Source author record

Ian Walker

Ian Walker appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

astro-ph.IM Artificial Intelligence astro-ph Computer Vision eess.IV Machine Learning physics.ins-det Robotics

Catalog footprint

What is connected

4works

8topics

4close collaborators

Actions

Connect this record

Open graph Browse works

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2022arXiv

A Failure Identification and Recovery Framework for a Planar Reconfigurable Cable Driven Parallel Robot

In cable driven parallel robots (CDPRs), a single cable malfunction usually induces complete failure of the entire robot. However, the lost static workspace (due to failure) can often be recovered through reconfiguration of the cable attachment points on the frame. This capability is introduced by adding kinematic redundancies to the robot in the form of moving linear sliders that are manipulated in a real-time redundancy resolution controller. The presented work combines this controller with an online failure detection framework to develop a complete fault tolerant control scheme for automatic task recovery. This solution provides robustness by combining pose estimation of the end-effector with the failure detection through the application of an Interactive Multiple Model (IMM) algorithm relying only on end-effector information. The failure and pose estimation scheme is then tied into the redundancy resolution approach to produce a seamless automatic task (trajectory) recovery approach for cable failures.

preprint2019arXiv

Causality matters in medical imaging

This article discusses how the language of causality can shed new light on the major challenges in machine learning for medical imaging: 1) data scarcity, which is the limited availability of high-quality annotations, and 2) data mismatch, whereby a trained algorithm may fail to generalize in clinical practice. Looking at these challenges through the lens of causality allows decisions about data collection, annotation procedures, and learning strategies to be made (and scrutinized) more transparently. We discuss how causal relationships between images and annotations can not only have profound effects on the performance of predictive models, but may even dictate which learning strategies should be considered in the first place. For example, we conclude that semi-supervision may be unsuitable for image segmentation---one of the possibly surprising insights from our causal analysis, which is illustrated with representative real-world examples of computer-aided diagnosis (skin lesion classification in dermatology) and radiotherapy (automated contouring of tumours). We highlight that being aware of and accounting for the causal relationships in medical imaging data is important for the safe development of machine learning and essential for regulation and responsible reporting. To facilitate this we provide step-by-step recommendations for future studies.

preprint2015arXiv

A passive THz video camera based on lumped element kinetic inductance detectors

We have developed a passive 350 GHz (850 μm) video-camera to demonstrate lumped element kinetic inductance detectors (LEKIDs) -- designed originally for far-infrared astronomy -- as an option for general purpose terrestrial terahertz imaging applications. The camera currently operates at a quasi-video frame rate of 2 Hz with a noise equivalent temperature difference per frame of $\sim$0.1 K, which is close to the background limit. The 152 element superconducting LEKID array is fabricated from a simple 40 nm aluminum film on a silicon dielectric substrate and is read out through a single microwave feedline with a cryogenic low noise amplifier and room temperature frequency domain multiplexing electronics.

preprint2004arXiv

An update on the SCUBA-2 project

SCUBA-2, which replaces SCUBA (the Submillimeter Common User Bolometer Array) on the James Clerk Maxwell Telescope (JCMT) in 2006, is a large-format bolometer array for submillimeter astronomy. Unlike previous detectors which have used discrete bolometers, SCUBA-2 has two dc-coupled, monolithic, filled arrays with a total of ~10,000 bolometers. It will offer simultaneous imaging of a 50 sq-arcmin field of view at wavelengths of 850 and 450 microns. SCUBA-2 is expected to have a huge impact on the study of galaxy formation and evolution in the early Universe as well as star and planet formation in our own Galaxy. Mapping the sky to the same S/N up to 1000 times faster than SCUBA, it will also act as a pathfinder for the new submillimeter interferometers such as ALMA. SCUBA-2's absorber-coupled pixels use superconducting transition edge sensors operating at 120 mK for performance limited by the sky background photon noise. The monolithic silicon detector arrays are deep-etched by the Bosch process to isolate the pixels on silicon nitride membranes. Electrical connections are made through indium bump bonds to a SQUID time-domain multiplexer (MUX). We give an overview of the SCUBA-2 system and an update on its status, and describe some of the technological innovations that make this unique instrument possible.