Better support for humans

[severin-lemaignan]

This issue tracks the development of human(s) support in MORSE.

Overview

Human support is an important feature of MORSE for HRI. Currently, MORSE only support a 'first-person' avatar, which is very complex (50 DoF, code with many dependencies on Blender internals,...).

We would like to eventually have:

• one 'accurate' human model for 'first-person shooter' style interaction, with simple and clean logic, and also support for up to 5 instances in the same simulation
• and simpler human models, not meant for direct mouse+keyboard interaction, but suitable for 'small-crowd' simulation (>20 characters)

  Autonomous navigation of humans in a scene

• develop a path object that would let other objects (like humans) to move along it.
• investigate detour | recast + new interface to define path to navigate

  Other comments:

• with a library of animations/behaviours (sitting, reaching, picking, waving hands...)
• Track multiples human with kinect input - Replay gestures from registered inputs (motion capture for exemple).
• Check the recent Character Physics type
• Current code for the human avatar need cleanup (in particular, remove the tight dependency on Blender internals)
• Simulation of sound ? -> Robot talks to the human

  Dependencies on other issues

• New human model: #655

And also: #116 (doc) #192 (unit-tests)

[severin-lemaignan]

Following the MORSE for HRI workshop (March 2014), this road-map has been updated as follow:

• instead of supporting 2 models for humans, we propose to have only one, with different levels of (mesh) accuracy. Those could be generated for MakeHuman, as suggested in #360. Issue #503 tracks this "multi-level" human model.
• we propose to completely remove the control logic for the so-called "FPS" interactive human avatar from MORSE, and move this controller outside of MORSE to a dedicated (pymorse) script. The human model in MORSE would only have regular armatures as sensors and actuators.
• this would remove a lot of Blender dependencies in MORSE code, and the external pymorse controller would provide a complete example of external control of the human avatar, that could be used as a base for alternative control modalities (Kinect, motion capture, Occulus Rift,...)

[severin-lemaignan]

Fixed by #682

[severin-lemaignan]

(re-opening until the PR #682 is merged)