From: Dan Koditschek [kod@umich.edu] Sent: Thursday, September 11, 2003 10:55 PM To: 'Karen J. Coulter'; Ron Fearing; Alfred Rizzi; Dan Koditschek; Kellar Autumn; Mark Cutkosky; Martin Buehler; Robert Full (Robert Full) Cc: 'Al Rizzi'; 'Uluc Saranli'; 'Jonathan Edward Clark'; 'G. Clark Haynes' Subject: RiSE Working Group 4.1 - Simulation Tool Recommendations Folks, On behalf of the RiSE Simulation Tool working subgroup 4.1, I submit the following report and recommendations for consideration at tomorrow's telecon ------------------------------------------- Group 4.1 Recommendations 9/12/03: RiSE Simulation Tools ------------------------------------------- After reviewing a large number of alternatives, as can be viewed on the OpenDiscussion page http://www-cdr.stanford.edu/twiki/bin/view/Main/OpenDiscussion Group 4.1 has whittled the choice down to ADAMS vs. Arachi's Dynamics Engine (DE)". To make the final choice between these two very different contenders, we observe that the RiSE project requires a simulation tool with the properties (in rough order of priority): A) good underlying mechanics or easy interface thereto B) easily driven by an external controller C) easy interface to a "creature description environment" D) output can be more or less sexily rendered as needed Regarding A) we are not confident that any presently available package contains the intrinsic capabilities required, for example (not an exhaustive list!): partially plastic and/or deformable solids (e.g., fluid filled footpads); various friction models (e.g., customizable support distribution models such as might be necessary to handle a gecko-hair-patch-ground interaction) model; etc. In this regard, although DE does not have nearly the accumulated bells and whistles of ADAMS, it has arguably better (at least more efficient) rigid body collision modeling and may have superior intrinsic contact mechanics. Most importantly, its design as an "engine" gives much greater confidence in the ease of interfacing to our own handbuilt contact models than does ADAMS. Regarding B), our own direct past experience with ADAMS indicates that it is dismally poor in this regard. Again, since DE is an "engine" the expectation is that this will be much more straightforward. Regarding C), ADAMS already has a viable (reasonably parametrized and readily learned) CDL, whereas DE does not and one would have to be handbuilt by project personnel. Both packages seem to be equally reasonable along the last axis, D). Overall, our view is that as a "complete environment" ADAMS is not a system designed for what we want to accomplish. Therefore, we will always run into problems originating from the difference in design philosophy in terms of what it is designed for and what we want to do. Furthermore, we will probably receive negligible support from MSC Software (as members of the group have already experienced) as a small and insignificant customer compared to their large commercial customer base. In contrast, DE, as a hungrier, smaller company seems to be considerable more amenable to providing the requisite flexible support. Our recommendation is to go with DE and identify people to take up the following tasks (and figure out what their timelines/must be relative to the larger Project Plan mpp file): 1) Immediately interact with Arachi and get the details about how to hook up the RiSE runtime development environment (increasingly looking like RHexLib) to DE, then if all remains seemingly good, make the *buy* decision and implement this connection. 2) Following the DE *buy* decision, start work on the design and implementation of a CDL for DE 3) In parallel with 1) & 2) collect a time-prioritized list of contact mechanics simulation capabilities (who will need what kind of models at what point in time). Devise a plan for the PIs inspection that implements the list within DE (again assuming the *buy* decision) in order of priority and then start working on the list. ========================================= Daniel E. Koditschek, Professor AI Lab & Controls Lab, EECS Dept. University of Michigan 170 ATL, 1101 Beal Ave. Ann Arbor, MI 48109-2110 TEL (734) 764 4307; FAX (734) 763-1260 Email: kod@umich.edu; http://ai.eecs.umich.edu/people/kod =========================================