In my problem, qo(2,1) is a CVX variable and q(2,1) is a constant, my expression is (1 / ||qo - q||). When I writed inv_pos(norm(qo(:,u)-w(:,k))), the error " Illegal operation: pow_p( {convex}, {-1} )" happend. How can I express (1 / ||qo - q||).

Can you multiply through by `norm`

and be in compliance with CVX’s rules?

`inv_pos`

requires its argument to be concave (or affine). So a non-affine convex argument such as `norm `

is not allowed.

In any event, have you proven your optimization problem is convex?