Opossum includes two of the best-performing,single-objective optimization algorithms in Grasshopper: model-based RBFOpt and evolutionary CMA-ES. It also includes the multi-objective RBFMOpt, and the multi-objective MACO (Ant Colony), MOEA/D, NSGA-II and NSPSO (Particle Swarm) algorithms from the Pygmo 2 library.
RBFOpt uses advanced machine learning techniques to find good solutions with a small number of function evaluations, i.e. simulations, while CMA-ES reliably finds near-optimal solutions when many function evaluations are possible. (Various papers with benchmark results are availabe here.) RBFMOpt won the “2-Objective Expensive” track of the international Black Box Optimization Competition 2019, surpassing the winner of the previous two years, the commercial algorithm Artelys Knitro.
Opossum 3 includes Performace Explorer, an interactive and intuitive visual tool for performance-informed design space exploration. Performance Explorer visualizes a single-objective optimization problem's fitness landscape with real-time performance feedback and interacts with Grasshopper to allow users to explore the corresponding designs. Papers on Performance Explorer are availabe here.
Opossum's GUI is similar to Galapagos. Opossum has a results table, which makes it easy to revisit all optimization results by double-clicking entries in the table.
Opossum requires a free license key, please email support to get one.
Opossum is currently developed at ICD University Stuttgart.
Opossum 3.0.1 2022-August-10
- Includes Performance Explorer
Opossum 2.2.5 2022-August-05
- Fixed a bug where license had to be reentered after Rhino update
- Minor bug fixes
Opossum 2.2.4 2021-April-07
- Opossum does not expire the whole canvas anymore
- "Tuned" default parameters for RBFMOpt
Opossum 2.2.3 2020-December-09
- Fixed a bug with serialization in Grasshopper
Opossum 2.2.2 2020-December-01
- Results table is more responsive
- Results table is sortable
- Fixed a bug where clicking the results table did not reset to the correct solution
Opossum 2.2.0 2020-October-15
- Faster interaction with Python back-end
- Added Paretorank and improved Hypervolume calculation for MOO results
- Added tooltips for Expert settings
- Added "augmented Tchebycheff sum", "epsilon", "Do Init?" options for RBFMOpt
- Close now resets Opossum
- Update to RBFOpt 4.2.1
- Fixed Rhino 7 WIP display issue
- Fixed crash when changing variables
Opossum 2.1.0 2020-May-23
- Added multi-objective algorithms: MACO, MOEA/D and NSPSO.
- Expert settings for all algorithms
- Bug fixes
Opossum 2.0.0 2019-September-09
- Multi-objective optimization with RBFMOpt and NSGA-II.
Opossum 1.7.1 2019-May-27
- Genepools for CMAES starting point.
Opossum 1.7.0 2019-April-28
- Includes CMA-ES algorithm
- Eliminates need for installer (just drop Opossum.gha and the "Solvers" folder into Libraries)
- Updated to RBFOpt 4.1.1: New options, more efficient initialization phase
Opossum 1.5.0 2018-March-16
- Canvas lockscreen
- Automatic "max_evaluation" setting
Opossum 1.4.3 2017-Sept-30
- Ensured Invariant Culture
Opossum 1.4.2 2017-Sep-23
- Bug Fix
Opossum 1.4.1 2017-Sep-17
- Update to RBFOpt 3.0.1: faster performance with improved local search
- Results tab allows examination of all optimization results
Opossum 1.2.3 2017-Jun-06
- Fixed bug caused by very long PATH variable
Opossum 1.2.2 2017-Jun-12
- Fixed license write-permission issue
Opossum 1.2.1 2017-Jun-12
- Updated to RBFOpt 2.1: faster performance
- Included example file
- Small bugfixes
Opossum 1.1.3 2017-Jun-12
- Opossum window always on top.
Opossum 1.1.2 2017-Jun-12
- More compact libraries
- Culture-independent license check
Opossum 1.1.1 2016-Oct-27
- GUI Improvements
- Installer Improvements
- Bug Fixes