improvements:
- when reseting to a new path, robot does not take example of multiple movements
- robot does not continue to explore after reaching goal during first run.
- update g_values if car encounters a location it has already been, and the g_values are inaccurate -- requires propagation through graph
- if the robot can see (utilizing more of the sensor data) a square it has already been to, we need to propagate the g-values as well.