reward_sample.py

from time import time


class Reward:
    def __init__(self, verbose=False):
        self.previous_steps = None
        self.initial_time = None
        self.verbose = verbose

    @staticmethod
    def get_vector_length(v):
        return (v[0] ** 2 + v[1] ** 2) ** 0.5

    @staticmethod
    def vector(a, b):
        return b[0] - a[0], b[1] - a[1]

    @staticmethod
    def get_time(params):
        # remember: this will not return time before
        # the first step has completed so the total
        # time will be slightly lower
        return params.get('timestamp', None) or time()

    def reward_function(self, params):
        if self.previous_steps is None \
                or self.previous_steps > params['steps']:
            # new lap!
            self.initial_time = self.get_time(params)
        else:
            # we're continuing
            pass

        steering_factor = 1.0

        if abs(params['steering_angle']) > 14:
            steering_factor = 0.7

        reward = float(steering_factor)

        self.previous_steps = params['steps']

        if self.verbose:
            print(params)

        return reward


reward_object = Reward()


def reward_function(params):
    return reward_object.reward_function(params)