7b2ff2d8345738ca4a9857863f375049a335ee02,cistar-dev/cistar/core/base_env.py,SumoEnvironment,setup_initial_state,#SumoEnvironment#,205

Before Change

        self.sorted_ids = self.sort_by_position()

        // collect headway, leader id, and follower id data
        vehicles = self.get_headway_dict()

        for veh_id in self.ids:
            self.vehicles[veh_id]["headway"] = vehicles[veh_id]["headway"]
            self.vehicles[veh_id]["leader"] = vehicles[veh_id]["leader"]
            self.vehicles[veh_id]["follower"] = vehicles[veh_id]["follower"]

        // dictionary of initial observations used while resetting vehicles after each rollout
        self.initial_observations = deepcopy(dict(self.vehicles))

After Change

        // so density in vehicles/km would be 1000 * self.density
        self.density = self.scenario.num_vehicles / self.scenario.net_params["length"]

    def setup_initial_state(self):
        
        Store initial state so that simulation can be reset at the end.
        TODO: Make traci calls as bulk as possible
        Initial state is a dictionary: key = vehicle IDs, value = state describing car
        
        // collect ids and prepare id and vehicle lists
        self.ids = self.traci_connection.vehicle.getIDList()
        self.controlled_ids.clear()
        self.sumo_ids.clear()
        self.rl_ids.clear()
        self.vehicles.clear()

        // create the list of colors used to different between different types of
        // vehicles visually on sumo"s gui
        self.colors = {}
        key_index = 1
        color_choice = np.random.choice(len(COLORS))
        for key in self.scenario.type_params.keys():
            self.colors[key] = COLORS[(color_choice + key_index) % len(COLORS)]
            key_index += 1

        for veh_id in self.ids:
            // import initial state from traci and place in vehicle dict
            vehicle = dict()
            vehicle["id"] = veh_id
            veh_type = self.traci_connection.vehicle.getTypeID(veh_id)
            vehicle["type"] = veh_type
            self.traci_connection.vehicle.setColor(veh_id, self.colors[veh_type])
            vehicle["edge"] = self.traci_connection.vehicle.getRoadID(veh_id)
            vehicle["position"] = self.traci_connection.vehicle.getLanePosition(veh_id)
            vehicle["lane"] = self.traci_connection.vehicle.getLaneIndex(veh_id)
            vehicle["speed"] = self.traci_connection.vehicle.getSpeed(veh_id)
            vehicle["length"] = self.traci_connection.vehicle.getLength(veh_id)
            vehicle["max_speed"] = self.traci_connection.vehicle.getMaxSpeed(veh_id)

            // specify acceleration controller
            controller_params = self.scenario.type_params[veh_type][1]
            vehicle["controller"] = controller_params[0](veh_id=veh_id, **controller_params[1])

            if controller_params[0] == SumoController:
                self.sumo_ids.append(veh_id)
            elif controller_params[0] == RLController:
                self.rl_ids.append(veh_id)
            else:
                self.controlled_ids.append(veh_id)

            // specify lane-changing controller
            lane_changer_params = self.scenario.type_params[veh_type][2]
            if lane_changer_params is not None:
                vehicle["lane_changer"] = lane_changer_params[0](veh_id=veh_id, **lane_changer_params[1])
            else:
                vehicle["lane_changer"] = None

            self.vehicles[veh_id] = vehicle
            self.vehicles[veh_id]["absolute_position"] = self.get_x_by_id(veh_id)
            // the time step of the last lane change is always present in the environment,
            // but only used by sub-classes that apply lane changing
            self.vehicles[veh_id]["last_lc"] = -1 * self.lane_change_duration

            // set speed mode
            self.set_speed_mode(veh_id)

            // set lane change mode
            self.set_lane_change_mode(veh_id)

            // Saving initial state
            // route_id = self.traci_connection.vehicle.getRouteID(veh_id)
            route_id = "route" + vehicle["edge"]
            pos = self.traci_connection.vehicle.getPosition(veh_id)

            self.initial_state[veh_id] = (vehicle["type"], route_id, vehicle["lane"],
                                          vehicle["position"], vehicle["speed"], pos)

        // collect list of sorted vehicle ids
        self.sorted_ids = self.sort_by_position()

        // collect headway, leader id, and follower id data
        // vehicles = self.get_headway_dict()

        for veh_id in self.ids:
            // self.vehicles[veh_id]["headway"] = vehicles[veh_id]["headway"]
            // self.vehicles[veh_id]["leader"] = vehicles[veh_id]["leader"]
            // self.vehicles[veh_id]["follower"] = vehicles[veh_id]["follower"]
            
            headway = self.traci_connection.vehicle.getLeader(veh_id, 200)
            if headway is None:
                self.vehicles[veh_id]["leader"] = ""
                self.vehicles[veh_id]["follower"] = ""
                self.vehicles[veh_id]["headway"] = self.scenario.length - self.vehicles[veh_id]["length"]
            else:
                self.vehicles[veh_id]["headway"] = headway[1]
                self.vehicles[veh_id]["leader"] = headway[0]
                self.vehicles[headway[0]]["follower"] = veh_id

        // dictionary of initial observations used while resetting vehicles after each rollout
        self.initial_observations = deepcopy(dict(self.vehicles))

        // contains the last lc before the current step
        self.prev_last_lc = dict()

Instances