notes from meeting

questions:
	- could machine learning (eg Ng helicopter maneuver stuff) be applied to problem of robustly combining open-loop + feedback to landing maneuvers for perching?
		- note: MIT has used macine learning for indoor
		- could ML strategy or results be implemented ONBOARD (vs. Ng, MIT)?
	- what is the best way to incorporate sensor information into a predominantly open-loop landing strategy?
	- how much sensor noise & uncertainty can we tolerate
		- how to do necessary signal processing in 50g package
		- possibility to adopt v. robust (but low computation) sensor processing strategies(neuro-inspired & preflexes) used by insects
	- are there simple modifications to plane and/or propellor that will immprove robustness
	- f landing without significantly reducing flight performance (robin vs. swallow)
	- How to operate while in wall contact?  Constrained sliding, motion planning w/ vectored thrust
	- How to tune semi-active suspenstion in this nonlinear domain 
		- different from usual freq. domain analysis of grand vehicle suspension (Dynamically change properties, eg brake)
		- EPAMs
	- strategies for recovery & regripping if lose grip due to large gusts(also to detect when losing grip)

- using coils as actuators
- investigate v. light actuators, eg coil, EPAMs, 
- pulse nitinol to grip/ungrip
- electroadhestion possibilities

what research tasks need to be added to address these questions properly?
alan: thus far to systems eg in air, not air/ground transition w/ large change in dynamics & eq of motion (2 different regimes)

-landing strategies
	- parachutes, squirrels, birds, landing gear
	- changing dynamics right at the end
	- damping for energy absorbtion (high damper landing gear)
	- land while in contact, eg add two dampers that create a rotation and thus a predictable motion
		-enable more open loop control

- multiple propellors allow more control with no windspeed