Why does this error occur when using rlDQNAgent "First argument must be a Q value function object"

2 次查看（过去 30 天）
Emmanuella
Emmanuella 2023-9-13
评论： Emmanuella 2023-9-20
I am new to the reinforcement toolbox and I am trying to simulate a mec environment that observes 3 states from the environment, that is, a scalar value SNR, a scalar value data, and a 1x3 row vector channel gain and reproduces an action from the range of 0 to 1 which determines the amount of data to be offloaded (offloading factor). My observation info is a cell array which I have concatenated to become a vector when I run my custom environment code. But I keep on getting this error " Error using rlDQNAgent
First argument must be a Q value function object, a Q value representation object, or
an observation specification created using 'rlNumericSpec' or 'rlFiniteSetSpec'
objects.
Error in DRL_trial_dLL (line 52)
agent = rlDQNAgent(layers, obsInfo, actInfo);" I have attached my codes

请先登录，再回答此问题。

回答（1 个）

Sachin Lodhi
Sachin Lodhi 2023-9-18
Hi Emmanuella,
Based on the information provided, it appears that the issue you are encountering is related to incorrect parameters being used with the “rlDQNAgent” function. The first parameter of “rlDQNAgent” should be an ObservationInfo object of type “rlFiniteSetSpec” or “rlNumericSpec”, or an array containing a combination of such objects. However, it seems that in your code, you have provided a layer as the first input argument, which is causing the error message to appear.
If your intention is to use the network (layers) that you have created, you will have to create a critic using the “rlVectorQValueFunction” and then pass that critic as the first argument instead of the layers. This adjustment should allow your code to run without encountering any errors.
I recommend referring to the following documentation that provides an example of creating a DQN Agent using a critic.
Deep Q-network (DQN) reinforcement learning agent - MATLAB (mathworks.com)
I hope this helps in simulating your environment and resolves the error.
  1 个评论
Emmanuella
Emmanuella 2023-9-20
Thank you for your response Sachin, but the error still persist even when I change the first parameter to my observationInfo object. I think the error is because my observationInfo contains both scalar values and matrices. When I use either scalar values only or matrices only, it works, but I get that error when I use both.

请先登录，再进行评论。

类别

在 Help Center 和 File Exchange 中查找有关 Reinforcement Learning 的更多信息

标签

产品


版本

R2022a

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by