Questions & Discussion

Questions

Why create Kinesis?

Kinesis was created mostly as a byproduct of attempting to create a base for further AI experiments in embodied cognition. I wanted an agent with a complete physical presence that could navigate and interact with a broader physical environment in an organic and “non-scripted” way. Although I didn’t think it would be a cakewalk, I simply misjudged the amount of work it would take to both implement and work with such a muscle-based system (at least, for someone like me without a graduate background in biodynamics, robotics, or mathematics). In hindsight, I could’ve picked a better approach for what I wanted to accomplish, but working on Kinesis was rewarding in its own right.

The reason I decided to clean up the implementation and release Kinesis on the Unity Asset Store was because I didn’t really find anything like it available for Unity, and it definitely would’ve been nice to have something like it when starting out. My hope is that fellow experimenters find it a useful starting point.

What does the muscle model in Kinesis ignore and what is it missing?

The muscle model in Kinesis is a somewhat simpler version of Hill-type muscle model. While it does handle segmented pathing, features that were not included in the initial implementation include:

• Modeling of muscle fiber pennation angles.
• Modeling of serial damping effects.
• Handling for more complex muscles such as those with multiple heads, insertions, and sheet-like muscles.

There are other phenomena that Hill-type muscles models don’t typically seem to address:

• Muscle mass, although there has been at least some exploration of this with Hill-type muscle models.
• The effects of previous activity (of various timescales) on muscle dynamics.

The model applies forces using the muscle segments as lines of action. Why use AddTorque() instead of AddForce()?

I did experiment with AddForce() in the beginning, but it quickly became apparent that the approach was inherently unstable, even though one would imagine they should thereotically generate the same results. Not only was using AddTorque() more stable, but it also simplified the code required for movement forces.