def step()

in roboschool/gym_forward_walker.py [0:0]

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

    def step(self, a):
        if not self.scene.multiplayer:  # if multiplayer, action first applied to all robots, then global step() called, then step() for all robots with the same actions
            self.apply_action(a)
            self.scene.global_step()

        state = self.calc_state()  # also calculates self.joints_at_limit

        alive = float(self.alive_bonus(state[0]+self.initial_z, self.body_rpy[1]))   # state[0] is body height above ground, body_rpy[1] is pitch
        done = alive < 0
        if not np.isfinite(state).all():
            print("~INF~", state)
            done = True

        potential_old = self.potential
        self.potential = self.calc_potential()
        progress = float(self.potential - potential_old)

        feet_collision_cost = 0.0
        for i,f in enumerate(self.feet):
            contact_names = set(x.name for x in f.contact_list())
            #print("CONTACT OF '%s' WITH %s" % (f.name, ",".join(contact_names)) )
            self.feet_contact[i] = 1.0 if (self.foot_ground_object_names & contact_names) else 0.0
            if contact_names - self.foot_ground_object_names:
                feet_collision_cost += self.foot_collision_cost

        electricity_cost  = self.electricity_cost  * float(np.abs(a*self.joint_speeds).mean())  # let's assume we have DC motor with controller, and reverse current braking
        electricity_cost += self.stall_torque_cost * float(np.square(a).mean())

        joints_at_limit_cost = float(self.joints_at_limit_cost * self.joints_at_limit)

        self.rewards = [
            alive,
            progress,
            electricity_cost,
            joints_at_limit_cost,
            feet_collision_cost
            ]

        self.frame  += 1
        if (done and not self.done) or self.frame==self.spec.max_episode_steps:
            self.episode_over(self.frame)
        self.done   += done   # 2 == 1+True
        self.reward += sum(self.rewards)
        self.HUD(state, a, done)
        return state, sum(self.rewards), bool(done), {}