Robotics Program
Multiverse-Mutiny Robotics Program
Top Grant: https://www.grants.gov/search-results-detail/351273 The Mind, Machine and Motor Nexus (M3X) Program
Other potentially related grants: - https://www.grants.gov/search-results-detail/348793 for tools - https://www.grants.gov/search-results-detail/306172 for tools - https://www.grants.gov/search-results-detail/320753 for education
Project Summary
Title: "Educational Robotics as a Scalable Data Collection Pipeline for Foundation Models"
Overview: This project addresses a critical challenge in robotics research: the scarcity of high-quality human demonstration data for training foundation models. We propose a novel approach that leverages an established educational robotics program to create a sustainable pipeline for robotic data collection while simultaneously advancing STEM education. By combining revenue-generating educational activities with research infrastructure development, we create a financially sustainable model that bridges the gap between robotics education and cutting-edge AI research.
Intellectual Merit: - Novel framework for collecting diverse human-robot interaction data through structured educational programs - Development of quality control methodologies for student-generated demonstration data - Research into how educational interactions can inform foundation model development - Investigation of curriculum design that optimizes both learning outcomes and data quality - Integration of latest foundation model architectures (like π0) with educational robotics
Broader Impacts: - Creates sustainable pipeline for robotics education and research data collection - Makes robotics education more accessible through revenue-generating program structure - Develops open-source tools for robotic data collection and validation - Establishes replicable model for combining education and research in robotics - Advances understanding of how humans learn and teach robotic systems
Budget Efficiency: The project leverages an existing educational program generating approximately $700,000 annually in operational funding. The requested NSF funding ($545,000 over three years) focuses specifically on research infrastructure and data collection capabilities, demonstrating significant cost-sharing and long-term sustainability.
Research Plan: 1. Development of data collection infrastructure integrated with educational robotics curriculum 2. Implementation of quality control systems for student-generated data 3. Creation of assessment tools measuring both educational outcomes and data quality 4. Integration of collected data with foundation model training pipelines 5. Validation through real-world robotic tasks
Funding Goals
| Category | Annual Cost | Covered by Program | Grant Request | Notes |
|---|---|---|---|---|
| Personnel | ||||
| 3 FT Professors ($120k ea) | $360,000 | $360,000 | $0 | Covered by course revenue |
| 1 FT Program Manager | $120,000 | $120,000 | $0 | Covered by course revenue |
| 3 PT Staff ($36k ea) | $108,000 | $108,000 | $0 | Covered by course revenue |
| 2 Research Assistants | $60,000 | $0 | $60,000 | New positions for data collection |
| Equipment | ||||
| Research Sensors/Tools | $40,000 | $0 | $40,000 | Year 1 only |
| Facility Usage | $60,000 | $60,000 | $0 | Covered by course revenue |
| Research Infrastructure | ||||
| Data Storage Systems | $20,000 | $0 | $20,000 | Year 1 only |
| Processing Hardware | $30,000 | $0 | $30,000 | Year 1 only |
| Quality Control Tools | $25,000 | $0 | $25,000 | Year 1 only |
| Educational Development | ||||
| Curriculum Development | $50,000 | $50,000 | $0 | Covered by course revenue |
| Assessment Tools | $25,000 | $0 | $25,000 | Year 1 only |
| Online Platform | $75,000 | $0 | $75,000 | Year 1 only |
| Travel | ||||
| Conferences | $15,000 | $0 | $15,000 | Annual |
| Site Visits | $10,000 | $0 | $10,000 | Annual |
| Workshops | $25,000 | $0 | $25,000 | Annual |
Total Grant Request by Year: | Year | Amount | Notes | |------|---------|--------| | Year 1 | $325,000 | Includes all infrastructure setup | | Year 2 | $110,000 | Ongoing research and travel | | Year 3 | $110,000 | Ongoing research and travel | | Total | $545,000 | 3-year total |
Key Points: 1. The educational program is self-sustaining for core operations (~$700k/year) 2. Grant focuses on research infrastructure and data collection 3. Major equipment and setup costs front-loaded in Year 1 4. Demonstrates significant cost-sharing through program revenue 5. Clear separation between educational operations and research activities
Let me analyze how our project aligns with the key requirements from the M3X grant program.
1. Human and Synthetic Actors - Required: Must involve interaction between human and synthetic actors with sensorimotor capabilities - Our Alignment: * Students (human actors) directly interact with robot arms (synthetic actors) * Teaching process naturally creates rich interaction data * Both actors engage in physical manipulation tasks * Multiple types of robots provide diverse interaction scenarios
2. Sensorimotor Interaction - Required: Must have bidirectional exchange through sensorimotor channels - Our Alignment: * Students provide demonstrations through teleoperation * Robots provide haptic and visual feedback * Physical tasks require continuous sensorimotor coordination * Real-time interaction through educational interface * Multiple sensory modalities (visual, haptic, proprioceptive)
3. Embodied Reasoning - Required: Actors must develop knowledge/expectations through physical interaction - Our Alignment: * Students learn robot capabilities through direct manipulation * Foundation models learn from human demonstrations * Curriculum progressively builds understanding * Both human and robot develop task strategies through interaction * Quality metrics capture growing mutual understanding
4. Physics-Based Environment - Required: Must involve real/simulated environment with physical laws - Our Alignment: * Real robot arms manipulating physical objects * Structured educational environment with clear physics constraints * Mix of simulation and real-world interaction * Progressive complexity in physical tasks * Measurable physical outcomes
Additional M3X Priorities Our Project Addresses: 1. Research Infrastructure - Creating open-source data collection tools - Building assessment frameworks - Developing quality control systems
- Multiple Modes of Interaction
- Direct teleoperation
- Language-based instruction
- Visual demonstration
-
Haptic feedback
-
Cross-cutting Perspectives
- Educational research
- Machine learning
- Robotics engineering
-
Human-robot interaction
-
Broader Impacts
- Democratizing robotics education
- Creating sustainable research pipeline
- Building foundation model resources
- Advancing understanding of human-robot learning
Revenue Breakdown Table
| Cohort | Prerequisite Revenue | 101 Revenue | 201 Revenue | 301 Revenue | Total Revenue |
|---|---|---|---|---|---|
| Cohort 1 - 30 students | $20,000 | $7,200 | $150,000 | $90,000 | $267,200 |
| Cohort 2 - 30 students | $20,000 | $7,200 | $150,000 | $90,000 | $267,200 |
| Cohort 3 - 30 students | $20,000 | $7,200 | $150,000 | $90,000 | $267,200 |
| Total | $60,000 | $21,600 | $450,000 | $270,000 | $801,600 |
Assumptions
- Cohort Size:
- Each cohort consists of 30 students.
-
There are 3 cohorts in a year, totaling 90 students.
-
Prerequisites:
- 1/3 of students take no prerequisites.
- 1/3 of students take a single prerequisite (Code, costing $500).
- 1/3 of students take two prerequisites (Code and one additional prerequisite like Prompt Engineering, totaling $1,500).
- Total prerequisite revenue for each cohort:
- ( \frac{30}{3} \times 0 + \frac{30}{3} \times 500 + \frac{30}{3} \times 1500 = $20,000 )
-
The Multiverse School captures this as revenue and uses it as initial bootstrapping cash, and for paying professors to get enough "robot-ready" learners together to stock the program.
-
Class Progression:
- Every student completes classes up to 301:
- 101: $240 per student.
- 201: $5,000 per student.
- 301: $3,000 per student.
- Total revenue per student: ( 240 + 5000 + 3000 = $8,240 ).
-
Each class will take a few weeks to teach, and there's a timing / rooms / physical equipment / professors problem here, (let Liz worry about it it's their bread and butter)
-
Revenue per Cohort:
- 101 revenue: ( 30 \times 240 = $7,200 ).
- 201 revenue: ( 30 \times 5000 = $150,000 ).
- 301 revenue: ( 30 \times 3000 = $90,000 ).
-
Total revenue for a 30-person cohort: $247,290
-
Total Revenue:
- Adding prerequisite revenue and class progression revenue across all cohorts (3) yields $741,600, with prerequisites yields $801,600 annually.
Robotics Program:
101
- Build a Rover
- Joaquin and I are running this 1st week of December in SF as a prototype
- parts list
- Roughly costs $31 per person just on amazon
- Robot Control
- Set up server for controlling robot
- $240 for the whole series (performing pricing and pacing)
201
- Build a Robot Arm (SO-100)
- Entry Level
- 3 sessions
- How to set up teleoperation
- Simple, one session after building the arm
- Train an action policy on the SO-100 (3 weeks)
- $5000 (1k/week) (needs pricing and pacing research)
301
- How to use types of robot arms that we got
- Worth $, Immediate job skill
- $3000? Needs pricing research, this is a gut call
401 - Research Program
- Train an action policy on the arms?
- for researchers, not so difficult, but physically dangeroussd
- How to set up a multi-step assembly process
- Worth $ but would probably need a more in-depth program for certs
- How to set up a sorting process
- Technical Risk? A Fast Follow on Covariant
Robotics Prerequisites
Some of the prerequisites (we can also teach these & have pricing and pacing on them), but also, crucially, feeder audiences who can often pay well for classes.
The Multiverse School can capture a lot of the Customer LifeTime Value (LTV) of a student who doesn't have some baseline readiness for a robotics program (at least one of the below, or hardware engineers can cross-pollenate)
- Code - 12 - 24 sessions
- $500-$3000 per student
- ML - 12 - 24 sessions
- $2000 - $5000 per student
- Prompt Engineering - 12 sessions
- $1200 per student
- Agent Engineering - 12 sessions
- $3000 - per student with robotics as a next step
- Data & Cloud Engineering - 12 sessions
- $3000 - per student
Professors
Who have confirmed they want in & posess the skills needed to meaningfully contribute
- Joaquin - Robots and CS (may or may not want to quit his job but would like to)
- Liz Howard (AI Liz) - Programs, CS, ML, RL, Robot Arms
- Liz Cheverra (Battlebots Liz) - Programs, Robot Arms, Fab, Battlebots
- JKL - CS, ML, Robots, CogSci
- Chris - Part Time while learning, is teaching, has other LOB
- Ryan - Programs, Biznaz, CS, ML, RL, Robot Arms (Part Time), CogSci/Neuro
Resources We Have For Sure:
- Robot Arms
- An Apartment Building That Needs Work To Be Inhabitable But Could Help
- A Factory To Host Classes and Robot Arms
- Humans and Brains
- An Audience
- Servers & 2080s
- SFCompute
Resources We Might Need:
- More Compute (probably get from SFCompute)
Program Costs - Idealized
Cost Parameters
- The Professors Need To Get Paid (everyone participating)
- 10k a month (120/yr) for Full Time
- 3k a month (36k / yr) for Part Time
- Andy probably should get compensated for the space (to support detroit nonprofits this will be an easy sell)
- ??? I dont know how much really.
- Let's Guess $5k / mo but obviously this can go up
- Maybe some physical equipment we're missing (say 30k-50k?)
- What Am I Missing?
Staff Costs:
- 4 FT * 120k = 420k / yr
- 3 professors at a time
- 1 program person
- 3 PT * 3k/mo = 36k / yr = 108k
- 5k/mo for Andy / DBC = 60k / yr ??? this can swing 100k+ bro idk plz tell me what you need and how to make this make sense for you
1 year = 588
2 years = 1.1m
- Robot Arms (1m) (handled)
-
Equipment (40k)
-
30% over what we need is 360000
| Category | Details | Cost |
|---|---|---|
| Staff Costs | ||
| - 4 FT * 120k/year | 3 professors, 1 program person | $420,000/year |
| - 3k/month * 12 * 3 | Additional staff | $108,000/year |
| - 5k/month for Andy/DBC | Dedicated support | $60,000/year |
| Subtotal (1 year) | $588,000 | |
| Subtotal (2 years) | $1,176,000 | |
| Additional Costs | ||
| ~~Robot Arms~~ | Already have them | ~~$1,000,000~~ |
| Equipment | Miscellaneous equipment | $40,000 |
| Buffer (30%) | Over what is needed | $360,000 |
| Total Target | Recommended amount to raise | ~$1,500,000 |