Ceres solver states everywhere that it can
[...] solve robustified bounds constrained non-linear least squares problems
and that it supports upper and lower bounds constraints on the parameter blocks (for example in http://ceres-solver.org/modeling_faqs.html it states Ceres Solver only supports upper and lower bounds constraints on the parameter blocks
), but somehow I can't find anywhere in the documentation how I can set these upper and lower bounds.
So, how do I set upper and lower bounds for parameter blocks in ceres solver?
Specifically, how do I do that in an AutoDiffCostFunction
? If I use if
statements to return a very big residual out of bounds, then that function isn't differentiable.
For example, this is the ceres Hello World:
struct CostFunctor {
template <typename T>
bool operator()(const T* const x, T* residual) const {
residual[0] = 10.0 - x[0];
return true;
}
};
int main(int argc, char** argv) {
google::InitGoogleLogging(argv[0]);
// The variable to solve for with its initial value.
double initial_x = 5.0;
double x = initial_x;
// Build the problem.
Problem problem;
// Set up the only cost function (also known as residual). This uses
// auto-differentiation to obtain the derivative (jacobian).
CostFunction* cost_function =
new AutoDiffCostFunction<CostFunctor, 1, 1>(new CostFunctor);
problem.AddResidualBlock(cost_function, nullptr, &x);
// Run the solver!
Solver::Options options;
options.linear_solver_type = ceres::DENSE_QR;
options.minimizer_progress_to_stdout = true;
Solver::Summary summary;
Solve(options, &problem, &summary);
std::cout << summary.BriefReport() << "\n";
std::cout << "x : " << initial_x
<< " -> " << x << "\n";
return 0;
}
how would I impose in this example a lower bound of 2.0 and an upper bound of 20.0 to the parameter x?
You can use methods setParameterLowerBound
and setParameterUpperBound
as defined here:
http://ceres-solver.org/nnls_modeling.html?highlight=setparameterlowerbound#_CPPv4N5ceres7Problem22SetParameterLowerBoundEPdid
In your case, I guess something like this:
problem.SetParameterLowerBound(&x, 0, lower_bound);
problem.SetParameterUpperBound(&x, 0, upper_bound);