Decision-tree based navigation library
For now, this is meant to help compare a decision-tree navigation algorithm with a simple lawnmower-style policy for a robotic kayak.
Maps can come from either hardware tests or from a our simulator.
If you have a Matlab file with data from the boat with Latitude and Longitude
fields, generate the map using tools/data2map.m. Invocation in Matlab is as
follows:
data2map(<PATH TO DATA FILE>, 'data/out.map')
If you wish to generate a simulated map, fiddle with the settings in
createmap.m or mass-produce some maps with multicreatemap.m. I'll probably
combine these later.
Once you have a map, run treemower as follows to generate a path:
./treemower <CSV costmap> <rh | lm> <budget> <lookahead> <random choice fraction> <uncertainty threshold> <output filename>
For example, to use the receding horizon path planner on data/lake.map with
3-step lookahead and 5000 units (explained below) of travel budget to reduce
uncertainty to 20%, your command might look something like this:
./treemower data/lake.map rh 5000 3 1.0 0.2 output/lake.path
1 unit of the travel budget corresponds to 1 tick or cell in the map that is
used by the path planner. The real-world value of this unit depends on what was
specified as the UNIT_LENGTH in data2map.m.
For there to be a meaningful comparison between the hardware map and the
generated paths, the value of SAMPLE_RADIUS in data2map.m and
include/config.hpp must be the same.