The research, education, and business communities in Australia and New Zealand must work together to address the innovation challenges presented by current and disruptive technologies.
How do we work together technologically to create an environment for innovation and development, while waiting for changing conditions?
Startups and established organisations can cooperate with universities to ensure that the next generation of engineers and scientists can implement concepts rapidly and purposefully. Together we can grow innovative ideas into world-class innovative products.
Electromagnetic radiation attenuates rapidly in the ocean, leaving sound as the only practical means of remotely sensing that environment—something that has been exploited by marine animals for millions of years. Human activities are now making a substantial contribution to the background noise in the ocean, and understanding how that additional noise affects the animals that live there is now an important research topic. The Centre for Marine Science and Technology, Curtin University (CMST) addresses this through the prediction and measurement of underwater sound levels from various human activities, and experiments aimed at quantifying the impacts of sound on marine animals. Additionally they use both passive (listening only) and active (transmitting and receiving) sonar systems to find out more about the animals themselves. These tasks require extensive, highly specialised data analysis, numerical computation, and data visualisation which we carry out using MATLAB®. This talk provides an overview of some of the tasks that researchers at CMST regularly carry out using MATLAB and shows examples of the results.
Most research laboratories rely on digital tools to collect and store data, run analyses, and publish results. Technical computing software such as MATLAB® can greatly expand the research capabilities and output of a lab; however, a bottleneck is created as each new researcher must learn to use existing (often uncommented) programs. University of Melbourne’s approach is to offer free, open-access beginner- and intermediate-level courses to researchers who need to learn MATLAB for their work. The success of these courses is attributed to the low-key atmosphere, as the trainers are all research students who volunteer their time, and to the “gamification” of the learning process, with regular code challenges interspersed throughout the classes. MATLAB is an ideal teaching platform, due to its extensive documentation files and playful nature (for example, Cody Challenges).
Open-access teaching works well primarily because it turns the learning process into a community-building exercise, where the community is based around learning and sharing coding skills. The greatest advantage when it comes to learning MATLAB is the highly active and resourceful online community. Replicating this communal environment in a lab or university is a key goal towards improving researchers’ MATLAB software skills and consequently their productivity.
Sports technology research has recently become popular through the advent of personalised monitoring with wearable technology. It’s at the intersection of sports science and engineering, with a lot of overlap into the health sciences and biomedical engineering. The SABEL Labs research group at Griffith University has been active in this area for over 15 years and needed to move with agility to respond to growing demand as well as to operate in the interdisciplinary environment, all set against a backdrop of tighter research budgets.
Some time ago, when SABEL Labs developed a standard sensor platform (SABEL Sense), theydecided to use and develop common tools and data standards in MATLAB®. While SABEL Sense is primarily a 9DOF inertial sensor platform, it has plenty of flexibility for additional sensors and customization and has seen a much wider utility in engineering research programmes as well. To ensure laboratory data is collected in a common framework a customized structure (Athdata), they developed a MATLAB toolbox (ADAT) and accompanying data analysis GUI, which have more recently evolved into a cloud-based solution. This approach ensures data and analysis techniques can be shared among members, which leads to a dramatic increase in operational efficiency, reduced on-boarding time for new research students, and faster execution of externally supported research activities, allowing SABEL Labs to meet the often time-critical demands of their partners.
SABEL Labs’ research partners include elite sports institutes, sports and health science researchers, industry organisations, and startups. For these partners, solutions need to be robust, user friendly, and highly customizable, rather than technology centric. MATLAB was central to building intimacy where individual requirements for simplified visual interfaces, graded access to data detail, heavy customization, and integration of other data sources were key. The flexibility of MATLAB also helped partners to meet their appetite for growing technological literacy.
This presentation details how MATLAB has been a core component in increasing operational effectiveness and growing intimacy with research partners, and has grown as we have. It also shares some successes in the sports world as illustrations along the way.
Although our sense of sight feels effortless, recognising objects and navigating complex environments are hard computational problems involving around half of the human brain. Steven’s work is on trying to understand how the brain does it, why it goes wrong, and what we can do to help when it does. This is done by quantifying the limits of our ability to perform simple visual tasks (like recognising a face, or judging the brightness of an object), comparing results with computational models, and figuring out what improves performance of the model. This discipline is known as visual psychophysics. In this talk, Steven describes a few of his lab’s projects and how MATLAB® is used to achieve them. A recent example is concerned with measuring acuity—the smallest letter you can read—in children, using a computer tablet. The lab first developed a new set of child-friendly characters that they optimised using modelling in MATLAB (to quantify mutual confusability) and psychophysical experiments (using PsychToolbox with MATLAB). A long-standing problem with testing children’s vision is that they fidget and do not maintain constant viewing distance. The lab’s procedure overcomes this by using the tablet’s front-facing camera to track the child’s head (using the KLT algorithm, part of Computer Vision System Toolbox™). This provides a moment-to-moment estimate of the child’s physical distance from the screen that can be used to correct estimates of acuity. Finally, MATLAB Compiler™ is used to deploy the program as an executable file that runs on suitable machines without sharing source code.
This talk highlights the capabilities and features of the most recent version of these robotics toolboxes, which are now over 20 years old. Capabilities include kinematics and dynamics for robot manipulator arms; modelling, path planning, localisation, mapping, and SLAM for mobile robots; basic image processing and feature extraction; camera geometric modelling; multiview geometry; and vision-based control.
Undergraduate students at University of Technology, Sydney undertake a year-long capstone project in which they apply the skills and knowledge acquired in their coursework to a practical project. It is an opportunity for students to demonstrate the level of performance expected of a professional engineer. MATLAB® can be a highly effective tool for training in civil engineering students without inundating them with low-level programming details. The projects shown in this presentation are examples of capstone projects that students have completed using MATLAB to generate useful recommendations for professional and practising civil engineers.
Two common methods for integrating safety factors into design of shallow foundations and earth-retaining structures against failure are the traditional global factor of safety and the partial factors of safety. In this study, through MATLAB programming, these methods were compared to assess the variability of design outcomes, and identify the conservativeness of the methods. For both methods, many types of foundations and retaining walls were analysed by applying various loading, geometry, and soil conditions. To achieve this comparison and develop design calculations with appropriate graphs and tables, detailed computer models were developed using Symbolic Math Toolbox™ and Optimization Toolbox™.
The models were developed to automate design calculations and create a graphical user interface to be applied to various design test cases, allowing effective comparison between different methods. Various sections of the models were developed and validated individually, and then integrated to form the overall program. Through the developed codes in MATLAB and the direct comparison of the results, advantages and disadvantages of both methods have been identified in terms of safety, commercial, and design contexts.
Teaching, research, and learning: these are the core tasks for any educational institution. As in industry, academics are striving to deliver these tasks better than they’ve been done before.
There are, of course, challenges. For teaching academics, an ongoing challenge is to motivate and stimulate students—to constantly look for new ways to communicate technical concepts and make strong associations with the applications of those concepts. For researchers, the challenge is to maximise the time and efforts of advancing the status quo—to access and apply leading-edge algorithms.
MATLAB® Release 2016a includes capabilities that will have an immediate impact on both how you communicate with your students and the productivity of your research. In this session, Bradley demonstrates the R2016a features that support teaching and research in machine learning, acoustics signal processing, and hardware connectivity for computer vision applications.
New in R2016a is the MATLAB Live Editor – a game changer for creating MATLAB scripts. The MATLAB Live Editor provides a new way to create, edit, and run MATLAB code. See your results together with the code that produced them. Add equations, images, hyperlinks, and formatted text to enhance your narrative. You can also share with others as interactive documents.
11:45 a.m.–12:30 p.m.
Machine learning is an integral part of data analytics, which deals with developing data-driven insights for better designs and decisions. In many data analytics applications, machine learning models are either deployed to the web or databases or integrated into enterprise systems for on-demand analytics. MATLAB® can help you rapidly develop, evaluate, iterate, and choose the best predictive analytics algorithm. It can easily integrate with your software, hardware, and operational processes. This session uncovers methods to:
- Handle structured and unstructured data
- Extract and preprocess text, signal, and image data
- Statistically uncover hidden patterns in data
- Discuss scaling machine learning algorithms
This session introduces Cody™, the web-based MATLAB® assignment grading tool. Through Cody, teaching staff can pose quizzes, tutorials, and assignments based in MATLAB and see the progress made by their students via automatic scoring charts and reports.
The new self-serve library of existing questions allows lecturers to quickly build their own quizzes and assignments, which can supplement existing assessment schedules. In this session, Bradley demonstrates the library, and attendees who bring their laptops can participate in a live Cody assignment.
Bradley also reviews the free MATLAB curriculum packages that teachers can access. Although it is an essential part of the teaching process, the development of curriculum material is often a time-consuming task. Learn how to supplement your own teaching material with content developed by your academic peers from around the globe (Moler, Strang, Palm, Chapra, etc.). During this session, Bradley reviews MATLAB courseware, an archive of science and engineering curriculum made freely available to educators.
Computer vision is an enabling technology that is driving the development of self-driving cars, augmented reality, autonomous robots, and other smart systems. Computer vision applies complex algorithms to images and video to detect, classify, and track objects or events in order to understand a real-world scene. MATLAB® can simplify the computer vision system design workflow from algorithm development to implementation on embedded systems. This session provides insight into:
- Deep learning and traditional machine learning for object classification
- 3D point cloud processing, stereo vision, and structure from motion for visual perception
This session reviews how engineering and science students use software simulation tools to develop a deeper understanding of complex multidomain applications. A quadcopter UAV example is used to showcase how the fundamental mathematics concepts introduced in the earlier years of a science or engineering degree work hand in hand with the higher-level numerical methods and control design concepts taught in the later years.
The session demonstrates:
- How the new MATLAB® Live Editor can demystify rotation matrix sequences and enliven the teaching of concepts associated with moments of inertia
- How Simulink® can model and solve the 6-DOF equations of motion of a rigid body (building, solving, implementing equations, and incorporating the rotation matrix concepts along the way)
- How to linearize a 6-DOF model and design controllers for altitude and attitude controls
Quite often, we run into situations where we are dealing with large data sets or running algorithms that are computationally expensive, or both. Distributing the computations to the available hardware resources is one approach to speed up the response time of our algorithms. In this session Mandar:
- Provides an overview of MATLAB® for distributed computing
- Demonstrates how users can distribute their computations on local and remote resources
- Discusses techniques to handle large data sets in MATLAB
- Presents ways to scale up applications to compute on the cloud, such as Amazon EC2
In this session, Bradley demonstrates the capabilities of the new Robotics System Toolbox™ and they enrich an engineering curriculum. In an era when many universities are struggling to maintain enrollments in the harder sciences, academics are continuously looking for ways to amplify the excitement of science and engineering. For many, a key factor for captivating the interest of students is through hardware connectivity and the development of autonomous mobile robotics applications. The challenge to this approach has been the prerequisite skills needed to implement “exciting” algorithms, and the interfacing and communication with the robotics platform.
In Robotics Systems Toolbox, design tasks associated with autonomous vehicles now become accessible to students at all levels, from undergraduate through postgraduate, by using MATLAB®. With some rudimentary MATLAB skills, even first-year students can experience the thrill of designing and implementing a task for an autonomous mobile robot. Robotics Systems Toolbox capabilities demonstrated in the session include:
- Map representation
- Path planning and following for differential drive robots
- Interfacing with the Robot Operating System (ROS) and ROS-enabled simulators such as Gazebo
Wireless engineering teams use MATLAB® to reduce development time from algorithm development through full system simulation and hardware implementation. In this session, Mandar shows:
- LTE and 802.11 a/b/g/n/ac simulation, signal generation, design verification, and performance assessment
- Live waveform transmission and reception with SDR hardware
- Carrier aggregation, beamforming, and antenna array modelling for MIMO systems
- Signal analysis and information recovery
- Design antennas for specific applications