Motion · Learning · Wireless systems · A small pause

Playground

Eighteen small, dependency-free games live inside the site rather than taking it over. Compact learners, conceptual research models, and quick puzzles turn wireless systems, AI, and research routines into hands-on interactions.

Card sized
Bilingual
Theme aware
Static hosting

01

Research ideas you can play

Demonstrate a route, inspect a latent forecast, or interpret a temporal monitor: each interaction makes a compact learning idea visible before introducing the terminology.

01

One-key runner · 20 sec–2 min

Reward Runner

Jump through positive rewards and clear penalty blocks as the episode accelerates. One click is the whole policy.

Click / Space
Live return
Dino-like

Open standalone

03

Human–Agent Q-learning · 20–60 sec

Routing Apprentice

Human and Agent explore fresh routing episodes as equals, and every transition updates the same compact Q-table. Let the Agent run once or keep learning until its next delivery.

19 decision states
Shared exploration
Terminal obstacles

Open standalone

05

Conceptual world model · 30–90 sec

Latent Look-Ahead

Move a UAV relay north, south, east, or west, adjust its altitude, and keep drifting maritime users inside coverage. A compact action-conditioned model previews three latent steps before the next observation and learns online from every reveal.

Seven flight actions
Latent rollout
Service lock

Open standalone

02

Bandit lab · 15 actions

Explore–Exploit Lab

Probe three unknown actions, update their value estimates, and decide when exploration should give way to exploitation.

Three actions
Q estimates
Expected regret

Open standalone

04

Discounted return · 6 rounds

Return Route

Compare two reward sequences under a changing discount factor, choose the better route, and see why delayed rewards sometimes win.

Discount factor
Return calculation
Six choices

Open standalone

06

Temporal-logic monitor · 30–90 sec

Semantic Sentinel

Tune a temporal-logic alert threshold, read a short observation–action–outcome stream, and decide whether to hold, repair locally, or probe longer. Sensitive monitoring catches faults earlier but raises false alarms; tolerant monitoring risks missed faults.

Adjustable threshold
Semantic margin
False alarm vs miss

Open standalone

02

Wireless systems you can steer

Simplified conceptual models make coverage, interference, backscatter, resilient telemetry, reflection, and secure beam steering visible and playable.

07

Antenna geometry · 1–4 min

Movable Antenna Lab

Steer twelve sector antennas across four well-spaced base stations, inspect coverage and interference, then watch a search algorithm improve the layout.

12 steerable beams
Coverage map
Animated search

Open standalone

10

Human–Agent anti-jamming · 30 sec–3 min

Hopping Apprentice

Human and Agent alternate fresh interference episodes, with every channel choice updating the same 18-state policy. Ask the Agent to explore once or continue until it delivers at least 9 of 12 packets.

18-state table
Shared exploration
Run until success

Open standalone

12

Semantic-link recovery · 30–90 sec

Predict Through the Outage

Choose how much semantic telemetry to send through a fragile link, fall back to predicted state during outages, and spend one forced resynchronization at the right moment. The rescue guarantees a fresh state but carries an explicit control cost.

Cached prediction
One rescue token
Service gap

Open standalone

08

Waveguide placement · 1–4 min

Pinching Antenna Lab

Slide four radiating pinch points along a waveguide, flip their emitting side, and balance served users against overlapping beams.

Drag positions
Flip direction
Animated search

Open standalone

09

Secure beam steering · 30 sec–3 min

Secrecy Beam Lab

Steer Alice's beam across a full-room local-average-energy map. Opaque obstacles remove blocked line-of-sight components; optional walls use the same lossy specular law, with at most three reflections across walls and obstacles. Compare Bob's and Eve's achievable rates, then maximize instantaneous secrecy rate.

Instantaneous secrecy rate
Energy map
Three-bounce reflections

Open standalone

11

Adversarial backscatter · 1–3 min

Signal Judo

Manage a finite queue and battery across ten slots. Probe a shifting opponent, read the last four observed carrier outcomes, and adapt when its signal is a threat, an energy source, or a carrier worth backscattering.

Queue + battery
Observation trail
Adaptive rhythm

Open standalone

03

Research easter eggs

Two-click link-acquisition and idea-mapping interactions, designed to hide beside papers, project notes, or quiet corners of the homepage.

13

Link acquisition · 10–20 sec

Link Acquisition

Start an orbital pass, then lock the satellite inside the acquisition window. Two clicks complete one link attempt.

Two clicks
Acquisition timing
Range cues

Open standalone

14

Idea map · 5–30 sec

Idea Constellation

Place seven idea nodes and connect a brief pause into a small research constellation.

Click / Enter
Seven nodes
No score

Open standalone

04

Take a longer research break

Signal-sequence, visual-memory, paper-growing, and resource-network puzzles for a longer break, all kept inside embeddable cards.

15

Signal memory · 30 sec–3 min

Signal Replay

Watch four channels pulse, then replay a signal sequence that grows by one step each round.

Keys 1–4
Ten rounds
Slow cues

Open standalone

17

Research merge · 3–10 min

Paper Garden

Merge equal research samples on a 4×4 board and grow one small idea into a finished paper.

Arrow keys
Local best
Idea → paper

Open standalone

16

Visual memory · 1–4 min

Pattern Recall

Treat a 4×4 symbol grid as a tiny visual-memory trial: recover eight pairs and beat your best move count.

Eight pairs
Local best
Keyboard

Open standalone

18

Resource orchestration · 1–5 min

OpenRaaS Mesh

Rotate a compact OpenRaaS network until compute, storage, and image nodes all reach the coordinator, then watch requests flow.

Rotate tiles
Always solvable
OpenRaaS

Open standalone