Use cases

Summer city

The main idea of the game is to simulate a real life experience for the players with the use of virtual environments and artificial intelligence. An intelligent NPC will be fully autonomous by its own inborn reasoner based on life goals and beliefs. Basically, each NPC has to eat, sleep and work so that it can survive and reach to the goal.

Summer City Episode1

Screen is divided to 4 areas : 4 cameras to follow each NPC

Summer City Use case Demo Explained

  • Game Purpose: to provide a use-case of life simulation by applying fA (showing a behavior of AI NPC which is decided by its desire/life-goal).
  • Game system Environment: Unity3D with ‘fA for Unity’ on window 64bit
  • Player (character): goals: this demo is focused on behaviors of NPC, so there is no player character.
  • NPC: goals and beliefs: there are 4 NPCs living in Summer City.
  • Individual goals: each NPC has a goal to maximize HQ or IQ or Wealth or SQ(not yet implemented). Basically, each NPC has to eat, sleep and work so that it can survive and reach to the goal.
  • Individualbeliefs: HQ(how rasted), IQ(how intelligent), SQ(how social), hunger(how hungry), and money, positions(houses, furniture, boxes, corns, truck, buildings), amount of corn in a box, state; occupied, fulled, picking, holding or not.
  • Initial capabilities of NPCs:
  • Reasonercapability (such as problem solving/learning/memory)
  • Percept processing (to beliefs) / belief-update – they can learn change in the perceivable state of the environment and remember that.
  • Belief-inference (with MA-PDDL) – they can infer new knowledge (beliefs) based on beliefs updated from perception.
  • Action planning (with fA NeuralPlanner) – they can decide (based on the current beliefs) about what actions to attempt next to achieve their goals.
  • Physical capabilities: Move, Pick, Hold, Drop, Study, Exercise, Eat, Rest, Sleep, Talk
  • Precepts’ of NPCs: seeing objects(box, furniture and machine), seeing other NPCs(within a limited vision), self-perception(own state and etc.)
  • (if you adopt OOA)
  • Box – hold / drop
  • Corn – pick
  • Training machines – exercise
  • Bed – sleep
  • Chair – rest / eat
  • Desk – study
  • NPC – talk
  • Constraints and others if any
  • if HQ / hunger of NPC is zero then NPC can’t work, exercise, study…
  • if a box is held by other it can’t be held again.
  • if corn-place/training machine/desk or chair is occupied by other, NPC has to find other place.
  • the HQ can’t be higher than maxHQ.
  • the Hunger can’t be lower than zero.
  • the amount of corn in a box can’t be higher than 100.
  • To eat NPC spends 10.

Stealth Fortress Demo Explained

In this demo we have a simulated Stealth Fortress (or facility / factory), which is populated with 25 AI guards. Each guard is fully autonomous, having it’s own AI made with ‘fAutonomy for Unity’. There is a single player controlled character, PC (a secret agent), who wants to escape/exit the fortress without being captured by the guards.

The goal of the player (blue character above) is to sneak through the fortress unnoticed, or at least not captured by the guards & exit.

  • Purpose: to provide a use-case of interactive agent-simulation (demoing interactive life-simulation capability) by applying fA.
  • Game system Environment: latest release version of Unity (2017.3.1f1 (x64) or later) with ‘fAutonomy for Unity’
  • PC (Player Character) goals: the single player (character) is a secret agent in a fortress, and now wants to escape with the obtained secret. Accordingly, the goal of the PC is to reach the EXIT.
  • NPC (Non-Player Character) goals and beliefs: there are 25 simulated AI NPC guards the fortress.
  • Individual goals: patrol a distinct part of the fortress, stay rested, and if any intruder is spotted, then catch that intruder (not allow any intruder to exit).
  • Individual beliefs: HQ (how rested); last seen position of intruder (if any); topology of the warehouse (positions of doors, incl. exit, and keys); alert state; intruder caught / exited or not…
  • Initial capabilities of Player/PC/NPCs:
  • PC can run (quick&loud); sneak (slow&silent); turn; open doors; cross open door (incl. exit); hide in compartments; pick-up keycards; unlock locked doors (with proper keycards); throw a small rock to distract guards; stun guards (putting them to sleep for a limited time).
  • Player can change perspective from 3rd person to top-view. Player sees: inventory (of keycards); direction of exit; field-of-view and position of close guards (on mini-map); alert state (with green, yellow and red color) of guards; types of doors (openable / locked with given type of key); 3D environment.
  • Actions (physical capabilities) of NPCs: walk; run; turn; rest for a time; wake up stunned / resting guards; catch intruder. They have all keycards, so they can open any door, and any compartment. They can also communicate via radio to exchange any type of known information.
  • Percepts of NPCs: seeing doors (with open/closed state); hearing sounds (door opening, PC / guard walking / running (from different distance)); seeing PC and other guards (within a limited vision cone); hearing other guards radio communication (not limited by distance); self-perception (e.g. own state, resting, awake, alert level, etc.).
  • Reasoning capability of NPCs (such as problem solving/learning/memory) :
  • Percept processing (to beliefs) / belief-update – they can learn change in the perceivable state of the environment and remember that.
  • Belief-inference (with MA-PDDL) – they can infer new knowledge (beliefs) based on beliefs updated from perception.
  • Action planning (with fA NeuralPlanner) – they can decide (based on the current beliefs) about what actions to attempt next to achieve their goals.
  • Physical capabilities over objects (OOA-style):
  • Door: open / cross / unlock
  • Compartment: open / enter / exit
  • Key: pick-up
  • Guard: stun / wake-up / distract

Constraints: the environment imposes various constraints on the state and timing of the execution of actions (e.g. open door cannot be re-opened; stunned guard cannot be stunned again; the inventory of the PC is of limited capacity; etc.)

Login
Loading...
Sign Up

New membership are not allowed.

Loading...