Djinni: A Templatized C++ Framework with Python Bindings for Heuristic Search

Djinni is a framework for implementing heuristic search algorithms. The core elements are coded in C++ and python bindings are provided to simplify the user interface. The current version of Djinni implements compressed annealing (Ohlmann et al., 2004), a generalization of the well-known simulated annealing algorithm, and includes code used by Ohlmann and Thomas (2007) to solve the traveling salesman problem with time windows (TSPTW). The Djinni framework uses C++ templates to separate code into three parts: a heuristic search algorithm, a problem model, and problem data. Thus, it is straightforward to apply compressed or simulated annealing to problems other than the TSPTW. Furthermore, it is not difficult to apply different heuristic search algorithms to the same problem.

Downloading Djinni

Djinni can be downloaded as a source code tarball. Documentation can be generated by running doxygen. A pre-built binary is also available for Debian.

Authors and Contact Info

Djinni's primary author is Robert Hansen, assisted in initial stages by Jeff Ohlmann, Barry Thomas, and Tristan Thiede. Justin Goodson assisted with some fine tuning and is the contact for Djinni.


Ohlmann, J., J. Bean, and S. Henderson (2004). Convergence in probability of compressed annealing. Mathematics of Operations Research, vol 29, 837-860.

Ohlmann, J. and B. Thomas (2007). A compressed annealing approach to the traveling salesman problem with time windows. INFORMS Journal on Computing, vol 19, 80-90.

Last modified 6 years ago Last modified on May 16, 2011 4:10:25 PM