A good place to start would be looking at various pharmacophores for each receptor and looking if overlap is possible.
Due to the explosion in research chemicals in these 3 classes, you can examine many scaffolds and look for commonalities between them.
I caution using swisstarget prediction to guide you, as only interactions with the likeliehood bar filled more than 90% will be druglike.
There are many structure activity relationship (sar) studies on agonists of each receptor that have done most of the work, and will tell you motifs necessary for binding each receptor.
I guess a more general thing is to familiarize yourself with
Lipinski's rule of 5, which is a heuristic for drugs that actually manage to get into the brain at an appreciable level.
Designing such a drug that you propose is honestly a herculean task. If somebody proposed this as a pHD project, it would be shot down due to unfeasibility to do within a 5-7 year timeline.
May be much easier trying to design molecules where there is already known overlap such as DAT ligands that are also NMDA antagonists.