An ILS, consisting of a localiser beam to provide lateral guidance (which can also be used on its own for a non-precision approach) and a glideslope beam to provide vertical guidance, is (currently) indispensable for precision approaches in IMC, but comes with a few disadvantages.
Chief among those is the fact that an ILS only supports approaches that are ramrod-straight all the way in from the outer marker, when such an approach path is often impossible or undesirable...
- ...sometimes because something too solid to fly through is in the way (for instance, the infamous approach to runway 13 at Kai Tak, which, due to an inconveniently-placed mountain, necessitated coming in at a large angle to the runway and then making a sharp right turn at very low altitude at almost literally the last moment; an instrument guidance system - basically an ILS without the “landing” part - was added in the 1970s to provide precision guidance for the part of the approach before the turn, but the turn itself and the final approach segment still had to be hand-flown visually, an exceptionally difficult task even in good weather)...
- ...and sometimes because someone else is using part of the airspace in question (for instance, the four major airports in the New York area [JFK, la Guardia, Newark, and Teterboro] have numerous intersecting approach paths, which, as the approaches can’t be bent or segmented to keep out of each others’ way, frequently forces la Guardia and Teterboro to land aircraft on runways with strong crosswinds [because the ILS approach paths for their more optimal runways would risk causing MACs with aircraft on ILS approaches into JFK and/or Newark]; on the other side of the continent, the approach paths coming into LAX from the east pass near or over Ontario, which has caused quite a number of near misses in the vicinity of the latter airport).
It doesn’t seem, conceptually, like it should be too hard to break up an ILS approach into two or more segments, each with its own localiser and glideslope, and with the beams from the pre- and post-bend segments intersecting at the bend in the approach path. As most aircraft are unable to turn on a dime, they would need to start turning a bit before the bend in the approach, but this could be easily handled by having another instrumental aid (for instance, a third localiser) indicate when to start turning; if the bend in the approach were sharp enough for the aircraft to have trouble locking onto the post-bend beams, another localiser and glideslope, with their beams tangent to the midpoint of the turn, could be used to assist the aircraft round the bend.
All of this would also work for segmented localiser approaches, with the added advantage of being easier to implement (no need for the multiple glideslope antennae, just the localiser ones).
Here be an illustration of what I’ve got in mind:
Why aren’t segmented ILS and localiser approaches used in practice?