. It says the optimal policy is to solve the linear-quadratic problems of the form in picture. But as I know, the objective function is loss function. So it should be “min”, but not “max”.
Could anyone can help me and tell me what does the objective function is? Is it loss function or welfare function?
If its the welfare function ,then it should be “max”.

In the OSR context, it must indeed be a min here, because we are minimizing a linear combination of second moments (think of variances). In the 4.5.7 manual this is correct.

Indeed, the objective function in OSR is the second moments, such as lambdavar(yt) + var(pit). But my objective is lambda(yt - y*)^2 + (pit - pi*)^2. Where y* is the target value of yt. So could you help tell me how I can write the code of this objective function?