TODO.txt

-------------
TO-DO

Validate inputs/outputs against matlab individually:
- costs (also look for ways to make this faster)
- compute_derivatives
- forward_pass
- backward_pass

- step 2 in iLQR_core
	- check for termination due to small gradient
		- the way they do it in matlab looks complicated - is there an easier way?

- integrate cost for change in control input

- write final control sequence and states to text file
- test whole thing and see if output looks reasonable - plot in matlab

-------------
FASTER
1. calculate_derivatives : this is taking most of the time
	- store auxiliary variables
	- change incrementing method?
	- take analytic derivatives from DDP-generator

2. backward_pass

3. forward_pass
	- vectorize for parallel line search?

- see links in notes
-----------
Later on:
- fix all the messy stuff
	- extract methods into smaller methods, especially in fw, bw, boxQP

- make it more readable
	- see google cpp style guide
	- refactor to completely separate iLQR class from Model class
		- make sure everything in iLQR is dimension-agnostic
		- there are currently some hacks
	- make it reusable
		- make skeleton for new Model class, for reference

- add documentation
- helper functions are all inline right now to prevent duplicate symbol error - move these to cpp file