EGO: 65.0 mph
TTC: 2.82 s
DMS: GLANCING
LAT ACCEL: 0.12 g
Stopping Distance
58.4 m
Includes 120ms stack latency
Time to Collision (TTC)
2.85 s
Hazard clearance: +18.2 m
Intervention Margin
+1.28 s
Surplus safety buffer
Max Lateral Jerk
1.4 m/s³
Passenger comfort threshold: < 2.5
Perception (0.12s)
Alert / DMS (0.35s)
Human Motor Reaction (1.10s)
Braking Slew (1.85s)
Buffer (+1.28s)
Perception & Tracking
Acoustic / Haptic Chime
Human Driver Takeover
Mechanical Deceleration
Residual Safety Margin
SAE Level 2 Supervision: The Dynamics of Human-Machine Transfer
Under SAE J3016 Level 2 classification (Supervised Driving Automation), the vehicle executes continuous longitudinal (acceleration/braking) and lateral (steering) control. However, legal accountability, constant hazard scanning, and immediate operational fallback rest entirely with the human driver. When consumers describe supervised systems as "unbelievable," they are reacting to end-to-end neural network planners executing human-like trajectory synthesis. Yet from a systems engineering perspective, the boundary between seamless assistance and disengagement hazards is governed by unforgiving kinematic equations.
Vision-Only Optical Latency vs Radar/LiDAR
Pure optical architectures rely on continuous multi-camera temporal video networks (occupancy networks and 3D voxel transformers). Object persistence, velocity vectoring, and occlusion resolution require multi-frame inference windows (typically 3 to 6 frames at 36 fps), introducing 80–180 ms of pipeline latency before kinematic planning can actuate the steering rack or brake booster.
Driver Monitoring Systems (DMS) & Cognitive Saccade
Driver monitoring cameras track cabin infrared eye gaze, head pose pitch/yaw, and blink frequency. NHTSA and SAE human factor research demonstrates that a driver who is visually disengaged (gazing at a center touchscreen or smartphone) requires between 1.4 to 2.8 seconds to reacquire situational awareness, recognize closing velocity, reposition hands on the yoke, and exert evasive torque.
Core Trajectory & Evasion Kinematics
When a stationary or decelerating obstacle is identified in the ego lane, the planner must solve a dual optimization: emergency braking distance versus lateral lane-change swerve corridor. Emergency braking distance ($d_b$) is governed by friction coefficient ($\mu$), grade, and actuation slew:
Stopping Distance: d_stop = v_0 · (t_perception + t_reaction) + (v_0²) / (2 · μ · g)
Lateral Evasion: y_lat(t) = (a_lat_max / 2) · t² - (jerk_lat / 6) · t³
Time to Collision: TTC = d_relative / Δv
Takeover Margin: t_margin = TTC - (t_perception + t_dms_alert + t_driver_motor + t_decel_ramp)
At 65 mph (29.06 m/s), a vehicle travels 29.1 meters every single second. A combined system perception lag (150 ms) plus a modest distracted-driver recognition delay (1.5 seconds) burns 48 meters (157 feet) of roadway before deceleration even begins. This fundamental constraint is why continuous supervisory vigilance is physically non-negotiable.
Frequently Asked Technical Questions
What is the difference between SAE Level 2 and SAE Level 3 / Level 4?
Under SAE Level 2, the human driver must constantly supervise the automation and is legally responsible for collision avoidance at every millisecond without delay. Under SAE Level 3 (Conditional Automation), the system handles the dynamic driving task and allows the driver to divert visual and cognitive attention; when the system encounters an operational boundary, it must guarantee a transition demand buffer (typically 10 seconds) during which the automated system maintains safety until the human resumes control. SAE Level 4 requires no human fallback whatsoever within its Operational Design Domain (ODD).
Why does the simulator evaluate both braking and lateral swerving?
At speeds above 45 mph (72 km/h) on dry asphalt, lateral evasion (swerving 3.5 meters into an adjacent clear lane) often requires less longitudinal distance than coming to a full physical stop. However, lateral evasion requires verified occupancy of adjacent lanes, strict lateral jerk limits (< 3.0 m/s³ to prevent rollover and loss of tire adhesion), and driver awareness of adjacent traffic.
How does road friction (μ) affect the intervention margin?
On dry asphalt with high tire friction ($\mu \approx 0.85$), maximum braking deceleration can exceed 8 m/s². On wet asphalt ($\mu \approx 0.50$) or snow/ice ($\mu \approx 0.15–0.25$), stopping distance increases inversely with $\mu$. As braking distance doubles, the required Time to Collision (TTC) for safe driver intervention expands dramatically, shrinking the available margin into negative territory if speeds are not reduced proactively.
