Self-building Robot Civ Game Framework

[josh]

Future directions could develop into a way for potential clients/customers to solicit bids/solutions in concrete/specific formats.

[josh]

There will be lots of different kinds/types of “source” code involved. One kind of maneuver to prevent time wasting in anticipation of change that is to hard to predict: say “significant” sections of source will be labeled with a source type & version number (explicitly or as an auto IDE function); say that the compilation/interpretation process will always include a stage that allows for mechanical translations of outdated sections to the current level wherever that is unambiguous & viable. Do that & don’t sweat too much about interesting changes.

[josh]

Notes:
a) In my mind, there wasn’t any specifically planned connection between the kernel distribute OS for “sequential process hacking” (getting at CRM in ‘1984’ type environments & others) & software systems for simulation of robotic tasks in the civ game framework. Of course there are always possibilities to share libraries, styles, & innovations, but I didn’t mean to propose any tight sharing with those pair of initiatives that I wrote about in near days time. Why not the same task?? The CRM like initiative is emphasizing rapid advancement of the application plugin architecture for contacting, tracking, evaluation, conversational AI, negotiation etc – we’re trying to achieve rapid development & long life for those types of plugins in order to meet urgent organizational goals & to fit with other commercially popular products that only start to seem like tough metal experiments from a softare systems POV, when we are dealing with modifying the databases that the apps depend on while ongoing conversations are running. That’s not an overlap with the needs of the robot simulation which is dealing with new, computationally taxing models of specific chunks of reality and hardware/software systems running there. In the robot world, there is mroe emphasis on hypervisor levels & PDE analyses, levels of approximation,….

b) For robot planning of advancements in simulation, it will be valid to include plans for new capabilities that start out at abstract levels of detail. Those plans will become challenged to deliver adequately realistic implementations & sometimes fail.

c) Low level griding on simulations can turn out to be a garden path if the high level plan is unsatisfactory. In my own experience, I’ve spent years battling EMF attacks with concrete experiments & limited successes of earlier experiments often were entirely thrown away in light of later analysis – e.g. putting up a thick plastic contractor bag provides immediate shielding, but it turns out the bag is nanotech enabled with its own charge up, & starts to attack later once it has absorded enough energy => earlier rough simulations of medium range magnetism guns turned out to be not so relevant when the attacks were delivered by IOT on objects inside my house, with wireless charging/hopping between them based on nanotech/IOT, etc. My threat model got updated to include all of this nanotech everywhere – little screws can shoot EMP for a surprising range, but that’s not the only problem. Also didn’t realize the FBI had a giant world wide gang of civilian minions with long history of investment in both sewer & underground burrowing terrorism.

[josh]

Valuable future tech for robotics in general:

Go from verbal descriptions, with meaningful modifiers & adjectives, to random generators of many test case simulation environments that can be used to examine the robustness of a given robot design & deliver useful analysis freom lots of simulated runs.

[josh]

I’m sort of hearing a lot of enthusiasm for the general project of matching constructive digital engineering/design with gaming. In order to tap into even more of the same, it makes sense to add aesthetic criteria into some elements of some competitions. It’s most constructive if the bulk of that effort goes into elements that are late in the cycle of “consumer” or end user acceptance. There is room for submitting surveys at those points (double blind) & ML models that attempt to match the aesthetic judgments of different classes of consumers – calibrated against the data.

[josh]

On a practical level it’s interesting to apply this concept to increasing the automation of factories that already involve manufacturing machines. Simulation can be used to model replacement for various human stages in the manufacturing process using full automation. Practical experiments outside of the simulation are expensive, so extra tolerances for noise/imprecision/variability/events should be built in to the simulation. Economic analysis can also be added to help speed up the most practical proposals for implementation. From the POV of mfg., it’s also valuable to see which forms can be applied in many situations.

[josh]

My hope is that the overall context of GT accelerates development in many dimensions, and one of these can be the open sharing of ideational plans & concepts rather than preserving them for private research or consulting. In this context, for example, wise mentors could help younger teams by suggesting potential steps of increasingly accurate simulation/tech achievement/testing. There could be a wiki with alternative ideas about that & teams could see which looks like the best option for them or get ideas of their own. The community can help break big problems down into manageable, testable pieces.

[Author]

Posts