This package and its programs are essentially now an abandoned prototype. We are extracting any goodness and ways of doing things into seperate packages it still remains our favoured demo and data visualisation app for the time being but is not now actively developed while we concentrate on a more modular system.
A package and of library and programs to chart and plot wind, tide, marks, tracks, events and other navigational stuff.
I hijacked this visualisation project to support our BlueBox GPS tracking and accelerometer logger device. We use this to visualise a track filtered by event acceleration and plot static features (marks). This was motivated by my reasearch on collision and grounding event detection for small craft.
We are currently migrating from flat file data files to using the SQLite database. We have experimented with SpatiaLite GIS module but the jury is still out on the utility vs. maintainability aspects (see below).
We have done some experiments using netCDF files which might be interesting way of standardising on layers. We'd really like to support grib as these are lighter and have some compression advantages if we have to get these at sea over a sat phone link etc.
We rely heavily on the geodetics package for everthing geodetic and we are working on completing universal transverse mercator zone defintions for worldwide mapping which we hope to contribute back to that great project.
BlueBox was updated with new accelerometer chip to provide time, GPS position and 9-axis (accelerometer, magnetic & gyro) we are not concerned with altitude in marine envionment but we could substitute depth if we can integrate that data from the navionics. The current push is to visualise the acceleration and orientation of the vessel from this data stream.
The added complexity of dependencies introduced by SpatiaLite module loading and accompanying custom sqlite build vs. the convenience/performance of spatial reasoning and search in the database may not really work for us. The idea was to transform our position data (GPS and charted objects) via. UTM zone transformation to cartesian grid at load time does not apply for realtime GPS tracking. The actual drawing (Gloss/OpenGL) uses pairs of floats. Given that transforms from typesafe dimensioned data (WGS84 positions) shows up as our bottleneck when profiling plotting we could factor a hybrid and more simple solution using vanilla SQLite and storing the graphic points and computing bounding boxes at the graphical level rather than the geodetic, while transforming the GPS WGS84 positions to UTMGrid points in realtime.
direct-sqlite Making
this work well and reliably across platforms requires some fiddling
as we need to link against an system installed sqlite library and
cautiously load mod_spatialite into SQLite. Track this on
feature/GIS branch (now closed).
Nano (ARM Cortex + Nvidia GPU + GPIO) on board system and integration (NMEA) bus.
The port to aarch64 using ghc 8.8.1
was merged from port/aarch64-minimal branch in
October and we are now actively developing on this platform using
ghc 8.8.1 as our primary compiler and similarly on OS X
BlueBox dataWe are considering a kickstarter style project for this as the second prototpype/seed unit is on the bench as I write and this could be of immediate value to the Yachting community especially coded vessels used in a commercial school or charter context, or as a tracker/performance analysis tool for racers and owners.
Started investigating writing a server (central in BluetoothLE terms) using bindings to bluez on Linux. Not no eay to do this in C or Haskell unless we can use the Core.Bluetooth framework on Mac OS unless libbluez is available.
We may look at Grenade RNNs (LSTM) e.g. can we distinguish types of event using a neural net. Can we teach the system to sail the boat? Win races?
We aim to do all this real time on board as we are yacht racers in our analog lives hence the choice of a Jetson Nano for this platform.