alexshafranov / derplanner

Hierarchical-Task-Network planner in C++

derplanner

derplanner is a:

• C++ game AI decision making library.
• Hierarchical-Task-Network Planner.
• Compiler, turning the custom domain description language into C++ code.

Here's a sneak peek at domain syntax:

domain turret {
  fact location(vec3, vec3)
  fact target(id32, vec3)
  
  prim fire!(id32)
  
  const Fov = cos(pi() / 4.0)
  
  task attack_visible() {
    each :sorted(1.0 - Dot)
      location(Pos, Dir) & target(Id, Tgt) &
      (Dot = dot(norm(Tgt - Pos), Dir)) & (Dot >= Fov) -> [ fire!(Id) ]
  }
}

Building

derplanner has no external dependencies, so it can be included into your own build system.

Alternatively, you can use premake5 executable shipped with the project to generate project files / makefiles.

• On Windows run premake5 vs2015. Any Visual Studio version, starting from Visual C++ 2008 should work as well.
• On Linux run ./premake5 gmake. Linux build is regularly tested on gcc 4.6.3 and clang 3.4.

Quick Intro

Types

• id32, id64 32-bit and 64-bit handles, to reference entities.
• int8, int32, int64 signed integers.
• float 32-bit floating point.
• vec3 3-component vector type.

Fact Database

Fact database is a collection of typed tuples, representing domain knowledge about the world.

Domain database requirements are specified in the following way:
fact { location(vec3) target(id32, vec3) }

Primitive Task

Plan is a sequence of primitive task instances, where task instance is a task type and an argument tuple:
goto!(A) melee!(T) goto!(B)

Primitive task is defined in the following way:
prim { goto!(vec3) melee!(id32) }

Compound Task

Compound tasks make the planning process hierarchical and recursive.

Compound task defines one or more cases, where each case is a pre-condition and a task list:

task attack(Bot, Enemy) {
  case pos(Bot, Src) & pos(Enemy, Dst) & dist(Src, Dst) < Close_Range
    -> [ approach(Src, Dst), melee!(Bot, Enemy), retreat(Bot) ]
  
  case line_of_sight(Bot, Enemy)
    -> [ select_weapon(Bot), fire!(Enemy) ]
}

Task list can contain both primitive and compound task instances. It defines the way to execute a parent compound task.

Planning Process

The plan is formulated in a simple left-to-right and top-to-bottom way:

• Initially, the plan is a single compound task (the first one defined in a domain).
• For each compoud task instance in the plan:
  • For each case, in order of definition:
    • If the case precondition is satisifed, replace the compound task instance with the task list.
  • If no satisified cases found, back-track.
• Once the plan consists of only primitive tasks, the planning process is finished.

License

derplanner is licensed under zlib license

derplanner's Issues

Supporting the HDDL format

https://ulm.de/fileadmin/website_uni_ulm/iui.inst.090/Publikationen/2020/Hoeller2020HDDL.pdf

I was wondering if it was worthy to support the HDDL format.

How to use?

Hello,
This looks very interesting, but I don't know how to use it.

I was able to compile block.txt and travel.txt, but don't know what to do after that. I assume you add block and travel's C++ source files to the derplanner-runtime section, but after that I have no idea.

A sample main.cpp file that calls them and gets the steps of the plan would be very helpful.

Question - effects and prediction

Hi,

Great work on the library. I don't see any reference to effects in the code. Did I miss them? If not, any guidance on how I could add it?