Hi，the problem of the reward function. about rl-agents HOT 3 CLOSED

• i notice that a variable which named “SPEED_COUNT” but i don't know the meaning

It is used to discretise the space of velocities between SPEED_MIN and SPEED_MAX, so that discrete actions FASTER and SLOWER can be used.
It is used in the reward function to compute a "velocity ratio" between 0 (min speed) and 1 (max speed) and the corresponding reward.

• how often the high velocity reward is returned when the agent is trained in the environment ？

the _reward() function is called at every step(action) in envs/abstract.py (= every action = every second)

from rl-agents.

* i notice that a variable which named “SPEED_COUNT” but i don't know the meaning

It is used to discretise the space of velocities between SPEED_MIN and SPEED_MAX, so that discrete actions FASTER and SLOWER can be used.
It is used in the reward function to compute a "velocity ratio" between 0 (min speed) and 1 (max speed) and the corresponding reward.

* how often the high velocity reward is returned when the agent is trained in the environment ？

the _reward() function is called at every step(action) in envs/abstract.py (= every action = every second)

Thanks for you help! I wish every success in your future !

from rl-agents.

You too!

from rl-agents.