Here is a very rough outline of our schedule. This list is just an indication and will be updated on a regular basis. We have a few more intermediate stages in mind.

Objectives and status

The following objectives are the most prominent ones. We have a few more but these you can find in our publications. As of version 1.0 most of these objectives have been reached but we keep going.
Merging engines
Objective: Merge the Aleph and pdfTeX codebase so that LuaTeX in DVI mode behaves like Aleph, and in PDF mode like pdf(e)TeX. Also allow Lua callbacks for file searching.
Status: This stage is mostly finished. LuaTeX 0.40 and higher start up as TeX82 and one has to explicitly enable extra primitives (normally this will be done by the macropackage).
Objective: Some functionality is still tagged as pdfTeX or Omega while in practice it concerns generic functionality. The low level code as well as some primitives can be stripped from these properties. At the node level this means that some whatsits will become first class nodes.
Status: This will happen from version 0.80 onwards and take some time. It is part of the general cleanup (and also involves removal of experimental code from older engines).
OpenType fonts
Objective: Make the Aleph functionality available in PDF mode and add support for OpenType fonts. Allow Lua scripts to control all aspects of font loading, font definition and manipulation. Core font support completely driven by Lua remains a core objective (in addition to the traditional, downward compatible TeX method).
Status: Most of this is finished. We use code from the FontForge project which has the advantage that one can use that open sourced tool alongside LuaTeX and get a similar view at the font. The already stripped FF library will be further stripped.
Tokenizing and node lists
Objective: Use Lua callbacks for various internals, complete access to tokenizer and provide access to node lists at moments that make sense.
Status: This stage is to a large extent completed (around version 0.50) but you can expect a few more callbacks when the internals have been cleaned up. An additional node access model has been introduced in 2013 and a better tokenizer interface has been implemented in 2015, including scanner functions.
Paragraph building
Objective: Provide control over various aspects of paragraph building (hyphenation, kerning, ligature building), dynamic loading loading of hyphenation patterns.
Status: Apart from some small details these objectives are met. Users can write their own paragraph builders (although this is a somewhat specialized task). At some point we will provide more context for each paragraph. We expect to have reorganized this in version 0.90.
Metapost (mplib)
Objective: Turn metapost into a library. Extend the library to arbitrary precision (megapost) and add features on the wishlist. Advanced usage in LuaTeX, like for instance runtime font generation and manipulation is not yet scheduled.
Status: The first stage of this project is finished with the first public beta and integration in LuaTeX. This subproject has been finished on schedule by the end of 2009. Future versions of LuaTeX will ship with the latest versions of the mplib library. For instance in 2013 we got support for doubles and in 2014 the decimal and binary number systems showed up. Later in 2014 there is access to Lua from this library. The PNG and SVC backend are no longer enabled in LuaTeX which greatly reduced the depedency of third party libraries.
Image handling
Objective: Image identification and loading in Lua including scaling and object management.
Status: This first version of the img library showed up in the 0.22 beta and some more features have been added after that. Inclusion of PDF has been optimized in 0.50. In 1.10 we expect to have replaced the dependency on Poppler for embedding external PDF images by a dedicated small lightweight PDF access library that will also replace the now somewhat inconsistent epdf interface. This is part of the onguing effort to make LuaTeX as independent if other code as possible (the long term stable TeX narrative).
Multi-directional typesetting
Objective: Cleanup the internals, remove bugs in the Aleph code and review the backend part. The model stays mostly as in Omega.
Status: This is an ongoing effort and the first major cleanup took place around versions 0.40 and 0.50. The code has already been debugged and simplified. We now only have four directions. Some aspects of the page, body, text and par model might change and information currently stored in whatsits might move to normal nodes. This is projected for versions 0.80 or later. The current versions have native direction nodes and a numeric as well as symbolic interface.
Special features
Objective: Cleaning up of HZ optimization and protruding, optimize resources in the PDF output. Get rid of remaining global font properties. Cleanup and optimize the backend code.
Status: Much of this was done in 0.50 (more efficient pdf output). In 0.78 a major change took place when the front- and backend aspects of this mechanism were separated and overhauled.
Objective: Extending math to Unicode and providing more control over rendering. Support OpenType math cf the Microsoft specification. Provide primitive support for some features that are currently dealt with by macros.
Status: This effort started at the end of 2008 and is mostly finished mid 2009 in version 0.40 and became quite stable in 0.50. We have a few extensions on the agenda but they are subprojects. These developments happen in parallel to the development of more math fonts in the TeXGyre project. We keep adding more control over rendering, also because some opentype fonts behave more like traditional fonts.
Page building
Objective: Control over page building and access to internals that matter. Access to inserts.
Status: This is on the agenda for 2014 at the earliest (maybe version 0.90).
TeX primitives
Objective: Access to and control over most TeX primitives (and related mechanisms) as well as registers will be provided. There will be a clear split between front and backend.
Status: This is an ongoing effort and much already has been done.
PDF Backend
Objective: Most backend related features will be opened up. Think of annotations and object management.
Status: The first code will show up later in 2009. In version 0.50 already quite some functionality is accessible at the Lua end. In 0.79 a lot of code has been normalized and cleaned up. There is still a todo list, so expect more.
Future stage
Objective: Convert the TeX Pascal source into C (CWeb) and start using Lua as glue language for components.
Status: This has been tested on MPlib. Parts of LuaTeX are already rewritten in C, for instance some memory management, IO, language support, par building, hyphenation, font loading and math. Version 0.60 is the first CWeb version because it is mandate for some of the things we have in mind. Improving the result is an ongoing effort.

Versions and projections

In addition to the mentioned functionality we have a couple of ideas that we will evolve and be implement along the road. We had some releases and there will be more:

Version 0.00
We first demonstrated LuaTeX in TUG 2005 in Wuhan (China). At that time we had only the scripting engine available and access to some registers. We could print back to TeX.
Version 0.10
The first major (beta) release was at TUG 2007 in San Diego (USA). This version could handle fonts and we demonstrated the first attempts of dealing with Arabic typesetting as part of the Oriental TeX project.
Version 0.20
The second major release (still beta) was around TUG 2008 in Cork (Ireland).
Version 0.30
This version showed up as prelude to version 0.40 and provided preliminary support for OpenType math. This version was the basis for successive (incompatible) beta versions that got more and more implemented of our relatively long wishlist.
Version 0.40
This is the BachoTeX 2009 release that will also be shipped with TeXlive 2009. At BachoTeX we will officially release the new math enhancements as this is the place where discussions started with regards to the new OpenType math companion fonts.
Version 0.50
Also known as the halfway version, version 0.50 will show up around EuroTeX 2009 that takes place in Delft in the Netherlands. This version will be a polished version 0.40 with some pending cleanup being done. It is the prelude to 0.60, the CWeb version.

For practical reasons the official release date has been moved forward to after TeXlive is frozen and distributed (TeXlive ships with version 0.4*). Version 0.50 has been released just before xmas 2009.
Version 0.60
This version does not offer fundamental new features compared to 0.50. Released begin April this iteration will go on TeXlive 2010 and for that stability is a condition. There are a few more callbacks and functions, some cleanups in the backend and the font inclusion has been improved. We are confident that this version is quite useable. At this point the source code has been converted back into web so that more detailed documentation can be added.
Version 0.65
This was an intermediate release with lots of small fixes and a few new features. The release was late 2010 (around the ConTeXt conference in Czech).
Version 0.70
This version came available around the BackoTeX (also EuroTeX) 2011 conference in Poland and will end up on TeXlive 2011. This is a stable version, suitable for production. No fundamental new features have been added. Macro (package) writers should be prepared to adapt (node and backend related Lua code) from version 0.71 onward.
Version 0.75
Around this version, which came available end 2012 we got Lua 5.2 on board. Experiments have shown us that this version runs somewhat faster than previous versions and also uses a bit less memory. Of course there are the usual bug fixed and some tracker items have been resolved. There is an updated MetaPost library that supports an additional number mode using doubles.

At the same time the first beta of LuajitTeX was released (version 0.72, Lua 5.1). This version uses LuaJIT. This version is not portable to all platforms. Some benchmarks with ConTeXt have demonstrated that runs can be 20-35% faster than a regular 0.72 version. Most of the speedup is due to the faster virtual machine as well as the garbage collector (which is similar to the one in Lua 5.2). You can use both version alongside but LuaTeX with Lua 5.2 is the official standard. This is a stable release.

Version 0.77
This is the version that is on TeXlive 2013. There are no fundamental changes and additions, but bugs have been patched. Because the engine is now more widespread we cannot make big changes unless macro package writers have time to adapt. Of course, we can still change aspects of the interfaces.
Version 0.78
This is the end-of-year version in 2013. Apart from some bug fixes, the most important change is the (first) cleanup of the hz optimization (expansion), which is a prelude to a better separation between front- and backend. There are also some speedups in node processing (that have been tested for more than half a year now). The remains of AAT font technology have been removed and the font loader is somewhat more robust. As usual, the latest MetaPost library is included.
Version 0.79
This is mostly be a bug-fix release. An experimental new token scanner interface has been demonstrated in 2014 and added to 0.79.3 onwards beginning of 2015.
Version 0.80
This version is the offical TeXlive 2015 version. Apart from some bug fixed the major new thing is a (new)token scanner that will replace the old one, so we have a transition stage. Lua 5.2.4 is used now.
Version 0.81
This version, published shortly after the 9th ConTeXt meeting, has a couple of bugfixes but more important, it is a first step in cleaning up some primitives: some have been removed, others have been promoted to core primitives. Lots of small improvements have been done and some extra helpers have been added at the Lua end. The manual has been overhauled too.
Version 0.82
This is just an experimental version for ourselves with lots of pending things done, cleanups and improvements.
Version 0.85
This end-of-2015 version is a prelude to 0.90 that will be available early 2016 and that will end up on TeXLive 2016. There is new math functionality, there are some extra callbacks, messages have been cleaned up, some extra helper are provided but most important is that the backend primitives are better isolated from the core code and that some extensions (and whatsit nodes) became core as well. There have been several incremental updates in for this version.
Version 0.90
This version is the promoted 0.85.5 and will serve as the test version for TeXLive 2016. This also means that there will be no fundamental changes. In fact, we're close to a stable 1.00 now and if there are drastic changes, they will happen after version 1.00 (apart from internal cleanups and cosmetic improvements).
Version 0.95
This version will carry some low priority (no showstopping) pending issues. It will be available around TeXLibe 2016 (and might even be the version shipped then).
Version 1.00
This version has been released at the 10th ConTeXt meeting in 2016. The version numbers will now use steps .01 and updates will mostly concern fixes or small extensions. New experiments and fundamental changes can happen in the experimental branch.
Version 1.04 and 1.05
This 1.04 version is the yearly TeXLive update. It has a couple of bugs fixed, a few helpers added, somewhat more control over formerly hard coded math properties, and some of the internals cleaned up. The 1.05 version will be released around the 10th ConTeXt meeting in 2017.
Version 1.07
TeXLive 2018 carries the 1.07 version. There have been no fundamental changes. Apart from bug fixes some more control over for instance math rendering has been introduced because we need to deal with fonts that can behave more like traditional fonts than opentype fonts.
Version 1.08 (context)
Because we had quite some pending small details (completion as well ass consistency) in the pipeline, this version was released shortly after TeXLive 2018 so that ConTeXt users could test it in production. There have for instance been some low level level changes in direction nodes. The interfaces are downward compatible. This version also defaults to Lua 5.3 which gives a small performance boost. This instance of 1.08 is not official and mostly tested as part if the ConTeXt distribution.
Version 1.08 (trunk)
Originally 1.09 was to be released around the ConTeXt meeting but we as we also want to support the arm architecture, we decided to push an updated 1.08 into trunk with the planned 1.09 functionity. This makes 1.08 an official release (only in trunk has the official releases!). Among the important changes are replacing poppler (used for PDF inclusion) and dropping binary mode from MPLib which simplfies the code base and compiling. We will push 1.09 a bit later when we're sure that arm also works well.
Version 1.09
This version will have the same functionality as 1.08 but with working ffi support for the arm platform. Of course bugs will be fixed too. This version is the prelude to 1.10 for TeXLive. Depending on the Lua 5.4 release date, we might add this update so that it can be tested.
Version 1.10
This is the TeXLive 2019 release. Because the engine is now used in several macro package, we cannot change the interface any longer. This is therefore what you get. Of course we will fix bugs and occasionally might add some functionality, improve performance, but for now stability of this engine has our priority.
Version 1.12
The 1.11 release was an interdiate bug fix release, while the 1.12 release is for TeXLive 2020. It uses pplib version 2. There are a few non-intrusive additions so compatibility should be okay.
Version 1.*
Further 1.* releases can happen around TeXlive code freeze moments, normally end February. There are seldom fundamental additions but for instance the used libraries get updated and the build gets adapted to changes in the environment.

We report progress on a regular basis at usergroup meetings and in usergroup journals so if you want to be kept informed, you might consider becoming a member of a TeX usergroup or visiting one of the TeX conferences.

For along while LuaTeX stayed beta till we are more or less functionally complete. It took a couple of years before we reached that state. Our original estimate was that we could provide version 1.0 in 2012 around TeX's 32th birthday, but we ended up with a lower number. Although major releases (x.x0 versions) are stable enough to be used for production work you need to keep in mind that only at version 1.00 most of the functionality and interfaces is considered to be stable. Don't automatically expect your old code to work with newer versions of LuaTeX: best update on a regular basis. With TeX being over 40 years old around LuaTeX 1.10 we can conclude that we're a bit over time, but intermediate versions always performed well.

It is important to know that starting with version 0.40 LuaTeX will start up with only TeX82 functionality. The \directlua primitive will be available for testing the available functionality and further initialization of features.

For testing we use an experimental version of ConTeXt tagged MKiV. Support is provided on the wiki and mailing list. ConTeXt users are advised to keep their LuaTeX binary in sync with the latest ConTeXt MKiV. The best way to do this is to use the minimals from the contextgarden as binaries are automatically updated there. Users of other macro packages can best consult the package related sites and lists.

The bigger picture

Although we frequently update the (potential) audience by means of articles in user group magazines and presentations at user group meetings, it makes sense to spend some words on topics that come up occasionally.
The landscape
There are currently three engines in use in the TeX community: pdfTeX, XeTeX and LuaTeX and they all serve an audience.

The pdfTeX engine is a stable extension to TeX that has eTeX on board as well as a few extensions. It provides hz and protruding and supports advanced PDF output. It is fast and reliable and sort of frozen with respect to functionality. It will stay around for ever but has only limited provisions for dealing with OpenType fonts.

Then there is XeTeX which supports Unicode as well as OpenType by means of third party libraries. It integrates nicely into the current infrastructure and support from macro packages is easy as there are no fundamental changes in interfaces and functionality. It uses the eTeX r2l typesetting while LuaTeX uses the Omega (Aleph) method. Support from macro packages does not demand changes in the core.

The LuaTeX project uses a different approach. Instead of using libraries it provides an extension mechanism. This keeps the program independent and permits the flexibility that we like so much. However, it comes at a price. First of all LuaTeX does not have the performance of its cousins, although for instance the average ConTeXt MkIV run can be faster in LuaTeX. LuaTeX also demands extensive support at the macro level if you want to profit from its benefits. Just adding some support for scripting is nice but the power of LuaTeX only shows when it's tightly integrated. This integration is also needed in order to provide macro users some stability as most of them are no TeX programmers. If such integration is not what you want, you might consider sticking to the other engines as there are no real advantages to using LuaTeX then.
As it possible to replace the current paragraph builder by a Lua function and this is the direction that we move to: LuaTeX as framework with only built in TeX82 support, some extended core functions (hyphenation and math) and provisions to extend the more debatable bit and pieces with your own code. One of the main objectives of the oriental TeX project is to add advanced r2l typesetting to ConTeXt MkIV and part of that extensions to the par builder written in Lua. There will be no extensions to the built in typesetting engine as we are confident that the Lua interface is flexible enough. An exception is math, where opening up as well as support for OpenType math demands some more control.
We ship with a minimal number of built in libraries in addition to the standard libraries: a fontloading library cf. FontForge, a few basic IO related libraries, some Unicode helper code. These have been adapted to our needs and are statically linked into LuaTeX. We try to keep the default featureset small and this is already a bit larger than we had in mind.

The 0.60 release provides can load additional libraries but these are not part of the projects as there are multiple choices as well as interfaces to these libraries possible. Demand and support is probably quite related to macro package demands and LuaTeX is (to be) independent of macro packages. There are some known limitation with respect to library versions related to the embedded Lua variant. These will be sorted out later.

Upto version 1.0 we use stock Lua 5.2 which will make it possible to provide stable library loading. We are experimenting with Lua 5.3 and will support compilaiton of both versions. This permits macro package writers to adapt when needed. When LuaTeX is used in production it makes sense to archive the version that you know works well. A formal switch to 5.3 will probably happen in TeXLive 2019.
We have chosen for a Lua based support for fonts. There are two solutions for handling fonts: using the internal functions that do what TeX has always done, or write a Lua function that does a different job. As there are multiple solutions possible and as we expect macro packages to have their own ways of dealing with fonts, there is not one solution for dealing with fonts anyway. Also, TeXies have always wanted full control over matters, and this is provided by the Lua solution.

There will be no hard coded solutions and experiments with helper code as show us that this is of no use as switching between Lua and TeX adds enough overhead to nil the advantage of such helper code. It's more efficient to stay either at the TeX end or at the Lua end. Of course we will continue to experiment with alternative ways but this is unlikely to happen before version 1.0 (after that we will formulate our LuaTeX 2.0 objectives and set a new agenda).
Plain TeX
As part of the project we provide some basic (non-intrusive) functionality like for instance OpenType support for plain TeX as we use this for testing some code. This basic code is derived from ConTeXT MkIV code which serves as our benchmark (the larger testbed is the Oriental TeX project). The code is part of the ConTeXt distribution but can be used independently. The first code showed up alongside version 0.40 (in TeXlive 2009) and is kept in sync with the ConTeXt code. Based on that support for LaTeX is provided.
You might have noticed that we really take our time to complete this project. This is partly because it is non-trivial to rewrite TeX in such a way that existing macro code still works. As LuaTeX development goes in parallel with rewriting ConTeXt (MkIV) we test everything new and opened up extensively. We spend much time discussing and trying out solutions. We do and redo when needed and try to stay generic as much as possible. We also document what we've done and report on it as well as maintain a manual. This simply takes time and we're talking of many (mostly unpaid) man-years here. We feel that this is the only way to arrive at what we aim at with LuaTeX, which is important as we ourselves are among its users. In the perspective of 30 years of TeX the timeline is quite feasible. Of course, as intermediate versions are quite useable. Although we originally planned to release version one in 2012, it happened end of 2016. We will add features and when possible improve performance. Experiments will be carried out in the ConTeXt development scene.
Traditional TeX has its own output format (DVI) which is then postprocessed into for instance specific printer code, the more generic PostScript or PDF. With pdfTeX one can directly produce PDF and as LuaTeX is derived from pdfTeX is also produced PDF. Because one has access to the internals additional backends are possible (using Lua).
Both engines now provide support for ffi. It is still somewhat experimental in LuaTeX but we aim at robust support in future versions.

This page has not been updated much since version 1.10. Not all objectives and developments are described here. As part of ConTeXt MkIV development, the most important LuaTeX developments have been reported at user group meetings and reported in articles.

In 2023 we're at version 1.16 and the engine is basically frozen. There are occasional extensions to keep it in sync with the other engine but nothing fundamental will be added. This is a prerequisity for usage in multiple macro packages that depend on long term stability. A side effect is that the roadmap will not be extended.

The LuaTeX project is closely related to ConTeXt and the focus of further (more fundamental )engine development is now shifted to LuaMetaTeX. Again there is more information in the ConTeXt related documents, articles and presentations.