def _step()

in gym/gym/envs/mujoco/ant_movement.py [0:0]

• 13 lines of code
• 11 McCabe index (conditional complexity)

    def _step(self, a):
        # print(self.data.qpos.shape)
        xposbefore = self.data.qpos[0,0] if (self.realgoal[0] == 0 or self.realgoal[0] == 1) else self.data.qpos[1,0]
        yposbefore = self.data.qpos[1,0] if (self.realgoal[0] == 0 or self.realgoal[0] == 1) else self.data.qpos[0,0]

        self.do_simulation(a, self.frame_skip)

        xposafter = self.data.qpos[0,0] if (self.realgoal[0] == 0 or self.realgoal[0] == 1) else self.data.qpos[1,0]
        yposafter = self.data.qpos[1,0] if (self.realgoal[0] == 0 or self.realgoal[0] == 1) else self.data.qpos[0,0]

        forward_reward = (xposafter - xposbefore)/self.dt
        # if self.realgoal[0] == 1 or self.realgoal[0] == 3:
            # forward_reward = forward_reward * -1
        side_reward = np.abs(yposafter) * 0.5
        ctrl_cost = .1 * np.square(a).sum()
        reward = forward_reward - ctrl_cost - side_reward
        done = False
        ob = self._get_obs()
        return ob, reward, done, dict(forward_reward=forward_reward, ctrl_cost=ctrl_cost, side_reward=side_reward)