Chris Melhuish

Whiskerbot: A Robotic Active Touch System Modeled on the Rat Whisker Sensory System

The Whiskerbot project is a collaborative project between robotics engineers, computational neuroscientists and ethologists, aiming to build a biologically inspired robotic implementation of the rodent whisker sensory system. The morphology and mechanics of the large whiskers (macro-vibrissae) have been modeled, as have the neural structures that constitute the rodent central nervous system responsible for macro-vibrissae sensory processing. There are two principal motivations for this project. First, by implementing an artificial whisker sensory system controlled using biologically plausible neural networks we hope to test existing models more thoroughly and develop new hypotheses for vibrissal sensory processing. Second, the sensory mode of tactile whiskers could be useful for general mobile robotic sensory deployment. In this article the robotic platform that has been built is detailed as well as some of the experiments that have been conducted to test the neural control algorithms and architectures inspired from neuroethological observations to mediate adaptive behaviors.

Extracting textural features from tactile sensors

This paper describes an experiment to quantify texture using an artificial finger equipped with a microphone to detect frictional sound. Using a microphone to record tribological data is a biologically inspired approach that emulates the Pacinian corpuscle. Artificial surfaces were created to constrain the subsequent analysis to specific textures. Recordings of the artificial surfaces were made to create a library of frictional sounds for data analysis. These recordings were mapped to the frequency domain using fast Fourier transforms for direct comparison, manipulation and quantifiable analysis. Numerical features such as modal frequency and average value were calculated to analyze the data and compared with attributes generated from principal component analysis (PCA). It was found that numerical features work well for highly constrained data but cannot classify multiple textural elements. PCA groups textures according to a natural similarity. Classification of the recordings using k nearest neighbors shows a high accuracy for PCA data. Clustering of the PCA data shows that similar discs are grouped together with few classification errors. In contrast, clustering of numerical features produces erroneous classification by splitting discs between clusters. The temperature of the finger is shown to have a direct relation to some of the features and subsequent data in PCA.

The BERT2 infrastructure: An integrated system for the study of human-robot interaction

Bristol Elumotion Robot Torso Version 2 (BERT2) is a humanoid robot currently in development at Bristol Robotics Laboratory (BRL). In this paper we present the current state of development and demonstrate how the integration of several advanced subsystems (of commercial and non-commercial nature) within a heterogeneous computing infrastructure enables us to construct a unique platform ideally suited to investigate complex human-robot interaction (HRI). We particularly focus on two important domains of non-verbal communication, namely gaze and pointing gestures in a real-world 3D setting and outline our thinking in terms of safety, ambiguities and further experimental work.

Sensitivity Analysis of a Parametric Hand Exoskeleton Designed to Match Natural Human Grasping Motion

This paper describes the simulated analysis of a fully scalable, parametrically designed hand exoskeleton previously developed as part of a stroke rehabilitation program within the Bristol Robotics Laboratory. The device is parametrically designed to match the location and trajectories of the joints within a normal healthy human hand. However, testing of fully scalable designs which can be custom fit to a person using parametric design can be costly, time consuming and potentially hazardous if ill-fitting. Here a method is presented which allows for the performance of a parametric design to be tested. A virtual mechanism with induced manufacturing tolerances is modelled and its interactions with the hand are simulated. The performance can then be assessed by the devices ability to achieve the objective trajectory within the simulation. The results show that for the designed hand exoskeleton, with a manufacturing tolerance of 0.2mm across parts the resulting average trajectory error is less than 0.2 degrees with an average tip error of less than 0.5 mm. The results also demonstrate that for a large tolerance of 1mm across all dimensions, the trajectory error can reach as high as 30.9 degrees. This result justifies the use of parametric design to develop mechanisms matching natural human motion. While the results are for a parametrically scalable hand exoskeleton, it is believed the methodology is applicable to any bio-compatible assistive device.

Assessment of human response to robot facial expressions through visual evoked potentials

The focus of this work is to investigate and quantify the ability of a humanoid ‘hybrid face’ robot to effectively convey emotion to a human observer by mapping their physiological (EEG) response to perceived emotional information. Specifically, we examine the event related response during two implicit emotion recognition experiments to determine the modulation of the face-specific N170 brain response component to robot facial expressions. EEG recordings were taken from a range of test subjects observing the BERT2 robot cycle through a range of facial emotions in each emotion recognition experiment. Results from both experiments demonstrate that the stimuli evoke the N170 component and that digital facial expressions with high correlations can be discriminated. Emotional expressions evoke a larger response relative to neutral stimuli, with negative evoking an increased amplitude and latency to positive emotions, and demonstrate that the response to robot facial expressions evoke similar brain activity to that of a human emotions. This study is the first of its nature to investigate and quantify the human physiological response to digital facial expressions as conveyed in real-time by a humanoid robot.

A Q-learning based Cartesian model reference compliance controller implementation for a humanoid robot arm

This paper presents the implementation (real time and simulation) of a model-free Q-learning based discrete model reference compliance controller for a humanoid robot arm. The Reinforcement learning (RL) scheme uses a recently developed Q-learning scheme to develop an optimal policy on-line. The RL Cartesian (x and y) tracking controller with model reference compliance was implemented using two links (shoulder flexion and elbow flexion joints) of the right arm of the humanoid Bristol-Elumotion-Robotic-Torso II (BERT II) torso.

A Novel Approach of Robust Active Compliance for Robot Fingers

In order to guarantee that grasping with robot fingers are safe when interacting with a human or a touched object, the robot fingers have to be compliant. In this study, a novel active and robust compliant control technique is proposed by employing an Integral Sliding Mode Control (ISMC). The ISMC allows us to use a model reference approach for which a virtual mass-spring damper can be introduced to enable compliant control. The performance of the ISMC is validated for the constrained underactuated BERUL (Bristol Elumotion Robot fingers) fingers. The results show that the approach is feasible for compliance interaction with objects of different softness. Moreover, the compliance results show that the ISMC is robust towards nonlinearities and uncertainties in the robot fingers in particular friction and stiction.

Advances in Autonomous Robotics Systems

AIMS project attempts to link the logistic requirements of an intelligent warehouse and state of the art core technologies of automation, by providing an awareness of the environment to the autonomous systems and vice versa. In this work we investigate a solution for modeling the infrastructure of a structured environment such as warehouses, by the means of a vision sensor. The model is based on the expected pattern of the infrastructure, generated from and matched to the map. Generation of the model is based on a set of tools such as closed-form Hough transform, DBSCAN clustering algorithm, Fourier transform and optimization techniques. The performance evaluation of the proposed method is accompanied with a real world experiment.We present a novel software tool intended for mobile robot mapping in long-term scenarios. The method allows for efficient volumetric representation of dynamic three-dimensional environments over long periods of time. It is based on a combination of a well-established 3D mapping framework called Octomaps and an idea to model environment dynamics by its frequency spectrum. The proposed method allows not only for efficient representation, but also reliable prediction of the future states of dynamic three-dimensional environments. Our spatio-temporal mapping framework is available as an open-source C++ library and a ROS module which allows its easy integration in robotics projects.

Bioinspired Control of Electro-Active Polymers for Next Generation Soft Robots

The emerging field of soft robotics offers the prospect of replacing existing hard actuator technologies with new soft-smart materials [7]. Such materials have the potential to form a key component of safer, more compliant and light-weight robots. Soft robots constructed from these advanced materials could be used in a progressively wide range of applications, especially those involving interactions between robots and people in unstructured environments such as homes, hospitals and schools. Electroactive polymer (EAP) technologies such as dielectric elastomer (DEA) actuators and ionic polymer-metal composites (IPMCs) are a class of smart materials that are of particular interest for use in soft robotics [2]. However, despite their great potential, EAP devices present a number of challenges for control. They are, for example, non-linear in behaviour, prone to degradation over time, and fabricated with wide tolerances. In this paper we describe a project that aims to develop novel bioinspired control strategies for EAPs addressing these key challenges.

Erratum: Reinforcement learning and optimal adaptive control: An overview and implementation examples (Annual Reviews in Control (2012) 36,1 (42-59))

The authors regret that in the above published article the reference details for the citations Bucak and Zohdy (2001) and Bucak and Zohdy (1999) were incorrect. The correct reference details should appear as follows: Bucak, I.O., & Zohdy, M.A. (1999). Application of reinforcement learning control to a nonlinear dexterous robot. In Proceedings of the 38th IEEE Conference on Decision and Control, 1999 (Vol. 5, pp. 5108–5113). Bucak, I.O., & Zohdy, M.A. (2001). Reinforcement learning control of nonlinear multi-link system. Engineering Applications of Artificial Intelligence, 14(5), 563–575. The authors would like to apologise for any inconvenience caused.