AI summary by TwinMind
Aicha Evans: Your self-driving robotaxi is almost here | TED
TED
11m 2s
Speaker Introduction and Personal Journey
- Aicha Evans from Senegal, West Africa
- Early love for technology, science, engineering
- Three formative experiences:
- Age 7: Flying unaccompanied between Dakar, Senegal and Paris
- Travel as portal to knowledge, different environments, adaptation
- Hacking father's rotary phone lock to call Paris friends
- Father tired of long-distance bills, put lock on phone
- Hacked lock, bills continued arriving
- Internet emergence
- Age 7: Flying unaccompanied between Dakar, Senegal and Paris
- Common thread: physical and virtual transportation at center of innovation flywheel
Current Mission and Industry Context
- Part of movement bringing transportation and technology together
- Goal: changing how we move people, goods, services
- Transformation involves robotaxis
- Addresses audience skepticism about driverless cars
- Always promised "next decade"
- Alphabet soup of companies, confusion about who does what
Problems with Current Transportation Models
- Personally owned cars: wasteful expense, pollution, urban traffic
- Self-driving shuttles: optimized for many, can't do point A to point B
- Current "personal self-driving" cars: still require human safety driver
- Evans owns one, acts as safety driver when using it
Robotaxis Solution
- Take you specifically from point A to point B
- When not using, somebody else uses them
- Being tested today
- Claims we're "on cusp of finally delivering that vision"
Computer Vision Technology Core
- Real-time digital representation of world and interactions
- Benefits from advances in computers, sensors, machine learning, software
- Camera systems at core:
- See agents (cars), locations, actions
- Pedestrians, locations
- Vehicle can see skeleton framework showing direction of travel
- Details like construction worker vs distracted pedestrian on phone
Camera Limitations
- Cannot match human brain's ability to understand/interpret environment
- Provide high-resolution imaging, continuous coverage
- Don't get fatigued, impaired, drunk
- Still can't see/measure everything
- Need supplementation for autonomous robotaxis
Sensor Fusion Technology
- Radar: direction of travel, measures agent movement within centimeters per second
- Lidar: objects/shapes using depth perception, long-range, night vision
- Long-wave infrared: see heat-emitting agents (animals, humans), especially at night
- Multiple sensor modalities at all four corners of vehicle
- 360-degree field of vision, continuous, redundant
- All outputs fused together for comprehensive view
Addressing Real-World Scenarios
- Edge scenarios always exist at boundary of real-world testing
- Some scenarios too dangerous to test on real streets
- Human beings underrated in abilities
- Solution: simulation
- Construct millions of scenarios in fabricated environment
- Test software reactions
- Can add/remove elements, see reactions
Human-in-the-Loop System
- Similar to aviation systems
- Teleguidance operators at control center
- When vehicle stuck or doesn't know what to do:
- Asks for guidance/help
- Receives remote assistance
- Proceeds with guidance
Historical Context of Vision Systems
- Vision systems assisting humans long time
- Examples:
- Microscopes: studying microbes, cells
- Telescopes: detecting galaxies millions light-years away
- Transformed industries: medicine, farming, astrophysics
Computer Vision Evolution
- Started as thought experiment: replicate human vision using cameras
- Now graduated with sensors, computers, AI, software innovation
- Goal: surpass what humans can see/perceive
- Progress made but more work needed
Safety and Testing Requirements
- Autonomous robotaxi must be safe, right, reliable every time
- Requires rigorous testing and optimization
- 94% of crashes caused by human error
- Computer vision enables move from problem-solving to problem-preventing
Future Vision
- Next generation scientists/technologists globally (Silicon Valley, Paris, Senegal, worldwide)
- Will be exposed to computer vision applied broadly
- All industries will be transformed
- Experience world differently
- Gift owed to next generation
- Computer vision will help solve many problems
