Dive mechanic: Bringing 3D virtual experimentation to elite level diving using the Workspace workflow engine

Abstract

Dive Mechanic is a custom 3D software application for modelling elite level diving. It is designed to assist coaches, athletes and sports biomechanists in understanding, visualising and improving diving performance for both platform and springboard diving. This application was developed by research scientists within CSIRO Data61 and was built upon the Workspace workflow engine. Biomechanics algorithms were developed within this graphical workflow environment and an intuitive graphical user interface (GUI) was developed to make the software accessible to end users. Workspace also simplified the packaging of Dive Mechanic by providing an installer and access to software licensing controls. This paper outlines how the Workspace workflow engine assisted in the process of both developing scientific algorithms and encapsulating them within a commercial product offering. Dive Mechanic was developed by CSIRO for use by Diving Australia as part of a larger project to investigate diving technique and injury risk. The aerial dynamics portion of the modelling was identified as sufficiently computationally-cheap to facilitate an interactive software solution. Dive Mechanic uses personalised body and motion data based on laser scans of the athletes and markerless motion digitisation from multiple camera video of the athletes performing dives. It employs a fast and detailed human biomechanics motion engine specifically developed for the application that allows diver technique to be modified interactively by coaches in order to explore technique or synchronisation improvements. Its GUI is powerful yet sufficiently simple and intuitive so as to provide a very positive user experience for non-IT diving experts. Dive Mechanic’s virtual experimentation offers advantages over physical experimentation in that it can help avoid injury and having divers training possibly worse techniques whilst experimenting. Dive Mechanic was used by coaches of the Australian Olympic diving squad in preparation for the Rio 2016 Olympics. The Workspace workflow engine was well suited to the development of Dive Mechanic. The first stage was to elicit requirements from Diving Australia and to formulate software specifications. Computational modelling of diving fits naturally within the workflow paradigm with the inputs of a physical diver model, diver technique and takeoff conditions being processed to output the complete diver aerial motion which feeds through to a visualisation workflow. From these specifications, the software architecture was determined in terms of functional blocks in a computational workflow and corresponding features in the user interface. Initial development involved creating the biomechanical file format readers, data structures, widgets and unit testing operations. Critical widgets for software usability including a kinematics modification widget and a simplified 3D OpenGL widget were then created. Next, the dynamics engine was developed and tested. Important features in the overall workflow were then connected with UI elements to produce the Dive Mechanic software. Incremental development of the workflow and the ability to easily inspect and interrogate data locally in the workflows using the same widgets used in the final product simplified software debugging and troubleshooting. Additionally the inherent modular nature of workflow operations maximised the reuse of existing operations and improved design of new reusable workflow operations by computational scientists.