Personal Background
I’ve always had a huge interest in Tesla and Waymo - they’re mostly what inspired me to join EUFS.
Objective & Background:
- Automate vehicle operation to reduce human intervention while ensuring safety, efficiency, and accessibility.
- Leverage advanced sensors and algorithms to mimic human driving capabilities while minimising human error. (Reliability)
Levels of Autonomy:
- L0: No autonomy. Driver handles everything. (Think old Porches without power steering)
- L1: Assistive functions like speed-keeping or lane-keeping (think helpers).
- L2: Partial automation. The car can handle steering and acceleration but needs the driver alert at all times. (Tesla Autopilot-ish)
- L3: Conditional automation. The car drives itself under specific conditions but can demand the driver to take over.
- L4: High automation. Fully autonomous in limited settings (e.g., urban centres).
- L5: Full automation. No steering wheel. No driver.
How It Works:
- 🟥 Perception: Sensors like LiDAR, cameras, and radar gather data about the car’s surroundings.
- Example: Detecting a pedestrian in a crosswalk.
- 🟧 Localisation: Using GPS and SLAM to pinpoint the car’s position on the map.
- 🟨 Prediction: Machine learning algorithms predict the motion of surrounding vehicles, pedestrians, etc.
- e.g., That cyclist will likely swerve left.
- 🟩 Planning: Develop the safest and most efficient driving path based on predictions and static maps.
- e.g., Slow down for the cyclist.
- 🟦 Control: Adjust acceleration, braking, and steering to execute the plan.
- Big up Neural Networks
Interestingly 🤔
Those are the exact sub-teams of Formula Student.
Hacking Risks:
- Cars get hacked: Self-driving cars introduce some novel hacking risks. Imagine driving down the road and your car just starts going on it’s own merry way. At the end of the day they’re computers on wheels. This brings on the same computer vulnerabilities we’ve seen over and over again.
- People try fuck with the roads: Computer Vision is vulnerable to statistics. What if people put mathematically specific pieces of tape on stop signs? Then they’re be perceived as “Go straight on”