This game has great potential, here is my wishlist for Crescent Loom / 2 if you intend on working on it. Note: I am not a neurologist, but I am a person who likes designing things - in particular I have 800 hours in From the Depths which is a vehicular building game.
- Optimisation. Half of the game's selling point is being able to program your creature using neurons, however trying to do more than animate your ragdoll creature (i.e. collision avoidance / hunting) severely bogs down performance for seemingly no good reason, even with performance boosts from running the game natively. I read that Crescent Loom 2 would be made in a tailored engine, probably to handle this simulation better, but a performance update would be nice. - - I assume that every neuron / axon's signal is being updated at frame time, which would be extremely bad for performance. Compiling signal strengths as functions over time, and using calculus to calculate when conditions are met (e.g. delay neuron) would greatly reduce iterations. - - I feel like the signal graphs of every neuron / axon updates even when I'm not looking (haven't tested fully), which would be a needless calculation. - - Every neuron's state change gets rendered on the brain even when you zoom out, LoD could be implemented here. - - If there's troubleshooting steps to be done on my end to improve performance, I haven't done anything so I would love to know. The game also just seems to be... running poorly, without having a high CPU load.
- Physics joints / fixtures. - - I feel that this is easier described with an example. I wanted to make a creature with a jaw that could open and close. I was able to animate it with muscles just fine, however both jaws could independently swivel left and right. If you imagine how gears rotate in opposite directions, that is the kind of stiffness I would prefer for opening and closing jaws - - All parts will always have a single point of failure because every part is extruded from the brain, being able to brace together parts (thinking Besiege) would be excellent. This mostly relates to using bones / structural parts as armour.
- Proper combat. Jaws instantly kill parts belonging to other creatures, including other jaws, which makes predator vs predator confrontations disappointing. It would be cool to see either physics based destruction (using f=ma for damage and perhaps using cross section to differentiate between part severing / blunt damage?). Either that or standardised parts that deal fixed damage would be fine too (thinking Spore / Sipho).
- Improved axon controls. I wish you could select an axon (edge) and edit its vertices. The controls are fine for really simple brains since you can simply paint the path, but the moment axons cross you need to rely on the pathfinding which doesn't always give you a nice tesselation. Do axons even need to take up space?
- Sandbox tools. Described in a comment below, but some sort of level editor or at the bare minimum the ability to load in creatures onto the cursor would make it easier to slam monsters against each other.
Footnote: I love Lancer Tactics. I haven't been able to get a party to play Lancer with, but already having limited build customisation is pretty awesome, and now I'm fixated on Lancer content :)
Thanks for the notes :) I had the opportunity to play CL at a party this weekend, and it was the first time in years that I've really sat down with it. I agree with a lot of your points — its difficulty running even a moderately complicated creature smoothing is a major issue. I did as much rendering optimization as I could back when I was developing it... unfortunately I think it's at the limits of its technology; the physics sim is the biggest bottleneck iirc.
Axons take up tiles because 1) they need to be selectable so you can inspect them (important for a lot of classroom applications) and 2) action potentials used to have to travel down the length of them and they were all different electrical compartments. The latter was removed to make creatures more responsive, but the engine still relies on them being that sort of entity.
Crescent Loom 2 is still a number of years off (gotta finish LT), but yeah more robust combat/opportunities for creature interaction and a map editor are going to be very high priorities for it. I really can't wait to get back to this. There's so many rough edges that a better underlying engine would solve.
I honestly think that axons should atleast be soft-separated from the grid system, as in it'd be a lot easier to interact with dense axons if they only became highlighted upon selecting a connected neuron. If you look at my creature, the brain code is very messy with a lot of overlapping axons. I ran into an issue where if I arranged the axons in a particular order (if an axon had to cross another axon at its synapse point) then I wouldn't be able to adjust the synapse strength of the axon. This shouldn't be a system where order matters.
Falling back on using functions and calculus again, you could probably use a function which describes the position of a signal along a given length of axon (linear interpolation my beloved) and calculate the time it takes for the signal to reach the next neuron (which would create support for either grid / gridless axons). You could then inspect along the length of said axon and it would basically render a 'slice' of this singular at a given point in the graph.
Most of the frametime stuff should just be rendering signals if the player happens to be looking in the general direction and zoomed in enough.
Once again: I'm not a neuroscientist and I'm guessing its not as easy as it sounds. I don't know any of the math involved in neurology so I would understand if simulation accuracy couldn't be maintained with simplified calculations.
Stugan is from 2018 and has since closed. I warned you to stop spamming these threads. I'm sorry but since you can't seem to control yourself, I'm going to have to ban you.
← Return to game
Comments
Log in with itch.io to leave a comment.
This game has great potential, here is my wishlist for Crescent Loom / 2 if you intend on working on it. Note: I am not a neurologist, but I am a person who likes designing things - in particular I have 800 hours in From the Depths which is a vehicular building game.
- Optimisation. Half of the game's selling point is being able to program your creature using neurons, however trying to do more than animate your ragdoll creature (i.e. collision avoidance / hunting) severely bogs down performance for seemingly no good reason, even with performance boosts from running the game natively. I read that Crescent Loom 2 would be made in a tailored engine, probably to handle this simulation better, but a performance update would be nice.
- - I assume that every neuron / axon's signal is being updated at frame time, which would be extremely bad for performance. Compiling signal strengths as functions over time, and using calculus to calculate when conditions are met (e.g. delay neuron) would greatly reduce iterations.
- - I feel like the signal graphs of every neuron / axon updates even when I'm not looking (haven't tested fully), which would be a needless calculation.
- - Every neuron's state change gets rendered on the brain even when you zoom out, LoD could be implemented here.
- - If there's troubleshooting steps to be done on my end to improve performance, I haven't done anything so I would love to know. The game also just seems to be... running poorly, without having a high CPU load.
- Physics joints / fixtures.
- - I feel that this is easier described with an example. I wanted to make a creature with a jaw that could open and close. I was able to animate it with muscles just fine, however both jaws could independently swivel left and right. If you imagine how gears rotate in opposite directions, that is the kind of stiffness I would prefer for opening and closing jaws
- - All parts will always have a single point of failure because every part is extruded from the brain, being able to brace together parts (thinking Besiege) would be excellent. This mostly relates to using bones / structural parts as armour.
- Proper combat. Jaws instantly kill parts belonging to other creatures, including other jaws, which makes predator vs predator confrontations disappointing. It would be cool to see either physics based destruction (using f=ma for damage and perhaps using cross section to differentiate between part severing / blunt damage?). Either that or standardised parts that deal fixed damage would be fine too (thinking Spore / Sipho).
- Improved axon controls. I wish you could select an axon (edge) and edit its vertices. The controls are fine for really simple brains since you can simply paint the path, but the moment axons cross you need to rely on the pathfinding which doesn't always give you a nice tesselation. Do axons even need to take up space?
- Sandbox tools. Described in a comment below, but some sort of level editor or at the bare minimum the ability to load in creatures onto the cursor would make it easier to slam monsters against each other.
Footnote: I love Lancer Tactics. I haven't been able to get a party to play Lancer with, but already having limited build customisation is pretty awesome, and now I'm fixated on Lancer content :)
La creatur :)
I will be saving this to the server as wm-predator1
Thanks for the notes :) I had the opportunity to play CL at a party this weekend, and it was the first time in years that I've really sat down with it. I agree with a lot of your points — its difficulty running even a moderately complicated creature smoothing is a major issue. I did as much rendering optimization as I could back when I was developing it... unfortunately I think it's at the limits of its technology; the physics sim is the biggest bottleneck iirc.
Axons take up tiles because 1) they need to be selectable so you can inspect them (important for a lot of classroom applications) and 2) action potentials used to have to travel down the length of them and they were all different electrical compartments. The latter was removed to make creatures more responsive, but the engine still relies on them being that sort of entity.
Crescent Loom 2 is still a number of years off (gotta finish LT), but yeah more robust combat/opportunities for creature interaction and a map editor are going to be very high priorities for it. I really can't wait to get back to this. There's so many rough edges that a better underlying engine would solve.
I honestly think that axons should atleast be soft-separated from the grid system, as in it'd be a lot easier to interact with dense axons if they only became highlighted upon selecting a connected neuron. If you look at my creature, the brain code is very messy with a lot of overlapping axons. I ran into an issue where if I arranged the axons in a particular order (if an axon had to cross another axon at its synapse point) then I wouldn't be able to adjust the synapse strength of the axon. This shouldn't be a system where order matters.
Falling back on using functions and calculus again, you could probably use a function which describes the position of a signal along a given length of axon (linear interpolation my beloved) and calculate the time it takes for the signal to reach the next neuron (which would create support for either grid / gridless axons). You could then inspect along the length of said axon and it would basically render a 'slice' of this singular at a given point in the graph.
Most of the frametime stuff should just be rendering signals if the player happens to be looking in the general direction and zoomed in enough.
Once again: I'm not a neuroscientist and I'm guessing its not as easy as it sounds. I don't know any of the math involved in neurology so I would understand if simulation accuracy couldn't be maintained with simplified calculations.
Stugan is from 2018 and has since closed. I warned you to stop spamming these threads. I'm sorry but since you can't seem to control yourself, I'm going to have to ban you.