditions on how many and where samples are needed over the space of problems to achieve a given approximation quality. The framework is applied to solve globally optimal collision-free inverse kinematics problems, wherein large solution databases are used to produce near-optimal solutions in a submillisecond time on a standard PC.

Posted in: Mathematics , Tagged:

Varying the number of particles in a PF in order to improve the speed of convergence, with a short related work about adapting the number of particles for other goals

June 22, 2017 10:51 ,

V. Elvira, J. Míguez and P. M. Djurić, “Adapting the Number of Particles in Sequential Monte Carlo Methods Through an Online Scheme for Convergence Assessment,” in IEEE Transactions on Signal Processing, vol. 65, no. 7, pp. 1781-1794, April1, 1 2017. DOI: 10.1109/TSP.2016.2637324.

Particle filters are broadly used to approximate posterior distributions of hidden states in state-space models by means of sets of weighted particles. While the convergence of the filter is guaranteed when the number of particles tends to infinity, the quality of the approximation is usually unknown but strongly dependent on the number of particles. In this paper, we propose a novel method for assessing the convergence of particle filters in an online manner, as well as a simple scheme for the online adaptation of the number of particles based on the convergence assessment. The method is based on a sequential comparison between the actual observations and their predictive probability distributions approximated by the filter. We provide a rigorous theoretical analysis of the proposed methodology and, as an example of its practical use, we present simulations of a simple algorithm for the dynamic and online adaptation of the number of particles during the operation of a particle filter on a stochastic version of the Lorenz 63 system.

Posted in: Bayesian filtering , Tagged:

Qualitative robot navigation

June 22, 2017 10:06 ,

Sergio Miguel-Tomé, Navigation through unknown and dynamic open spaces using topological notions, Connection Science, DOI: 10.1080/09540091.2016.1277691.

Until now, most algorithms used for navigation have had the purpose of directing system towards one point in space. However, humans communicate tasks by specifying spatial relations among elements or places. In addition, the environments in which humans develop their activities are extremely dynamic. The only option that allows for successful navigation in dynamic and unknown environments is making real-time decisions. Therefore, robots capable of collaborating closely with human beings must be able to make decisions based on the local information registered by the sensors and interpret and express spatial relations. Furthermore, when one person is asked to perform a task in an environment, this task is communicated given a category of goals so the person does not need to be supervised. Thus, two problems appear when one wants to create multifunctional robots: how to navigate in dynamic and unknown environments using spatial relations and how to accomplish this without supervision. In this article, a new architecture to address the two cited problems is presented, called the topological qualitative navigation architecture. In previous works, a qualitative heuristic called the heuristic of topological qualitative semantics (HTQS) has been developed to establish and identify spatial relations. However, that heuristic only allows for establishing one spatial relation with a specific object. In contrast, navigation requires a temporal sequence of goals with different objects. The new architecture attains continuous generation of goals and resolves them using HTQS. Thus, the new architecture achieves autonomous navigation in dynamic or unknown open environments.

Posted in: Robot motion planning , Tagged:

A new Kalman Filter that is more robust under certain deviations of the gaussian hypothesis

June 22, 2017 09:46 ,

Badong Chen, Xi Liu, Haiquan Zhao, Jose C. Principe, Maximum correntropy Kalman filter, Automatica, Volume 76, February 2017, Pages 70-77, ISSN 0005-1098, DOI: 10.1016/j.automatica.2016.10.004.

Traditional Kalman filter (KF) is derived under the well-known minimum mean square error (MMSE) criterion, which is optimal under Gaussian assumption. However, when the signals are non-Gaussian, especially when the system is disturbed by some heavy-tailed impulsive noises, the performance of KF will deteriorate seriously. To improve the robustness of KF against impulsive noises, we propose in this work a new Kalman filter, called the maximum correntropy Kalman filter (MCKF), which adopts the robust maximum correntropy criterion (MCC) as the optimality criterion, instead of using the MMSE. Similar to the traditional KF, the state mean vector and covariance matrix propagation equations are used to give prior estimations of the state and covariance matrix in MCKF. A novel fixed-point algorithm is then used to update the posterior estimations. A sufficient condition that guarantees the convergence of the fixed-point algorithm is also given. Illustration examples are presented to demonstrate the effectiveness and robustness of the new algorithm.

Posted in: Bayesian filtering , Tagged: ,

A promising survey on robust estimation methods aimed at robotic applications

June 22, 2017 09:35 ,

Michael Bosse, Gabriel Agamennoni and Igor Gilitschenski (2016), “Robust Estimation and Applications in Robotics”, Foundations and Trends® in Robotics: Vol. 4: No. 4, pp 225-269. DOI: 10.1561/2300000047.

Solving estimation problems is a fundamental component of numerous robotics applications. Prominent examples involve pose estimation, point cloud alignment, or object tracking. Algorithms for solving these estimation problems need to cope with new challenges due to an increased use of potentially poor low-cost sensors, and an ever growing deployment of robotic algorithms in consumer products which operate in potentially unknown environments. These algorithms need to be capable of being robust against strong nonlinearities, high uncertainty levels, and numerous outliers. However, particularly in robotics, the Gaussian assumption is prevalent in solutions to multivariate parameter estimation problems without providing the desired level of robustness. The goal of this tutorial is helping to address the aforementioned challenges by providing an introduction to robust estimation with a particular focus on robotics. First, this is achieved by giving a concise overview of the theory on M-estimation. M-estimators share many of the convenient properties of least-squares estimators, and at the same time are much more robust to deviations from the Gaussian model assumption. Second, we present several example applications where M-Estimation is used to increase robustness against nonlinearities and outliers.

Posted in: Probability and statistics , Tagged: