def after_step()

in gym-compete/gym_compete/new_envs/agents/ant.py [0:0]

• 22 lines of code
• 3 McCabe index (conditional complexity)

    def after_step(self, action):
        xposafter = self.get_body_com("torso")[0]
        forward_reward = (xposafter - self._xposbefore) / self.env.dt
        if self.move_left:
            forward_reward *= -1
        ctrl_cost = .5 * np.square(action).sum()
        cfrc_ext = self.get_cfrc_ext()
        contact_cost = 0.5 * 1e-3 * np.sum(
            np.square(np.clip(cfrc_ext, -1, 1))
        )
        qpos = self.get_qpos()
        agent_standing = qpos[2] >= 0.28
        survive = 1.0 if agent_standing else -1.
        reward = forward_reward - ctrl_cost - contact_cost + survive

        reward_info = dict()
        reward_info['reward_forward'] = forward_reward
        reward_info['reward_ctrl'] = ctrl_cost
        reward_info['reward_contact'] = contact_cost
        reward_info['reward_survive'] = survive
        reward_info['reward_move'] = reward

        done = not agent_standing

        return reward, done, reward_info