FrodoBots

FrodoBots is a robotic gaming platform that crowdsources real-world training data for embodied AI by letting people remotely control physical sidewalk robots from anywhere in the world. The company's thesis is direct: "Embodied AI will be solved by gamers." Rather than hiring professional teleoperation workers or constructing expensive simulation environments, FrodoBots turns the data collection problem into a consumer game — one that pays players for their time and produces high-quality, real-world movement data in the process.

Founded in February 2022, the project was publicly showcased at Solana Breakpoint 2024, where co-founder Michael Cho presented the Earth Rovers concept to the Solana developer community, signaling the project's intent to build its incentive and ownership layer on Solana.

Earth Rovers: Drive-to-Earn at Street Level

Earth Rovers is FrodoBots' flagship product and is currently in public beta. The concept is described as "like Mario Kart, but IRL" — a global scavenger hunt format where players connect a game controller, racing wheel, or keyboard to their home computer and take the wheel of a 4G-enabled sidewalk robot deployed somewhere in the world.

The robots are purpose-built for continuous outdoor operation. They are weatherproof for rain and snow, equipped with toggleable headlights for night driving, fitted with terrain-gripping tires, and carry a 4G SIM card for real-time control over the internet. Hardware is available for purchase directly, priced between 199 and 249 USD per unit, and hardware designs are released as open-source. Robots require a GPS signal and must remain outdoors more than 50 percent of the time to qualify for the platform.

At the time of its Breakpoint 2024 presentation, Earth Rovers had been deployed across more than ten cities including Taipei, Madrid, Vienna, Wuhan, Singapore, London, Stockholm, Berlin, and locations in California. That geographic spread is intentional: diverse urban terrain produces richer training data than any single city could provide.

How Earnings Work: FrodoBot Points

Players earn FrodoBot Points (FBPs) through a formula tied to real driving output: points are proportional to distance driven multiplied by a difficulty score. Difficulty scores range from 1 to 10 and are calculated based on terrain complexity and environmental factors — narrow sidewalks, obstacles, weather, and slope all factor in. A longer drive on a challenging route earns far more than a flat loop in an empty plaza. The design aligns player incentives with researcher needs: the most scientifically valuable data (complex real-world environments) is also the most rewarding to collect.

FBPs can be earned in several ways beyond direct driving. A friend-sharing system lets players create time-limited links to their robot feed with a customizable FBP split, letting others try the experience while the robot owner still earns. Operators who enroll in the school program can offer robot access to trainees and earn both FBPs and Time Credits (TCs), which are each valued at one US dollar. Robot owners can also partner with Earth Rover Pro Driver NFT holders — a category of credentialed professional drivers on the platform — who complete pre-defined missions on behalf of robot owners, with revenue shared through an NFT-gated arrangement.

The FrodoBots website states that a token-incentivized network rollout is planned to reward participants at scale, with the project building toward a broader community ownership model. The presentation at Solana Breakpoint 2024 and the NFT-based driver credentialing system both point to Solana as the intended settlement layer for this incentive network.

The FrodoBots-2K Dataset

The direct output of the Earth Rovers game is the FrodoBots-2K dataset, publicly hosted on Hugging Face. The dataset represents approximately 2,000 hours of teleoperated driving accumulated across more than 9,000 individual sessions in cities across multiple continents. Data types collected per session include control inputs from gamers at 10 Hz alongside wheel RPM readings; GPS location and timestamps at 1 Hz; 9-DOF IMU sensor readings covering acceleration at 100 Hz, gyroscope at 1 Hz, and magnetometer data; rear-facing video at 540x360 resolution and 20 frames per second; front-facing video at 1024x576 at 20 frames per second; and audio captured from onboard microphone and speaker at 16 kHz.

This multi-modal dataset is designed for embodied AI researchers building autonomous navigation and computer vision systems intended to operate in real-world outdoor environments. By making the data freely available and releasing hardware designs as open-source, FrodoBots positions itself as infrastructure for the embodied AI research community rather than a walled proprietary data vendor.

Upcoming Products

Octo Arms is the next product in development. It adapts the Earth Rovers formula to robotic arm manipulation: a 3D puzzle game where players use robotic arm controllers to complete challenges, generating manipulation training data alongside the gameplay. Octo Arms was listed as invite-only at time of writing, with a public waitlist open through the FrodoBots website.

ET Fugi is an additional game variant of the Earth Rovers experience, described separately on the FrodoBots product roadmap.

Ecosystem Position on Solana

FrodoBots occupies a niche at the intersection of physical infrastructure networks (DePIN), consumer gaming, and AI data markets. Its model is structurally similar to how companies like Tesla crowdsource driver behavior data — the key difference being that FrodoBots' collection is explicit, gamified, and rewarded rather than passively harvested.

The project has a natural relationship with the broader Solana DePIN ecosystem, where token-incentivized physical networks have attracted significant developer activity. The Earth Rover Pro Driver NFT already represents on-chain credentialing tied to real-world operator roles — a pattern consistent with Solana's approach to tokenizing access and participation rights in physical systems.

Related to FrodoBots, [[PROJECT:2001]] (BitRobot Network) operates in the same robotics-on-chain space, describing itself as "The World's Open Robotics Lab." The BitRobot Access ID system launched publicly on Solana in August 2026, with FrodoBots listed as an associated project.

Summary

FrodoBots inverts the usual data collection model: instead of paying contractors to drive robots through scripted scenarios, the platform makes robot driving genuinely entertaining, rewards gamers for the quality and difficulty of their routes, and routes that output into open AI training datasets. With hardware in more than ten cities, over 2,000 hours of publicly released driving data, and a planned token-incentivized network anchored to Solana, FrodoBots represents one of the more concrete attempts to apply on-chain incentives to the embodied AI data problem.