Behavior modulations

Behavior modulation calls virtual method pre (C++, Python) before and virtual method post (C++, Python) after a behavior is evaluated.

Specializing a new behavior modulation is a simpler alternative to specialize a new behavior, when the specialization should not change the actual methods called to compute a behavior but modulates the parameters (or even state) based on the state and/or the computed command.

Like for behavior, the command produced by post should be kinematically feasible, but it is not strictly required. You should also take into account that modulation are performed as a stack: when behavior b is associated to a sequence of modulations [m_1, m_2, ..., m_n], b.compute_cmd will call them like

m_1(b, time_step)
...
m_n(b, time_step)
cmd <- b.compute_cmd_internal(...)
cmd <- m_n(n, time_step, cmd)
...
cmd <- m_1(n, time_step, cmd)

Note

Although not enforced, each other, modulations should reset in post any change they have performed on the behavior in pre.

Virtual methods

C++ method	Python method	override
pre	pre	can
post	post	can

Class skelethon

C++Python

#include "navground/core/behavior_modulation.h"

namespace core = navground::core;

struct MyBehaviorModulation : public core::BehaviorModulation {
  // CAN override
  // modulates the behavior by setting parameters while caching their original value
  // void pre(core::Behavior &behavior, ng_float_t time_step) override;
 
  // CAN override
  // modulates the cmd and resets parameters to their original values
  // core::Twist2 post(core::Behavior &behavior, ng_float_t time_step, const core::Twist2 & cmd) override;
};