I've been advocating a routable peoplemover paths as the answer for years. Think 'automated baggage handling system'. The system would include both low speed and high speed zones, full automation, and vehicles destinations are determined by guest input.
The problem with peoplemover in general is the distances involved. The system must be multispeed and needs multiple routes to be practical. The ability to reuse the same pathes for different routes would be essential in costing it all.
It's already been done in smaller scale - it just needs to be built up to family sized carriages with the safety systems needed for humans.
At grade with elevated crossings is probably preferred as it minimizes evacuation issues, but introduces new obstacle avoidance and breakdown issues. A possible compromise would be to have an elevated track, but not one like the monorail where it's beam only and no need for it to be so high the entire distance.
If you go to Bombardia's website, you will see that they have higher capacity peoplemover systems, like the Innova 300, so near-monorail capacity is what I have in mind. If more capacity is required, just add more trains. The on-demand service you're thinking of is called PRT. There is one PRT system in the US and that is in Morgantown, WV and has been highly successful. Another PRT system is being built in Abu Dabi, UAE. That system runs underground, uses rubber tires and is powered with electric batteries. It's like an automated golf cart.
The problem with having low capacity PRT at WDW is that it won't meet the required capacity demand, even if they served only the resorts. GPRT (Group Personal Rapid Transit) is more viable, but still won't meet capacity demand. The Universal Health peoplemover system is a better fit, and it even has PRT-like features, where people could press a button at stations to call a train to stop there and retrieve those waiting at the station.
In Broward County, the Port Authority is planning on building a six-mile peoplemover system connecting the port to Ft. Lauderdale International Airport. At the same time, the state is looking at building an at-grade light rail that will connect the local Tri-Rail commuter rail system with a new commuter rail line using FEC tracks, located on tracks a few miles east. Also, a local group sponsored by Broward County plans on building a network of street cars for a circular around downtown Ft. Lauderdale. All of these systems have corridor routes that overlap. All will use different tracks - none of which are shared.
The above is an example of pure stupidly on the planning level, in my opinion. The Port Authory's peoplemover system will be one of the largest peoplemover systems in existence, if not the largest. Since all these Broward systems overlap, why not just built the other systems as extensions to the Port Authority"s system? It would be more cost effective, since a single track would only have to get built, instead of several, wherever there would have been overbuilt track segments. I bring this up because this is something like what you were talking about, but I agree with you when applied to my Broward example.
At-grade can work if an enclosure is built over it. This is best applicable, if built on or over the service road of Future World on the west side. My only question is if this is done, how would the west side of Future World be serviced? There could be special servicing peoplemover vehicles that could run on the same tracks, or build the tracks above the service road.
Or run the peoplemover through the underground utilidores.