Review of: VISION AND TOUCH GUIDED GRASPING
Organization: MIT
Principal Investigators: Salisbury and Slotine

Reviewed by: Fabio Cozman, fgcozman@cs, x2007

Objectives
----------
Development of robotic systems capable of grasping objects using visual and
	touch information:
	a) Visual techniques for localizing target objects.
	b) Grasp planning techniques.
	c) Grasp quality assessment, through touch and vision.
	d) Adapting to dynamic variation in grasp (friction, mass, energy
		variations).

Justification
-------------
Straight remote control of manipulators is impractical in planetary operations
	(time delay and bandwidth). Local ability is necessary to insure
	safety and remove excessive load from Earth operators.

Approach
--------
Equipment: WAM arm/Talon End-Effector + FastEyes stereo system.
	Both the arm and the stereo system are available at the MIT AI lab.
The FastEyes stereo system will be used to construct 3D maps of terrain.
Other sources of 3D data will be used if available (and if budget permits).
The 3D data will be analyzed to detect graspable objects. An analysis of 
	possible grasps will be made. Grasps will be monitored by the 
	vision system.
Grasp must be controlled properly in order to compensate for unknown mass
	properties of objects.

Deliverables
------------
They imply that they will test a system which can acquire objects and grasp
	them robustly, and they will test this system in simulated planetary
	surface environment.
They promise to transfer results (algorithms, techniques and evaluations) to
	JPL Remote Scientist program.
They mention that they may have to build a visual interface to the system,
	which would them be relevant to the JPL UOI and ARC VEVI programs.

Logistics Summary
-----------------
a) Number of years: 3
b) Money total: 759K
c) Money per year: FY96-220K, FY97-264K, FY98-275K
d) Commercial partners & money they put in: none
e) Participants (organizations involved in research/ development): MIT only
f) Identified customers: The JPL Remote Scientist and Rover projects, NASA/JSC
g) This project continues work currently on progress (during FY-94, FY-95);
	they seem to have developed coherently infra-structure so far.

Your Opinion
------------
The title reflects very well the central objective: they are concentrated
	on sensing abilities for grasping.
A weak point is that, even though they mention "touch", they are not
	explaining what they will do with it. Is it ready? Or totally unknown?
The justification of the project is coherent with planetary exploration, and
	the approach taken outlines a sensible, incremental path of research.
On the TRIWG group, there seems not to be competitors to the manipulator
	centered approach this project is interested in. There is parallel
	work in the Robotics community as a whole, for example Mat Mason at
	CMU is looking at vision for manipulators.
The proposal does not communicate very well what they intend to transfer or
	produce to NASA. They intend mostly to study a problem (grasping with
	sensing) and see what they can do; there is no specific objective
	in sight.