ISAM2Params

This header configures iSAM2’s nonlinear update strategy. ISAM2Params holds global incremental settings and one of three wrapped inner-step configurations: Gauss–Newton, Dogleg, or Dogleg line search.

Authors: Frank Dellaert, et al. (see THANKS for the full author list)

See LICENSE for the license information

try:
    import google.colab
    %pip install --quiet gtsam-develop
except ImportError:
    pass

import gtsam
import numpy as np
from gtsam.symbol_shorthand import L, V, X

`ISAM2GaussNewtonParams`¶

Gauss–Newton uses wildfire thresholding to skip small back-substitution updates. Lower thresholds do more work but approximate the full update more closely.

gauss_newton = gtsam.ISAM2GaussNewtonParams(0.001)
gauss_newton.setWildfireThreshold(5e-4)
print("wildfire threshold:", gauss_newton.getWildfireThreshold())

`ISAM2DoglegParams`¶

Dogleg adds a trust-region radius and an adaptation mode. setInitialDelta(), setAdaptationMode(), and setVerbose() tune that local nonlinear step.

dogleg = gtsam.ISAM2DoglegParams()
dogleg.setInitialDelta(1.0)
dogleg.setAdaptationMode("SEARCH_EACH_ITERATION")
dogleg.setWildfireThreshold(1e-4)
print("initial delta:", dogleg.getInitialDelta())

`ISAM2DoglegLineSearchParams`¶

The line-search variant bounds the trust-region radius and controls step size and sufficient decrease. It is useful when the standard adaptation rule is too coarse.

line_search = gtsam.ISAM2DoglegLineSearchParams()
line_search.setMinDelta(1e-4)
line_search.setMaxDelta(10.0)
line_search.setStepSize(0.5)
line_search.setSufficientDecreaseCoeff(1e-4)
print("delta interval:", line_search.getMinDelta(), line_search.getMaxDelta())

`ISAM2Params`¶

The main object selects one optimization configuration and controls relinearization, factorization, caching, detailed results, and factor-slot reuse. Relinearization threshold and skip are the most important accuracy/performance controls.

params = gtsam.ISAM2Params()
params.setOptimizationParams(gauss_newton)
params.setRelinearizeThreshold(0.05)
params.relinearizeSkip = 1
params.enableDetailedResults = True
params.evaluateNonlinearError = True
params.setFactorization("CHOLESKY")
print("factorization:", params.getFactorization())

Source¶

ISAM2Params.h

AI assistance caveat¶

AI was used to help draft this documentation, and inaccuracies could be present.

ISAM2GaussNewtonParams¶

ISAM2DoglegParams¶

ISAM2DoglegLineSearchParams¶