| Andrew Cooke | Contents | Latest | RSS | Twitter | Previous | Next

C[omp]ute

Welcome to my blog, which was once a mailing list of the same name and is still generated by mail. Please reply via the "comment" links.

Always interested in offers/projects/new ideas. Eclectic experience in fields like: numerical computing; Python web; Java enterprise; functional languages; GPGPU; SQL databases; etc. Based in Santiago, Chile; telecommute worldwide. CV; email.

Personal Projects

Lepl parser for Python.

Colorless Green.

Photography around Santiago.

SVG experiment.

Professional Portfolio

Calibration of seismometers.

Data access via web services.

Cache rewrite.

Extending OpenSSH.

C-ORM: docs, API.

Last 100 entries

Calling C From Fortran 95; Bjork DJ Set; Z3 Example With Python; Week 1; Useful Guide To Starting With IJulia; UK Election + Media; Review: Reinventing Organizations; Inline Assembly With Julia / LLVM; Against the definition of types; Dumb Crypto Paper; The Search For Quasi-Periodicity...; Is There An Alternative To Processing?; CARDIAC (CARDboard Illustrative Aid to Computation); The Bolivian Case Against Chile At The Hague; Clear, Cogent Economic Arguments For Immigration; A Program To Say If I Am Working; Decent Cards For Ill People; New Photo; Luksic And Barrick Gold; President Bachelet's Speech; Baltimore Primer; libxml2 Parsing Stream; configure.ac Recipe For Library Path; The Davalos Affair For Idiots; Not The Onion: Google Fireside Chat w Kissinger; Bicycle Wheels, Inertia, and Energy; Another Tax Fraud; Google's Borg; A Verion That Redirects To Local HTTP Server; Spanish Accents For Idiots; Aluminium Cans; Advice on Spray Painting; Female View of Online Chat From a Male; UX Reading List; S4 Subgroups - Geometric Interpretation; Fucking Email; The SQM Affair For Idiots; Using Kolmogorov Complexity; Oblique Strategies in bash; Curses Tools; Markov Chain Monte Carlo Without all the Bullshit; Email Para Matias Godoy Mercado; The Penta Affair For Idiots; Example Code To Create numpy Array in C; Good Article on Bias in Graphic Design (NYTimes); Do You Backup github?; Data Mining Books; SimpleDateFormat should be synchronized; British Words; Chinese Govt Intercepts External Web To DDOS github; Numbering Permutations; Teenage Engineering - Low Price Synths; GCHQ Can Do Whatever It Wants; Dublinesque; A Cryptographic SAT Solver; Security Challenges; Word Lists for Crosswords; 3D Printing and Speaker Design; Searchable Snowden Archive; XCode Backdoored; Derived Apps Have Malware (CIA); Rowhammer - Hacking Software Via Hardware (DRAM) Bugs; Immutable SQL Database (Kinda); Tor GPS Tracker; That PyCon Dongle Mess...; ASCII Fluid Dynamics; Brandalism; Table of Shifter, Cassette and Derailleur Compatability; Lenovo Demonstrates How Bad HTTPS Is; Telegraph Owned by HSBC; Smaptop - Sunrise (Music); Equation Group (NSA); UK Torture in NI; And - A Natural Extension To Regexps; This Is The Future Of Religion; The Shazam (Music Matching) Algorithm; Tributes To Lesbian Community From AIDS Survivors; Nice Rust Summary; List of Good Fiction Books; Constructing JSON From Postgres (Part 2); Constructing JSON From Postgres (Part 1); Postgres in Docker; Why Poor Places Are More Diverse; Smart Writing on Graceland; Satire in France; Free Speech in France; MTB Cornering - Where Should We Point Our Thrusters?; Secure Secure Shell; Java Generics over Primitives; 2014 (Charlie Brooker); How I am 7; Neural Nets Applied to Go; Programming, Business, Social Contracts; Distributed Systems for Fun and Profit; XML and Scheme; Internet Radio Stations (Curated List); Solid Data About Placebos; Half of Americans Think Climate Change Is a Sign of the Apocalypse; Saturday Surf Sessions With Juvenile Delinquents; Ssh, tty, stdout and stderr; Feathers falling in a vacuum; Santiago 30m Bike Route

© 2006-2015 Andrew Cooke (site) / post authors (content).

Efficient Collision Detection with Pessimistic Measures

From: "andrew cooke" <andrew@...>

Date: Sat, 15 Sep 2007 16:16:32 -0400 (CLT)

I'm modelling a set of colliding lines (rods in 2D) and using Napito to
plot their trajectories.

Modelling individual lines is easy; the problem is reliably detecting
collisions.  In particular, I need to find the *first* collision amongst
all lines (since the movements change after a collision I am simply
restarting after that point - a future optimisation might do something
more sophisticated).

So this is some kind of search.  The trouble is that I don't have an
analytic solution for collisions.  Instead I have a fairly efficient but
pessimistic (it may give false positives) method for determining whether
any two lines touch within a given interval and an iterative method for
determining the time of intersection within some interval that is only
reliable in the limit of small intervals.

Finding the first possible pair of lines ("candidates") is easy - just use
the pessimistic method over progressively smaller time intervals.  But the
"limit of small intervals" makes everything uncertain and possibly
expensive - if you narrow down on an interval to search, and then find
that that it was unsuccessful (which is possible, since the initial
process is pessimistic), then what do you do?  It seems that you are
forced into a fairly detailed systematic search over time (and of
necessarily small intervals).

Fortunately, I can implement the candidate test in a way that makes it
progressively less pessimistic as the time interval decreases (it's
asymptotically correct).  So an explicit search is not that expensive. 
What becomes an issue then is the efficient broadening of the search to
include other candidates if the initial "best" fails.

This suggests that the best approach is to use a depth first search in
time (consider the tree that successively divides intervals in half),
discarding lines when they are not part of the overlapping group as you
descend, pruning when there are no candidates, and using the iterative
method as a test only at the very bottom "leaf" intervals.

In retrospect that seems obvious, but it's taken me a heck of a time to
see it clearly.  I keep being tempted to use the iterative method sooner
(when it works it converges rapidly, but when it fails I am left with no
way to fold that knowledge back into the search).

I guess there may be a future optimisation which uses the iterative
approach in some kind of speculative manner.  Perhaps when I am more
confident about its properties (it's just an analytic solution to a linear
approximation).

Andrew

Subtle, but Correct (I Hope)

From: "andrew cooke" <andrew@...>

Date: Sat, 15 Sep 2007 17:27:17 -0400 (CLT)

I forgot to mention an additional concern.  It's not really a concern,
since I think the previous argument is correct, but it was one reason I
took so long to get this clear.

Since the initial restriction is pessimistic you may, for any finite
interval, be "blocked" from finding a correct pair of colliding the lines
by the presence of one or more "confusing" lines, which are incorrectly
included.

This is resolved by searching all time ranges in progressively smaller
intervals and relying on the asymptotically correct behaviour to
eventually weed out the confusion.

What I had been trying to do, without much success, was remove the
confusion by pushing information back "up" the search tree.

Andrew

Comment on this post