drl-supply-chain-env-step.py

class SupplyChainEnvironment(object):
    ...
    def step(self, state, action):
        demands = np.fromfunction(lambda j: self.demand(j+1, self.t), (self.warehouse_num,))
    
        # Calculating the reward (profit)
        total_revenue = self.unit_price * np.sum(demands)
        total_production_cost = self.unit_cost * action.production_level
        total_storage_cost = np.dot( self.storage_costs, 
                np.maximum(state.stock_levels(), np.zeros(self.warehouse_num + 1)) )
        total_penalty_cost = - self.penalty_unit_cost * ( 
                               np.sum( np.minimum(state.warehouse_stock,
                                                  np.zeros(self.warehouse_num)) )  +  
                               min(state.factory_stock, 0) )
        total_transportation_cost = np.dot( self.transporation_costs,               
                                            action.shippings_to_warehouses )
        reward = total_revenue - total_production_cost - total_storage_cost 
                               - total_penalty_cost - total_transportation_cost

        # Calculating the next state
        next_state = State(self.warehouse_num, self.T, self.t)
        next_state.factory_stock = min(state.factory_stock + action.production_level - 
                                       np.sum(action.shippings_to_warehouses), 
                                       self.storage_capacities[0]) 
        for w in range(self.warehouse_num):
            next_state.warehouse_stock[w] = min(state.warehouse_stock[w] + 
                                                action.shippings_to_warehouses[w] - 
                                                demands[w], self.storage_capacities[w+1])    
        next_state.demand_history = list(self.demand_history)

        self.t += 1
        self.demand_history.append(demands)

        return next_state, reward, self.t == self.T - 1