def step()

in gym_pusht/envs/pusht.py [0:0]

• 19 lines of code
• 2 McCabe index (conditional complexity)

    def step(self, action):
        self.n_contact_points = 0
        n_steps = int(1 / (self.dt * self.control_hz))
        self._last_action = action
        for _ in range(n_steps):
            # Step PD control
            # self.agent.velocity = self.k_p * (act - self.agent.position)    # P control works too.
            acceleration = self.k_p * (action - self.agent.position) + self.k_v * (
                Vec2d(0, 0) - self.agent.velocity
            )
            self.agent.velocity += acceleration * self.dt

            # Step physics
            self.space.step(self.dt)

        # Compute reward
        coverage = self._get_coverage()
        reward = np.clip(coverage / self.success_threshold, 0.0, 1.0)
        terminated = is_success = coverage > self.success_threshold

        observation = self.get_obs()
        info = self._get_info()
        info["is_success"] = is_success
        info["coverage"] = coverage

        truncated = False
        return observation, reward, terminated, truncated, info