Mohsen Raoufi

✨ What is LARS?

LARS (Light Augmented Reality System) is a cross-platform, open-source framework for experimentation, education, and outreach in collective robotics.
It leverages Extended Reality (XR) to seamlessly merge the physical and virtual worlds, projecting dynamic visual objects, such as gradients, fields, trails, and even robot states, directly into the environment where real robots operate.

LARS enables indirect robot-robot communication (stigmergy), while preserving all real-world constraints. It turns “invisible” collective dynamics into tangible, interactive experiences for researchers, students, and the public.

Poster_ANTS.pdf

🛠️ Key Features

Projection of Virtual, Visual Objects 🔦
Marker-Free, Cross-platform Detection and Tracking System (based on ARK) 🎯
Real-Time Performance ⚡
Standalone System 🖥️
Scalability to Collective Size 📈
Indirect Robot-Robot (Stigmergy) 🐜 and Human-Robot Communication
Direct Communication Possibility 📡
Ease of Setup and Robustness 🛠️
Logging for Post-Production 📝
Cost Effective 💶
Open Source 🔓

🎯 Why LARS? (Objectives)

For Research:
- Supporting reliability, reproducibility, and flexibility in collective robotics experiments
- Improving human-robot interaction
- Enriching environments with virtual objects—without sacrificing realism or robot constraints
For Education:
- Making abstract information observable to humans
- Promoting science communication and public engagement by showcasing embodied collective behaviors of robots

🏹 From ARK to LARS:
Advanced Multi-Robot Tracking & Visual Augmentation

LARS features a robust, real-time tracking module based on the ARK (Automatic Robot Kinematics) algorithm, but goes far beyond:

⚡ Dramatically improves speed and accuracy, supporting dense populations (100+ robots) at >35 FPS
🔎 Robustly handles occlusions, variable lighting, and noisy backgrounds
🔵 Generalizes to any robot that appears approximately circular from above—including Kilobots, Thymio, e-puck, and others—without the need for tags, markers, or hardware modifications
🏷️ Automatically preserves and recovers robot identities even during close interactions or when robots briefly leave the field of view

Beyond tracking:
LARS projects virtual visual objects (gradients, cues, signals) in real time—directly onto the arena and the robots themselves.
This enables:

Simulation of virtual environments without hardware changes
Dynamic, spatially precise feedback to individual robots or entire collectives
Exploration of new paradigms in human-swarm and robot-environment interaction

🏗️ Architecture Overview

LARS is built on the classic Model-View-Controller (MVC) pattern:

🧩 Model: World state, physics, and objects
🖼️ View: GUI, visualization, and projector output
🎮 Controller: Experiment orchestration, tracking, and logic

🧱 Hardware Setup Overview

🔦 Projector
A ceiling-mounted projector, redirected by an adjustable mirror, casts visual objects onto the arena surface, creating the augmented “world” the robots see.
👁️ Global Sensing for Detection and Tracking
An overhead camera, attached to the frame, tracks robot positions, closing the loop between robots, visuals, and control software.
🖥️ GUI and Overhead Controller
A side display and a controller arm provide the human interface to LARS, for starting experiments, adjusting parameters, and monitoring the system in real time.
🧩 Scaffolding
An aluminum frame and enclosed base hold all components in a stable, portable structure.

🧑‍🔬 Example Scenarios

🗳️ Collective Decision-Making: Track and visualize 100+ Kilobots in a noisy, projected environment
⏰ Swarm Synchronization: Record robot states and group dynamics in real time
🕹️ Interactive Demos: Let visitors steer/interact with swarms and see collective behavior
🧑‍🏫 Educational Labs: Manipulate real experiments to teach robotics, physics, and complexity

(top, left:) GUI snapshot of 42 Kilobots synchronizing on a grid with their internal binary state being detected by the color of their LED in blue or red,
(top, mid-left:) user view of 63 Kilobots making a collective decision on a tiled environment with projected dynamic noise
(top, mid-right:) GUI Snapshot of 109 Kilobots with the trace of their random movement decaying over time
(top, right:) GUI snapshot of two active balls randomly moving in the bounded arena, being tracked by LARS without the need for any markers
(bottom, left:) GUI snapshot of two Thymios with different colors locating the center of the light distribution (projected by LARS). The trace of each robot shows the consistency of the color detection of each robot over time, even after a collision
(bottom, right:) User view of Thymios moving randomly, with their centroid, the projection of their trajectory (light blue trails), their Voronoi tesselation (black lines) and the corresponding network (green lines).

Shaping LARS, One Step at a Time

Behind LARS is a long sequence of small decisions, experiments, fixes, and rethinking. These notes reflect the day-to-day work: ideas tried out, problems found, and tasks crossed off. This is the work behind LARS...

Page updated

Google Sites

Report abuse