skynet week 5 h4d stanford 2016
TRANSCRIPT
SkynetSam Gussman (creative generalist), Alvin Goh (hacker), Olga Musayev (project
development), Kevin Mott (project manager)
Liaisons/Mentors: Steve Behmer, Ryan Blake, Wayne Chen# Interviewed this week: 11
6 Users, 2 Buyers, 2 Experts, 1 Competitor# interviewed total: 54
# New Blog Followers: 1, # Total: 21 Competitor, 1 Professor
Developing autonomous drones for situational awareness. Helping prevent battlefield fatalities by
pinpointing friendly and enemy positions.
Customer DiscoveryHypotheses Results Actions
Computer vision technology is currently capable of doing the detecting/recognition we need
Interviews, research, talking to experts
Partially validated. Detecting people from drone feeds can be done.
Continue researching software options to see if classifying enemy/friend/civilian is possible with computer vision technology today.
Acquisition: there are ‘critical paths’ available for getting our product into SOCOM
Interviews, sponsor engagement
Validated, multiple mechanisms for getting our product into SOCOM. We can get technology developed and into hands of users within 1 year
Talk to people involved in SBIR process, understand the requirements and more detailed hurdles we need to jump through
SOCOM is the preferred agency through which to buy our product
Interviews, research Partially validated. If we work with RRTO or another office specifically designed to innovate, can be flexible on agencies. But SOCOM is great too
Interview people at RRTO, CTTO, CTO, etc. to see if we can work with them
Experiments
Border control is a dual-use. Interview BP Agents None yet Validate hypothesis with more interviews
Next Week:
MVP
Source: Gaszczak, A., Breckon, T., & Han, J. (2011). Real-time people and vehicle detection from uav imagery. SPIE.
Detections from thermal images (people)
Detections from optical images (vehicles)
What we have now:
Envisioned product:
Hardware:DIY custom drone kitCamera (color / thermal)
Software:Open-sourced firmware
Dro
ne
ATAK Cursor On Target*
*Schema much like JSON
And
roid
plu
gin Skynet
Computer Vision (openCV) Drone navigation
Object detection
Object recognition DJI sdk / open-sourced sdk
System Architecture (draft)
Android device
MVP
Acquisition Fastest RoutesSBIR: Develop bare idea (TRL 1) to prototype (TRL 7)
Uses different pot of money (same color) required to be spent on small businesses
Phase 1 feasibility study done in 6 months, phase 2 brings beyond prototype, phase 3 (not covered) is commercialization and scaling up
Phase 2 can be initiated even 10yrs later, by any agency, without competition
Timeline: 1 month to hear back; 2-4 months to set up contract; 6 months of funding
S&T Process: Develop idea/prototype (TRL 3/4) to product (TRL 9)
Uses existing capability gaps posted on fedbizops
Evaluated by operators working with PM and an engineer
Benefits from engaging with operators and PMs ahead of time
Timeline: From close of BAA, 2 months to hear back, 4-8 months to set up contract, 2 years of funding
Find existing contract: Get money to buy stuff
Find company that has on-going contract with military; any other agency can give money
May be limited to purchasing materials and paying engineers
Timeline: After signing contract with company, can get money from agency same week, funding only for duration of the existing contract
Deployment Timeline ComparisonSBIR
S&T
ExistingContract
Notific
ation
Fundin
g
Fundin
g
Expire
s
1 Year (6 months funding)
Notific
ation
Fundin
g
Fundin
g
Expire
s
Fundin
g
Contra
ct Funding Expires
34 Months(2 Years funding)
Max 2 Years(19 Months, 3 Weeks Funding)
6 121
102 34
2, +1 week 24
Route 1: S&T Process-- Prototype to Product
BAA for existing gap
White Paper
White Paper
White Paper
S&T PM(TBD)
S&T PM’s Group of Operators
Other operators who talk to operators in PM’s group
Bureaucracy line
Other operators who talk to operators in PM’s group
S&T “Council”: S&T PMs +
Director Tony Davis
Engineer Steve D.
Proposal
Contract Negotiations between business and PM: Statement of Work (COW); Contracts Data Requirements List (CDROLS)
Step 1: find gap on fedbizops + submit whitepaper
Step 2: All whitepapers get evaluated by operators and engineer
Step 3: S&T PM brings selected white paper to “council” of PMs. They choose which projects to prioritize and fund.
Step 4+5: Submit proposal and negotiate deliverables (CDROLS) and what you will do (COW)
Step 0: “Free flowing” communication with operators and PM to show off product and get selected and prioritized
Route 2: SBIR- Small Business Innovation Research
SBIR Topic Area
Proposal Proposal Proposal
SBIR PM Sean M.
SBIR PM’s Group of Operators
(all components)
Other operators who talk to operators in PM’s group
Other operators who talk to operators in PM’s group
Engineers
Proposal Contract for Phase 1 Feasibility study
Step 1: find idea on SBIR topic area
Step 2: SBIR PM reaches out to whoever submitted topic; they evaluate
Step 3: SBIR PM brings selected white paper to “council” of PMs.
Step 4+5: Phase 1 feasibility study. If successful, can go to phase 2
Step 0: “Free flowing” communication with operators and PMs to influence process
SBIR Council: S&T Director Tony Davis + PMs
Get to phase 2 sole source... at any point... with any agency
Route 3: Find existing contract
Step 1: Find company that has existing contract with any agency
Step 2: Contract with company
Step 3: Get money through Military Interdepartmental Purchase Request
Ex. Stanford contracts with company which contracts with DIUX which lets SOCOM give us $10k for development.
Mission Model Canvas- Fly autonomously on route- Object detection, recognition, classification - Identifying and tracking friendlies, enemies, civilians- Integration with ATAK
- SOF (green beret, ranger, SEALS)
- SOCOM, RRTO, CTTSO, JCTD, CTO
- Programmable drone manufacturers (eg: Solo)
- ATAK Program
- Neurala (or other computer vision firm)
-Stanford Computer Vision Lab
Tactical operators - SOF (SF, Ranger, SEALS, MARSOC/Recon, drone operators)
Private sector and academic partners (Neurala, drone manufacturer, Stanford lab)
ATAK Program Manager
SOCOM Small UAV S&T/Acquisition Program Managers
Dual Use: Border Patrol
- Keep track of friendly, potential hostile and neutral actors/equipment automatically
- New contracts and revenue streams- Opportunities for tech development
-Added value to current program/technology
-Cost effective, agile and quick method of filling capability gaps
- Low-cost detection of movement and people at border
-Improved situational awareness for small units in combat
-reduce workload for tactical operators and feed-monitoring operators
-decrease strain on other traditional aerial assets
Options:1). S&T Process
2). SBIR Process
3). Find existing contract (short term funding)
Fixed:- Equipment - drones, development toolkits- Software design & engineering
Variable:- Costs of maintenance, updates, and training
-Demonstrate utility- gain support from senior personnel-Show opportunities for new revenue streams-Add utility to current program of record to increase adoption-Provide low cost product with new capability filling gapSaboteur - competition within SOCOM PMs
Beneficiaries
Mission AchievementMission Budget/Costs
Buy-In/Support
Deployment
Value PropositionKey Activities
Key Resources
Key Partners
- Programmable Drone
- Computer vision SDK (preferably neurala)
- ATAK API + source code
__= New
SkyNet: Value Proposition Canvas
- Image recognition software applied on video feed- Aerial perspective
Customer Jobs
Monitor the border for suspicious
activity- border area is too large, unable to ensure full coverage- Impossible to look at the whole border 24/7
Border Patrol
Gains
Pains
Gain Creators
Pain Relievers
- Quickly respond to suspicious activity - Survey more area without needing more men
- utilize low-cost drone swarm to increase border coverage
- distributed autonomous drones ensure high refresh rate across the whole border
Products& Services
Drone with visual recognition
software
SkyNet: Value Proposition Canvas
- Knowing the locations of friends/enemies/civilians- Reduced ambiguity - autonomous assessment of aerial perspectives and video feed Custome
r JobsShoot bad
guys without getting shot- Cognitive overload from
existing tools-Manpower losses - Hassle of systems not communicating
SOF Elements (combat)
Gains
Pains
Gain Creators
Pain Relievers
-Increased situational awareness on the battlefield- Reconnaissance of inaccessible places
- Integrate with current systems that soldiers are familiar with- Frees up another soldier
Products& ServicesSituationally
awaredrone with visual
interface
SkyNet: Value Proposition Canvas
- Automatic patrol- Facial Recognition of people identified in vicinity- Automatic threat detection to flag security threats Custome
r JobsForce protectionDay-to-day security patrols
are manpower intensive
Difficult to be always aware of surroundings
SOF Elements (non-combat)
Gains
Pains
Gain Creators
Pain Relievers
- Reliable 24/7 Security Awareness- Knowing who lives around you, and who doesn’t- Ability to respond quickly to threats
- Swarms of drones can replace patrol units
-24/7 situation awareness of surroundings
Products& ServicesSituationally
awareautonomous
drones
SkyNet: Value Proposition Canvas
- Access into SOCOM acquisitions
- building company’s presence in military
Customer Jobs
Make more moneyBuild customer relationships
- Difficult to get large contracts, technology still maturing, only have limited consumer market share
Commercial Partners
Gains
Pains
Gain Creators
Pain Relievers
-New revenue stream-Potential long term partnership
- Contracts with military, willing and eager to adopt new technology (early adopters) even if tech is not fully matured
Products& ServicesPartnership /
contract
SkyNet: Value Proposition Canvas
- Integrating drone camera feed enhanced with image recognition
- augmenting ATAK interface with image processing
Customer Jobs
Manage ATAKSystem under
cost- Proprietary technologythat does not talk to othertech- low adoption rate, ‘fancy new toy’ that’s not used much
ATAK Project Manager
Gains
Pains
Gain Creators
Pain Relievers
-Increased value to ATAK program
- Easy integration with ATAK- Promote culture of open-source- improving receptiveness to ATAK
Products& ServicesAutonomous
drone with live video feed
integrated with ATAK
SkyNet: Value Proposition Canvas
-Image recognition software first of its kind in the military
- tool could potentially be applied across platforms (as a software plugin)
Customer Jobs
Manage ProgramsFill gaps
- Tech is new and recognitionof potential enemy may triggerblowback- new technology always too expensive, sometimes they don’t get used after purchasing
SOCOM Small UAS PM
Gains
Pains
Gain Creators
Pain Relievers
- Provides new tool that existing programs don’t provide - Potential to scale
-Retain man-in-the-loop, test software during training stateside and slowly scale to overseas.- Maintain low costs by utilizing COTS, consistently engage user feedback at all stages of development.
Products& Services
COTS Drone with visual recognition
software
Backup Slides: Acquisition Process
Saboteurs + Other DangersProject Manager for Personal Reconnaissance Devices
There is a jurisdictional fight between the PMs for Personal Reconnaissance Devices and Small UAV Projects
Raven ProgramSmall fixed wing reconnaissance drones, the Raven chief amongst them, stand to lose both funding a prestige
Ourselves Any misclassification of a civilian that leads to injury is likely to completely derail the project.
‘Personal’ saboteursThis week we learned about an angry secretary who almost killed a project’s funding.
Tactical Beneficiary Archetype
Special Forces - they enter an area for few months and live with locals for a while, train them, advise them on combat.
Drone Operators - they pilot and monitor drone feeds to conduct targeted air strikes and protect troops on mission
Rangers - they enter a specific area, conduct a raid, and exit.
JTAC - responsible for calling and targeting air support.
Sergeant of the Guard (SOG) - they are responsible for force protection.
Value Proposition BriefProblem Existing solutions Our solution
Identify and locate hostile and neutral actors/equipment
Plot on a mapDrop enemy icons on JCR/BFTSoldier awareness
Real-time updates/tracking of persons of interest through computer vision
Special Forces monitoring of allied force
High level drone feed fed to SF, but limited 2-way communication.ATAK may show locations of allied force
Drone extends communication networks and allow audio and video communicationMonitors locations of allied force to prevent airstrike casualties
Collateral Damage Soldiers mark out potential CD on mapsGround force commander's’ judgementDrone operator in the US runs algorithm to determine CD from potential airstrike
Drone uses existing database to mark civilian buildings on ATAK map; Drone identifies potential civilians and marks location in ATAK using Computer Vision
Fratricide, especially when calling in airstrikes and artillery
ATAK, GPS in radios, 9-line call format, strobe lights on uniforms: all identify friendly locations
Low-flying drone detects friendly locations and conveys to JTAC and pilot. Could also run algorithms determining risk of fratricide; AR shows soldiers locations of friendlies
Force protection duties (guard) are manpower intensive
Patrols and Soldiers at static positions are a staple in force protection duties
Autonomous drone would circle compound and recognize dangers. Can form fleet of autonomous drones to complement patrol duties