Calling the superclass constructor 'rl.env.MA​TLABEnviro​nment' after an object use or after a return statement is not supported.

We aim to maximize the sum rate through the joint optimization of power allocation, UAV mobilty, and SIC decoding order for a UAV-NOMA system using reinforcement learning DDPG optimization. The following errors arise in the learning environment, please can anyone rectify and stabilize the learning environment: 'Calling the superclass constructor 'rl.env.MATLABEnvironment' after an object use or after a return statement is not supported'. I attached the learning environment script: classdef NOMAEnv < rl.env.MATLABEnvironment

properties

% System parameters

numSubchannels = 6;

P_max = 20; % Maximum power (watts)

N_max = 5; % Number of mobile NOMA users

timeSteps = 10; % Number of time steps

uavSpeed = 5; % UAV speed (m/s)

flightHeight = 100; % UAV flight height (meters)

radius = 100; % Area radius (meters)

% Additional properties

userPositions;

currentStep;

T = 100; % Total duration

delta_t; % Duration of each time slot

% Optimal solutions (for comparison)

optimalPowerAllocation;

optimalTrajectory;

optimalDecodingOrder;

optimalSumRate;

end

properties(Access = protected)

State;

IsDone = false;

end

methods

function this = NOMAEnv()

% Load optimal solutions

load('Training_data.mat', 'optimalPowerAllocation', 'optimalTrajectory', 'optimalDecodingOrder', 'optimalSumRate');

this.optimalPowerAllocation = optimalPowerAllocation;

this.optimalTrajectory = optimalTrajectory;

this.optimalDecodingOrder = optimalDecodingOrder;

this.optimalSumRate = optimalSumRate;

this.delta_t = this.T / this.timeSteps;

% Initialize Observation and Action info

ObservationInfo = rlNumericSpec([3 * this.N_max + 3, 1]);

ObservationInfo.Name = 'User distances, angles, current UAV position, and time step';

ObservationInfo.LowerLimit = [-inf * ones(2 * this.N_max, 1); zeros(this.N_max, 1); -this.radius * ones(2, 1); 0; 0];

ObservationInfo.UpperLimit = [inf * ones(2 * this.N_max, 1); ones(this.N_max, 1); this.radius * ones(2, 1); this.flightHeight; this.timeSteps];

ActionInfo = rlNumericSpec([this.N_max + 2, 1], 'LowerLimit', -1, 'UpperLimit', 1);

ActionInfo.Name = 'Actions for power allocation, UAV movement, SIC order';

% Call superclass constructor

[e-mail](ObservationInfo, ActionInfo);

% Initialize environment

this.reset();

end

function [Observation, Reward, IsDone, LoggedSignals] = step(this, Action)

LoggedSignals = [];

% Update power allocation, UAV position, and decoding order based on the action

[powerAllocation, uavPosition, decodingOrder] = this.interpretAction(Action);

% Update environment state

this.updateState(uavPosition, decodingOrder);

% Calculate reward (sum rate)

Reward = this.calculateSumRate(powerAllocation, uavPosition, decodingOrder);

% Check if episode is done

this.currentStep = this.currentStep + 1;

if this.currentStep >= this.timeSteps

this.IsDone = true;

end

IsDone = this.IsDone;

Observation = this.State;

end

function InitialObservation = reset(this)

% Initialize user positions

this.userPositions = rand(this.N_max, 2) * 2 * this.radius - this.radius;

% Initialize UAV position

uavInitialPosition = [0, 0, this.flightHeight];

% Initialize decoding order

initialDecodingOrder = randperm(this.N_max);

% Reset current step

this.currentStep = 0;

% Update state

this.updateState(uavInitialPosition, initialDecodingOrder);

InitialObservation = this.State;

this.IsDone = false;

end

end

methods (Access = protected)

function updateState(this, uavPosition, decodingOrder)

% Calculate distances and angles to users

relativePositions = this.userPositions - uavPosition(1:2);

distances = vecnorm(relativePositions, 2, 2);

angles = atan2(relativePositions(:, 2), relativePositions(:, 1));

this.State = [distances; angles; decodingOrder / this.N_max; uavPosition'; this.currentStep];

end

function [powerAllocation, uavPosition, decodingOrder] = interpretAction(this, Action)

% Extract power allocation (first N_max elements)

powerAllocation = (Action(1:this.N_max) + 1) / 2 * this.P_max / this.N_max;

% Extract UAV movement (next 2 elements)

uavMovement = Action(this.N_max+1:this.N_max+2) * this.uavSpeed * this.delta_t;

currentUAVPosition = this.State(end-3:end-1);

uavPosition = currentUAVPosition + [uavMovement; 0];

uavPosition(3) = this.flightHeight; % Maintain constant height

% Decode SIC order (remaining elements, if any)

decodingOrderAction = Action(this.N_max+3:end);

[~, decodingOrder] = sort(decodingOrderAction);

end

function sumRate = calculateSumRate(this, powerAllocation, uavPosition, decodingOrder)

% Implement your sum rate calculation here

% This is a placeholder implementation

distances = vecnorm(this.userPositions - uavPosition(1:2), 2, 2);

channelGains = 1 ./ (1 + distances).^2; % Simplified channel model

sumRate = sum(log2(1 + powerAllocation(decodingOrder) .* channelGains(decodingOrder)));

end

end

end

Answer 1

[Steven Lord]

In the Guidelines for Constructors section on this documentation page, one of the guidelines is "If your constructor makes an explicit call to a superclass constructor, this call must occur before any other reference to the constructed object and cannot occur after a return statement."

Your code violates this guideline; you call the superclass constructor after assigning values to the constructed object's properties. Move the "this.<property>" assignments to after you call the superclass constructor.

[Felix Obite]

Many thanks Steven for your support and prompt respose. However, the issue still persist

classdef NOMAEnv < rl.env.MATLABEnvironment

properties

% System parameters

numSubchannels = 6; % Number of subchannels

P_max = 20; % Maximum power (watts)

N_max = 5; % Number of mobile NOMA users

timeSteps = 10; % Number of UAV time steps

uavSpeed = 5; % UAV speed (m/s)

flightHeight = 100; % UAV flight height (meters)

radius = 100; % Area radius (meters)

% Additional properties

userPositions;

currentStep;

T = 100; % Total duration

delta_t; % Duration of each time slot

% Optimal solutions (for comparison)

optimalPowerAllocation;

optimalTrajectory;

optimalDecodingOrder;

optimalSumRate;

end

properties(Access = protected)

State;

IsDone = false;

end

methods

function this = NOMAEnv()

% Initialize Observation and Action info

ObservationInfo = rlNumericSpec([2 * this.N_max + 2 * this.numSubchannels + 3, 1]);

ObservationInfo.Name = 'User distances, angles, current UAV position, time step, and subchannel states';

ObservationInfo.LowerLimit = [-inf * ones(2 * this.N_max, 1); zeros(this.numSubchannels, 1); -this.radius * ones(2, 1); 0; 0];

ObservationInfo.UpperLimit = [inf * ones(2 * this.N_max, 1); ones(this.numSubchannels, 1); this.radius * ones(2, 1); this.flightHeight; this.timeSteps];

ActionInfo = rlNumericSpec([this.N_max + this.numSubchannels + 2, 1], 'LowerLimit', -1, 'UpperLimit', 1);

ActionInfo.Name = 'Actions for power allocation, subchannel selection, UAV movement, SIC order';

% Call superclass constructor

this = [e-mail](ObservationInfo, ActionInfo);

% Initialize properties after superclass constructor

this.delta_t = this.T / this.timeSteps;

% Load optimal solutions

load('optimalPowerAllocation.mat', 'optimalPowerAllocation', 'optimalTrajectory', 'optimalDecodingOrder', 'optimalSumRate');

this.optimalPowerAllocation = optimalPowerAllocation;

this.optimalTrajectory = optimalTrajectory;

this.optimalDecodingOrder = optimalDecodingOrder;

this.optimalSumRate = optimalSumRate;

% Initialize environment

this.reset();

end

function [Observation, Reward, IsDone, LoggedSignals] = step(this, Action)

LoggedSignals = [];

% Update power allocation, subchannel selection, UAV position, and decoding order based on the action

[powerAllocation, subchannelSelection, uavPosition, decodingOrder] = this.interpretAction(Action);

% Update environment state

this.updateState(uavPosition, decodingOrder, subchannelSelection);

% Calculate reward (sum rate)

Reward = this.calculateSumRate(powerAllocation, uavPosition, decodingOrder, subchannelSelection);

% Check if episode is done

this.currentStep = this.currentStep + 1;

if this.currentStep >= this.timeSteps

this.IsDone = true;

end

IsDone = this.IsDone;

Observation = this.State;

end

function InitialObservation = reset(this)

% Initialize user positions

this.userPositions = rand(this.N_max, 2) * 2 * this.radius - this.radius;

% Initialize UAV position

uavInitialPosition = [0, 0, this.flightHeight];

% Initialize decoding order and subchannel selection

initialDecodingOrder = randperm(this.N_max);

initialSubchannelSelection = randi([1 this.numSubchannels], this.N_max, 1);

% Reset current step

this.currentStep = 0;

% Update state

this.updateState(uavInitialPosition, initialDecodingOrder, initialSubchannelSelection);

InitialObservation = this.State;

this.IsDone = false;

end

end

methods (Access = protected)

function updateState(this, uavPosition, decodingOrder, subchannelSelection)

% Calculate distances and angles to users

relativePositions = this.userPositions - uavPosition(1:2);

distances = vecnorm(relativePositions, 2, 2);

angles = atan2(relativePositions(:, 2), relativePositions(:, 1));

this.State = [distances; angles; subchannelSelection / this.numSubchannels; uavPosition'; this.currentStep];

end

function [powerAllocation, subchannelSelection, uavPosition, decodingOrder] = interpretAction(this, Action)

% Extract power allocation (first N_max elements)

powerAllocation = (Action(1:this.N_max) + 1) / 2 * this.P_max / this.N_max;

% Extract subchannel selection (next numSubchannels elements)

subchannelSelection = round((Action(this.N_max+1:this.N_max+this.numSubchannels) + 1) / 2 * (this.numSubchannels - 1)) + 1;

% Extract UAV movement (next 2 elements)

uavMovement = Action(this.N_max+this.numSubchannels+1:this.N_max+this.numSubchannels+2) * this.uavSpeed * this.delta_t;

currentUAVPosition = this.State(end-3:end-1);

uavPosition = currentUAVPosition + [uavMovement; 0];

uavPosition(3) = this.flightHeight; % Maintain constant height

% Decode SIC order (remaining elements, if any)

decodingOrderAction = Action(this.N_max+this.numSubchannels+3:end);

[~, decodingOrder] = sort(decodingOrderAction);

end

function sumRate = calculateSumRate(this, powerAllocation, uavPosition, decodingOrder, subchannelSelection)

% Implement your sum rate calculation here

% This is a placeholder implementation

distances = vecnorm(this.userPositions - uavPosition(1:2), 2, 2);

channelGains = 1 ./ (1 + distances).^2; % Simplified channel model

sumRate = sum(log2(1 + powerAllocation(decodingOrder) .* channelGains(decodingOrder)));

end

end

end

[Felix Obite]

I attached the script for the learning environment with the error message: Error: File: NOMAEnv.m Line: 42 Column: 20

Calling the superclass constructor 'rl.env.MATLABEnvironment' after an object use or after a

return statement is not supported.

[Steven Lord]

Let's look at the constructor up to and including the superclass constructor.

function this = NOMAEnv()
% Initialize Observation and Action info
ObservationInfo = rlNumericSpec([2 * this.N_max + 2 * this.numSubchannels + 3, 1]);
ObservationInfo.Name = 'User distances, angles, current UAV position, time step, and subchannel states';
ObservationInfo.LowerLimit = [-inf * ones(2 * this.N_max, 1); zeros(this.numSubchannels, 1); -this.radius * ones(2, 1); 0; 0];
ObservationInfo.UpperLimit = [inf * ones(2 * this.N_max, 1); ones(this.numSubchannels, 1); this.radius * ones(2, 1); this.flightHeight; this.timeSteps];
ActionInfo = rlNumericSpec([this.N_max + this.numSubchannels + 2, 1], 'LowerLimit', -1, 'UpperLimit', 1);
ActionInfo.Name = 'Actions for power allocation, subchannel selection, UAV movement, SIC order';
% Call superclass constructor
this = [e-mail](ObservationInfo, ActionInfo);

Anywhere in the copied code, do you have a return statement? No.

Anywhere in the copied code, do you use the this object? Yes. There are five such references (used in a couple places) that I can see.

this.N_max
this.numSubchannels
this.radius
this.flightHeight
this.timeSteps

If those were Constant properties and referred to them as NOMAEnv.N_max, NOMAEnv.numSubchannels, etc. I believe that you would be able to use them to create your ObservationInfo and ActionInfo variables without MATLAB considering them to be uses of the object for purposes of that guideline. It looks like you never assign a value to any of those properties other than in the properties block so making them Constant may be appropriate.

You could also refer to those properties using an instance of the object later on, so you don't have to modify the rest of your code. NOMAEnv.N_max and this.N_max (assuming this isa NOMAEnv object) should both work to access that property (except in the constructor while you're in the process of instantiating the object, where you want to access it via the class name not an instance.)

[Felix Obite]

Many thanks Steven, this has helped to resolve the issue. I appreciate your support