Robotin Network Docs
1. Overview
1.1 Introduction
Over the past decade, artificial intelligence has profoundly transformed our world. From revolutionizing search engines and translation services to powering sophisticated recommendation algorithms and facial recognition systems, AI has evolved from a niche academic pursuit into an indispensable part of our daily lives. This era has been defined by the rise of "brain-like" AI—systems that can process information, understand language, and recognize patterns within the digital realm. However, this is only the first chapter of the AI story.
The next great leap in intelligence will not just be about thinking, but about acting and interacting in the physical world. This is the dawn of Embodied Intelligence. Technologies such as autonomous vehicles, advanced robotics, and immersive virtual reality systems are no longer the stuff of science fiction. They are becoming a reality, and their development marks a fundamental shift from AI that simply processes information to AI that navigates, perceives, and operates within our three-dimensional world.
Yet, a critical and largely unaddressed challenge lies at the heart of this revolution: a severe and fundamental scarcity of high-quality, real-world, Embodied AI data. The AI models that have dominated the digital space—like Large Language Models (LLMs)—have thrived on the vast, publicly available corpora of text and images scraped from the internet. In contrast, the data required for embodied AI is far more difficult to acquire. It demands an unprecedented volume of physical-world training data, captured and contributed by thousands of decentralized devices.
This data scarcity is the single greatest bottleneck holding back the embodied AI revolution. Consider the field of robotics. Vision-Language-Action (VLA) models, which enable robots to connect what they see with what they understand and how they act, are the future of the industry. These models require massive, multimodal datasets to train. However, as many robotics experts have highlighted, collecting this data is currently a costly, complex, and inefficient process, often requiring specialized and expensive setups—like a company's own warehouse filled with similar materials with different appearances to train a robot on a single task. The datasets that do exist are fragmented, small-scale, and cannot be easily scaled to meet the demands of real-world environments. Frankly speaking, without the piece of the puzzle of large-scale embodied intelligence data, humans are still a long way from truly usable next-generation physical AI.
This problem is not unique to robotics. It spans the entire spectrum of physical AI. Whether it’s a humanoid robot performing complex tasks in a factory, or a VR system seamlessly reconstructing a physical space, each application relies on the same core infrastructure: a foundation of rich, accurate, real-world data. The shortage of this critical resource is impeding the training of end-to-end models, slowing down innovation, and delaying the widespread adoption of technologies that will one day redefine our relationship with the world.
Therefore, the core mission is clear: to build the foundational data infrastructure that will empower this next era of embodied intelligence. We must move beyond the limitations of centralized, costly, and inefficient data collection methods and create a new paradigm—one that allows for the scalable acquisition of the high-quality, real-world data needed to train the machines of tomorrow. This is not just a technological challenge; it is the cornerstone upon which the entire future of physical AI will be built.
1.2 Robotin’s Vision and Technological Philosophy
Robotin aims at building a decentralized embodied intelligence data platform, collecting crucial data for embodied intelligence training. We believe that data harvested from devices scattered across ordinary households, rather than the vast data collection factories of the tech giants, is the true foundation for the future of embodied intelligence development.
1.2.1 Embodied Intelligence and VLA
The future of intelligence lies not only in seeing and speaking but in doing and solving within the physical world. This is the era of Embodied Intelligence, a paradigm shift that requires a new type of data. Unlike conventional computer vision datasets focused on static classification, embodied AI systems must learn intricate perception-action loops to navigate, understand, and interact with dynamic environments. This demands an unprecedented volume of multimodal, high-fidelity data that captures real-world interactions. However, this data is extremely limited, fragmented, and expensive to collect, making its scarcity the biggest bottleneck for the evolution of Vision-Language-Action (VLA) models in domains like household robotics.
The Robotin Network is a DePIN project designed to tackle this data problem head-on. We are building a community-driven, decentralized network for collecting and generating high-quality embodied AI data. Our project establishes a data pipeline specifically for indoor household domains, enabling the training of VLA models that can generalize across complex navigation and manipulation tasks. By creating a scalable and cost-effective data infrastructure, we empower the development of robust and general-purpose robots for the real world.
1.2.2 Open and Collaborative Ecosystem
The Robotin ecosystem is highly transparent and open to everyone. When the ecosystem reaches maturity, anyone can design, build, or customize hardware devices using our open-source framework. Participants are free to join and leave the network as they wish, and they can help collect, store, and process data without needing special permission.
Community and Governance
Our strength comes from our diverse community, which includes data providers, developers, hardware makers, and more. Together, they drive innovation and ensure the ecosystem continues to grow.
While we are a decentralized network, Robotin foundationoversees the project. The Foundation manages software and hardware development, ensures data quality, and guides network operations. We use Robotin Governance Proposals (RGPs) to let the community participate in decision-making, ensuring a transparent and fair process.
Privacy and Data Security
Privacy and security are at the heart of the Robotin project. We protect user privacy by ensuring all data is anonymized. For example, our device and system can precisely identify human facial features and personal information, then automatically blurs them. This allows us to create valuable data while fully protecting user privacy.
1.3 Community Contributor Participation and Network Value
The Robotin network makes it easy for anyone to contribute, creating an open and efficient ecosystem. Contributors can use Device A to collect data and participate in the community.
Our first device – MineLens is a highly integrated, compact sensor system designed for household indoor data collection. It includes a High-resolution RGB camera, data processing module and a Web3 identity and data validation module. It functions as a mobile data platform by attaching to a home cleaning robot, such as a vacuum. When the robot begins its cleaning and navigation cycle, MineLens captures continuous video streams of the floor-level environment. This valuable data is then uploaded to the Robotin network, where contributors earn $RTIN tokens as a reward.
A core principle of our network is that only data from a moving robot is considered valid. Submissions from a static device will not be accepted. This ensures the data is always dynamic and accurately represents real-world household environments.
In the future, we will provide more devices such as robot arm modules utilized for Embodied AI data collection, enabling higher data network value and contribution rewards.
1.4 Data Processing, Storage, and Commercialization
ROBOTIN INC serves as the primary entity for transforming raw data into valuable commercial products. It is responsible for the critical steps of cleaning, structuring, and converting raw data into Household Indoor Modeling or VLA Training Dataset and other premium data products for distribution to clients.
A key part of this process is the use of our intelligent backend system, which automatically filters out redundant and low-quality data. This ensures that only high-quality, valuable information is used to create our final data products.
To promote a more open and scalable ecosystem, Robotin will gradually decentralize its data processing capabilities. In the near future, qualified third parties will be empowered to access Robotin's open data, perform their own secondary processing, and commercialize the resulting products.
1.5 Project Governance and Core Responsibilities
The Robotin project is a collaborative effort led by two key entities: Robotin foundationand ROBOTIN INC This structure ensures both the decentralized integrity of the network and the effective commercialization of its data products.
Robotin foundation
The Foundation's primary role is to maintain the health and sustainability of the entire ecosystem. Its responsibilities include:
Ecosystem and Tokenomics Management: Ensuring the project's economic model remains stable and fair.
Open-Source Development: Providing open-source hardware designs and software frameworks for the community.
Infrastructure Oversight: Managing the global, decentralized data collection and storage network.
Technical Leadership: Guiding the project's technical roadmap and upholding core development principles.
ROBOTIN INC
ROBOTIN INC is focused on converting the raw data into valuable, market-ready products. Its main tasks are:
Data Productization: Transforming the collected data into commercial products, for instance, Household Indoor Modeling or VLA Training Dataset, etc.
Client Delivery: Supplying these data products to commercial customers.
Over time, this data productization and monetization role will be opened up to other parties within the Robotin ecosystem. This shift will foster a more decentralized and scalable model, empowering the community to contribute to the project's growth and value.
1.6 Data Privacy, Security, and Compliance
Robotin is committed to a global data ecosystem that is both innovative and fully compliant with international regulations. Our top priority is protecting user data and respecting local laws concerning privacy, national security, and fair competition.
Due to strict regulations in certain regions, our hardware and network are not permitted for use in the following countries: Mainland China, North Korea, Cuba, Iran, Venezuela, Russia, Belarus, Syria, and Ukraine. Any data originating from these locations will be rejected, and the corresponding contributors will not be eligible for rewards.
A Focus on Data Localization and Security
Our commitment to legal compliance extends to how we manage and store data. We follow these core principles:
Local Data Storage: We prioritize storing data on servers within the same country or region where it was generated. This approach minimizes the legal complexities associated with sending data across borders.
Secure Global Transfers: When cross-border data transfer is necessary, we use advanced encryption and security protocols to guarantee the data's integrity and confidentiality during transit.
Protection of Sensitive Information: For any sensitive data, we employ strict anonymization and de-identification methods to protect personal and proprietary details in compliance with local regulations.
Collaborative Compliance: We work with partners worldwide to build a secure and compliant system for data storage and processing that meets the diverse legal standards of each region.
We rely on our community members to play an active role in this effort. All contributors are responsible for understanding and adhering to the data collection and usage laws in their local jurisdiction to ensure a smooth and risk-free experience for everyone.
By balancing cutting-edge innovation with a strong commitment to legal and ethical standards, Robotin aims to build a sustainable future for Physical AI and Embodied Intelligence.
2. Data Production and Hardware
MineLens
MineLens is a compact, high-resolution sensor system designed for household indoor data collection, enabling accurate 3D scene reconstruction.
Key Functions and Workflow:
Data Capture:
MineLens attaches to a household cleaning robot and captures continuous floor-level video during the robot’s navigation. The high-resolution RGB camera ensures precise indoor modeling and scene reconstruction.
Data Authentication and Security:
Each device includes a secure encryption chip holding a private key. Video data is digitally signed at the hardware level, and the Robotin backend verifies signatures to prevent tampering or fraudulent submissions.
MineLens supports a wide range of applications, including embodied AI and robot training, long-term 3D modeling for smart homes, change detection and security monitoring, optimization of cleaning and service robots, and household behavior analysis. Future upgrades may incorporate additional sensors, variable camera heights, and enhanced data collection capabilities to enrich the network’s data diversity and value.
3. Time-Series Household Indoor Data and Embodied AI Data-related Technologies
3.1 Time-Series Household Indoor Data
3.1.1 Background
Most existing indoor datasets (e.g., Matterport3D, ScanNet, Replica) only provide static, one-time scans of household environments. However, in the real world, robots and intelligent systems operate in dynamic households, where objects shift, lighting changes, and obstacles appear or disappear every day. Currently, academia and industry lack long-term, continuous, and realistic household datasets that capture these temporal dynamics—limiting progress in embodied AI, robotics, and smart home research.
Key Problem: Static datasets fail to reflect dynamic, real-world household changes.
3.1.2 Core Features of the Dataset
Low-level perspective (10–40 cm): Relevant for service and cleaning robots.
Temporal continuity: Daily updates capturing natural variations.
Dynamic changes: Subtle differences in object placement, lighting, and furniture.
Realistic household settings: Authentic, non-synthetic environments.
3.1.3 Potential Applications
Embodied AI & Robot Training
Supports incremental SLAM and dynamic environment navigation.
Improves robots’ robustness to environmental changes.
Enables cross-day task planning.
Long-term 3D Modeling & Smart Homes
Builds time-evolving NeRF/3D reconstructions.
Analyzes furniture usage and movement patterns.
Supports health monitoring and layout optimization.
Change Detection & Security
Detects new or missing objects automatically.
Supports elderly safety monitoring and anomaly detection.
Cleaning & Service Robot Optimization
Identifies high-frequency obstruction zones.
Enables predictive obstacle avoidance.
Behavior & Habit Modeling
Infers household activity from object/furniture dynamics.
Enables personalized smart home services.
3.1.4 Market & Research Value
Academic value: Fills the gap of time-series indoor datasets, rare in today’s research.
Industrial value:Training material for robots, cleaning devices, and smart homes.
Long-term potential: Scalable to multi-household, multi-region, multi-season deployments.
Vision:Building a global embodied intelligence data asset through time-series
household datasets.
This dataset transforms household environments from static snapshots into continuous environmental stories. It accelerates research breakthroughs while delivering practical training data for robotics and smart home industries, paving the way for robust embodied intelligence.
3.2 Embodied AI Data and Training
The core of our project lies in the integration of embodied intelligence, enabling robots to perceive, reason, and act within real-world household environments. Unlike traditional AI systems that focus solely on static tasks, embodied intelligence demands dynamic interaction with environments, integrating visual perception, physical action, and natural language understanding. By leveraging a Vision-Language-Action (VLA) framework, we aim to enhance robots' abilities to execute complex manipulation tasks in indoor settings, relying heavily on data from continuous visual and action streams.
Our data collection pipeline captures continuous video and image sequences from household environments, where robots navigate and manipulate objects. This visual data includes not only static elements like room layouts and furniture but also dynamic factors such as object placements, occlusions, and lighting changes. The goal is to create rich, context-aware models that enable robots to navigate these environments effectively, even as they change. By collecting high-quality data from real-world environments, we aim to train models that understand the spatial relationships between objects, enabling robots to reason about how to move within these spaces and adapt to the ongoing changes within them.
In parallel, we gather robot manipulation data through wrist-mounted cameras on robot arms, which capture the robot's interaction with objects. The data collected includes visual streams, end-effector dynamics, and action sequences—such as grasping, moving, and placing objects—across a wide range of tasks. This combination of high-precision tracking and video data allows us to train models capable of executing precise manipulation actions in real-world environments. Through this, the system learns to perform tasks such as object sorting, moving items between locations, or interacting with novel objects, all while adjusting for environmental variables like clutter and lighting.
Training the models involves combining these visual and action data with natural language instructions. The integration of pretrained large language models (LLMs) and vision-language models (VLMs) enables robots to interpret open-ended, human-given commands and translate them into actionable tasks. This process makes use of multimodal fusion, where the robot learns to combine information from visual inputs and semantic instructions, mapping them to physical actions. The system understands natural language commands like “move the red vase from the kitchen to the shelf in the living room,” enabling it to navigate, interact, and manipulate objects across various household contexts.
Our end-to-end training pipeline is designed to process and merge multimodal data, allowing robots to not only navigate unfamiliar environments but also to perform precise manipulation tasks, such as placing or adjusting objects in response to spoken commands. By grounding the language in visual perception and action prediction, robots learn how to handle open-vocabulary tasks, meaning they can adapt to previously unseen objects without requiring specific training for each new scenario. This capacity for zero-shot generalization ensures that the system can perform tasks across diverse environments and respond flexibly to user needs.
The underlying model relies on expert demonstrations for imitation learning, where robots observe and replicate human actions. This method enables the system to learn generalizable policies that can later be applied to novel tasks or environments. Over time, the system evolves through continual data collection, improving its ability to understand complex household environments and complete tasks with increasing proficiency.
In summary, this approach combines natural language processing, visual perception, and robotic manipulation in a unified framework, allowing for intelligent interaction in dynamic, real-world environments. The system is capable of handling tasks with varying complexity, from simple object movements to more intricate organizational tasks, all while being adaptable to changes in the environment and open to new, unseen scenarios.
3.3 Data Usability Validation
The Embodied AI data we generated has completed usability validation. This is an demo arm trained by our data. It can automatically sense debris on the ground and store it in the designated location.
4. Combating Fraud on a Trustless Network
In decentralized networks, cheating and data manipulation are significant risks. Robotin addresses this with a trustless system that verifies data authenticity without relying on a centralized authority. This approach, built on our experience, makes a simple blacklist unnecessary. The core principle is that only verified data is accepted by the network.
4.1 Secure Digital Signature System
Every Robotin device includes a dedicated encryption chip that securely holds a unique private key. This key is not accessible from the outside, ensuring it cannot be stolen or copied.
The data signing process is straightforward:
When a device generates data, it uses the private key in its encryption chip to create a digital signature.
The data package sent to the network contains three key components:
The collected data itself.
The digital signature created by the encryption chip.
The public key corresponding to the device’s private key.
The Robotin backend system then uses the provided public key to verify the digital signature. If the signature is valid, the data is accepted as authentic. Any data that fails this verification is automatically rejected, preventing forged or tampered information from corrupting the network.
4.2 Data Content and Quality Analysis
Beyond behavioral checks, we will analyze the intrinsic properties of the video itself to detect anomalies indicative of fraud.
Metadata and Timestamp Validation: Each uploaded data packet will be verified against a server-side timestamp. Any data with manipulated or out-of-sync metadata will be automatically discarded. This ensures that the submitted data is not pre-recorded or tampered with before upload.
4.3 Behavioral and Environmental Verification
We will validate the data by cross-referencing the video content with the expected physical behavior of a robotic vacuum. This includes:
Movement Pattern Analysis: The system will analyze the camera's trajectory to confirm it aligns with typical robotic vacuum cleaning paths, such as "S-shaped" or wall-following patterns. Any long-time unnatural movements, such as abrupt jumps, rotations in place, or a static viewpoint, will be flagged as suspicious.
Environmental Consistency Checks: We'll analyze consecutive video frames for consistent environmental cues. A genuine stream will show subtle, natural variations in lighting, shadow, and object placement over time. Data that shows a completely static scene for an extended period or features abrupt, unrealistic changes in lighting will be considered fraudulent.
Sensor Cross-Validation: The system will correlate visual data with other sensor readings from the device, such as those from an Inertial Measurement Unit (IMU). The camera's detected motion and rotational changes must align with the corresponding sensor data.
4.4 Stricter Penalties for Cheating
To maintain fairness in the network, for verified cheating, Robotin enforces strict penalties:
1st Offense
Account and device frozen for 3 months
2nd Offense
Account and device frozen for 999 years (permanent ban)
Robotin takes cheating very seriously and is committed to preserving a fair and secure network environment.
5. $RTIN Allocation
Total fixed supply:10 billion
Allocation Structure:
40% as rewards for contributors, to participate in building the Robotin Network
20% for project investors
15% for project founding team members, and future global core contributors for project R&D and system construction
15% allocated as a long-term reserve managed by the Foundation to ensure the sustainability, resilience, and strategic flexibility of the Robotin Network.
10% for Robotin Network ecosystem including allocation includes liquidity, market operations, promotions, and other aspects essential for sustaining ecosystem development.
Category
% of Supply
Token Amount
Community Contributors
40.00%
4,000,000,000
Investors
20.00%
2,000,000,000
Team
15.00%
1,500,000,000
Foundation Treasury
15.00%
1,500,000,000
Ecosystem
10.00%
1,000,000,000
Allocation Structure:
40% Mining Rewards:Released progressively based on users’ data duration and data quality.
20% Investors:6-month cliff, followed by 24-month linear vesting.
15% Team Allocation:9-month cliff, followed by 36-month linear vesting.
15% Foundation Treasury:48-month linear vesting.
10% Ecosystem:0.02% unlocked at TGE; remaining tokens vested linearly over 12 months.
5.1 $RTIN Calculation
5.1.1 Base Rewards
MineLens: 32 $RTIN per hour
Starting from TGE, the base rewards for $RTIN will be halved every year
1
32 $RTIN
2
16 $RTIN
3
8 $RTIN
4
······
5.1.2 Data Quality Tiers & Reward Ratio
The quality of the collected data (e.g., clarity, brightness, frame redundancy, etc.) will impact the amount of $RTIN received. The final assessment will be categorized into quartiles:
A - Excellent
100%
B - Good
75%
C - Average
50%
D - Below Average
25%
F - Failing
0
*”F” will be assigned to any user using multiple MineLens at the same spot.
5.1.3 Daily Data Redundancy Rewards Decay
The whole network will see a 50% reward reduction for every 4000 hours data collections.
0 ~ 4,000 hours
100%
4,000 ~ 8,000 hours
50%
8,000 ~ 12,000 hours
25%
12,000 ~ 16,000 hours
12.5%
······
······
Data Redundancy are calculated globally.
Data Redundancy Records Reset Daily: the records will be renewed at 00:00 GMT.
$RTIN will be transferred within 1 week after data upload, data older than 2 weeks won't be accepted.
The previous week's rewards will be calculated by Tuesday and displayed in the app. These rewards will be distributed evenly across users from Monday to Friday of the current week.
*Tip: It is recommended to upload data promptly to maximize your rewards.
*Rewards Multiplier based on Customer Requirement
The additional $RTIN reward multiplier will be set based on the requirement of future data customer, using specific areas and tasks as criteria.
5.2 $RBTN Burn Mechanism
$RTIN operates on a permanent burn mechanism, with buyback funds sourced from data consumers purchasing Robotin Network's data products:
80% is used to buy back circulating $RTIN and permanently burn it.
20% supports Robotin Network's operational expenses.
6. Disclaimers
Living Document Disclaimer
The technologies and concepts detailed in this whitepaper reflect Robotin's vision and best efforts at the time of publication. However, the world of decentralized physical infrastructure and artificial intelligence is evolving at a rapid pace. As we encounter new challenges and make new discoveries, our design decisions may be revised to accommodate these changes. This document is a living framework, intended to inform and empower the community to participate in the development of the Robotin Network.
Legal Disclaimer Regarding Token Associated Risk Factors
This whitepaper is intended for informational purposes only. It does not constitute a prospectus, an offering of securities, or a solicitation for investment in any jurisdiction.
Prior to acquiring any tokens, each participant must carefully evaluate all information and risks detailed herein and in any related materials.
Robotin makes no guarantees regarding the future value or performance of its tokens. As with any digital asset, token prices are subject to significant volatility and may fluctuate dramatically. We make no promises of future performance, inherent value, or any specific value retention. Robotin Network tokens do not represent any equity interest, ownership, or rights in any business or assets.
It is important to be aware of the inherent risks of blockchain technology, especially as the issuance and holding of tokens may be new concepts in some jurisdictions. The token distribution and network operations could be subject to regulatory actions, including potential restrictions on ownership, use, or possession. Regulatory or other competent authorities may require us to modify the functionality of our tokens to comply with new legal requirements.
No regulatory authority has reviewed or approved this whitepaper. The publication of this document does not imply compliance with or the applicability of any laws or regulatory requirements.
This whitepaper may reference third-party data and industry publications. While we believe this information is accurate and our estimations are reasonable, we provide no assurances as to the accuracy or completeness of this data. Information and data from third-party sources have not been independently verified, nor have we confirmed the underlying assumptions relied upon by these sources.
This whitepaper is not intended for distribution to, or use by, any person or entity in a jurisdiction where such activities would be contrary to law or regulation.
Robotin shall not be held liable for any losses or damages (whether direct, indirect, consequential, or otherwise) that arise from the use of, reference to, or reliance on the contents of this whitepaper.
Forward-looking Statements
This document contains forward-looking statements that are based on our current expectations of future events. These statements are projections about future events and business trends that we believe are relevant to the development and success of the Robotin Network.
Forward-looking statements are based on assumptions and analyses made by the Robotin team in light of our experience and perception of historical trends, current conditions, and expected future developments. These assumptions are subject to risks, uncertainties, and other factors that could cause our actual results, performance, or achievements to differ materially from the expectations expressed or implied. Given these risks, prospective participants should not place undue reliance on these forward-looking statements.
Last updated