Portfolio Shabaz Sultan

Interactive Exploration of Gravity, Numerical Methods and Integration

MSc. thesis

An experiment in interactively showing the fundamentals of physics, numerical algorithms and code implementation internals in an integrated environment. Intended to offer a glimpse into what is possible in interactive teaching environment, new ways to interactively communicate both simulation results and the way a simulation's internals relate to it.

The environment allows one to interactively change initial conditions and see how this changes simulation outcomes. Or one can change integrators, integration parameters and even code a new integrator on the fly to see how this affects outcomes of a system.

The stepwise nature of an integration is made visible and explorable in the environment. The system offers a live coding option to create own integrators and get a feel for the code one is creating and its relation to a simulation in a novel way. A converging integrator is used as a reference calculation to give insight into the exact error characteristics of other integrators.

The system offers a basic explanation of fundamental concepts like the point mass abstraction and Newtonian gravity in an interactive environment, to bring someone from understanding the basic physics concepts to the fundamentals of real scientific n-body simulations.
Both simulation parameters and initial conditions can be interactively changed and explored in the environment. When the integrator is changed (e.g. from euler to leapfrog) the underlying update pattern is visualized in a timeline on the bottom of the screen and in steps on the right side of the screen. The substeps of an integrator (e.g. force interaction, position update) can be selected and visualized in the 3d environment.
A user can program their own integrator, which is fully integrated with all aspects of the environment. During development we experimented with developing an own programming language so integration on the AS could be done, but in the end it was decided to use Python for ease of use. The exact update pattern is analysed and correctly visualized completely on the fly.
A fully custom visualisation and simulation framework was developed. A flexible simulation backend that uses python to talk to various real scientific n-body codes allows simulations to be remotely triggered. Initial conditions and simulation parameters are send over a websocket connection and new integrators can be easily integrated (including custom integrators that are programmed on the fly). The frontend has a fully custom 3d engine deeply integrated with various visualisations of other parts of the simulation, such as the energy conservation, error and stepwise update pattern of a particular integrator.