An implementation of the Simulated Annealing stochastic nonlinear optimization method. More...

#include <frc/optimization/SimulatedAnnealing.h>

Public Member Functions
constexpr	SimulatedAnnealing (double initialTemperature, std::function< State(const State &)> neighbor, std::function< double(const State &)> cost)
	Constructor for Simulated Annealing that can be used for the same functions but with different initial states. More...

State	Solve (const State &initialGuess, int iterations)
	Runs the Simulated Annealing algorithm. More...

Detailed Description

template<typename State>
class frc::SimulatedAnnealing< State >

An implementation of the Simulated Annealing stochastic nonlinear optimization method.

Solving optimization problems involves tweaking decision variables to try to minimize some cost function. Simulated annealing is good for solving optimization problems with many local minima and a very large search space (it’s a heuristic solver rather than an exact solver like, say, SQP or interior-point method). Simulated annealing is a popular choice for solving the traveling salesman problem (see TravelingSalesman).

See also: https://en.wikipedia.org/wiki/Simulated_annealing

Template Parameters

State The type of the state to optimize.

Constructor & Destructor Documentation

◆ SimulatedAnnealing()

template<typename State >

constexpr frc::SimulatedAnnealing< State >::SimulatedAnnealing	(	double	initialTemperature,
		std::function< State(const State &)>	neighbor,
		std::function< double(const State &)>	cost
	)

inlineconstexpr

Constructor for Simulated Annealing that can be used for the same functions but with different initial states.

Parameters

initialTemperature	The initial temperature. Higher temperatures make it more likely a worse state will be accepted during iteration, helping to avoid local minima. The temperature is decreased over time.
neighbor	Function that generates a random neighbor of the current state.
cost	Function that returns the scalar cost of a state.

Member Function Documentation

◆ Solve()

template<typename State >

State frc::SimulatedAnnealing< State >::Solve	(	const State &	initialGuess,
		int	iterations
	)

inline

Runs the Simulated Annealing algorithm.