To get fusion there is a couple of ways: There is gravitational confinement (that is how stars do it, but it is impractically large for humans), there is inertial confinement (get enough energy out before it explodes, but that only works for hydrogen bombs and NIF, not for a reactor) and magnetic confinement (that is what we intend to use for reactors).
Magnetic confinement works because a plasma consists of charged particles that gyrate around the magnetic field lines. So (to first order) they can not escape across field lines. But they can move along the field lines and hit the end of the device. The particles move fast, so simply making a linear device long enough is hard. So the next idea was to bend the magnetic field into a torus (the shape of a donut), because that way there is no end to the magnetic field lines. Unfortunately that configuration is unstable, the plasma donut will bend and twist until it hits a wall, stops being a plasma and falls to the ground. There is basically three option out of this problem:
1.) Tokamaks [1] such as ITER. Here we induce an additional current inside the plasma that goes around the hole of the donut. That produces a small additional field that stabilized the current. But driving that current can be hard (using what is called a plasma transformer work only for a limited times, but AFAIK that is not the first limit on a discharge that ITER will hit, wall heating limits single plasma "shots" to shorter times anyway). The upside is that the design is rather simple, which implies you don't need millions of core hours to design and the fields coils are reasonably easy to produce.
2.) Stellarators [2] such as W7-X. Here the field coils that produce the field lines in the donut are intentionally twisted to produce a more complicated magnetic field. The upside is that we do not need the current in the plasma and get better performance (for a device of similar size), but the design of the field coils is hard (impossible back in the 60ies and still really hard even with modern computers) and the production of the coil is not simple either. You can actually include "producability" as an optimization goal along with plasma performance in your design code, but even then you will have to build a large number of different coil designs.
3.) Active control. It takes some time for the plasma donut to bend and twist. Typically a few milliseconds. So if you stick a large number of sensors and computer controlled coils around the plasma you might be able to continuously keep the plasma confined, just like balancing a pencil on its tip. This was of course utterly unimaginable back in the 60ies, and even today there is mayor problems. Sensors aren't fast enough, optimal (or even good) control algorithms are unknown and rapidly ramping megaamperes in the control coils is hard if you don't want to rip them out of the device accidentally. And you probably only have ~ 10 failed attempts before the wall of the vacuum device is compromised and you need a new device. Consequently there is some small scale research on that (sorry I don't have a cool link handy), but nobody is trying that on large devices for now.
[1] https://en.wikipedia.org/wiki/Tokamak
[2] https://en.wikipedia.org/wiki/Stellarator