AD-02 — What Is Automated Driving? The Automation Spectrum

FACT — AD-01 established the Sense → Perceive → Plan → Act pipeline, which answers how a decision gets made. This lesson asks a different question: who makes the decision, and who is accountable for its outcome? To measure that split precisely, we decompose SAE J3016’s core conceptual tool, the Dynamic Driving Task (DDT) — the direct foundation for the L0–L5 taxonomy in AD-03 and the ODD/fallback deep dive in AD-16/17.

In this lesson, you will learn to:

• Decompose the Dynamic Driving Task (DDT) into lateral control, longitudinal control, and OEDR.
• Distinguish the three architectural pillars underlying SAE J3016’s levels: DDT execution, DDT fallback, and ODD.
• Justify why the engineering community increasingly avoids the word “autonomous” in favor of “automated.”
• Analyze automation complacency and mode confusion at the L2/L3 boundary.
• Compare driver assistance against conditional/high/full automation across responsibility and real-world examples.

1. The Dynamic Driving Task (DDT): Three Core Sub-Tasks
2. SAE J3016’s Three Architectural Pillars: DDT Execution, DDT Fallback, ODD
3. “Autonomous” or “Automated”? Why a Word Choice Matters
4. Human Factors at the L2/L3 Boundary: Automation Complacency and Mode Confusion
5. Comparison: Driver Assistance vs. Conditional/High/Full Automation

---

1. The Dynamic Driving Task (DDT): Three Core Sub-Tasks

FACT — SAE J3016 defines the Dynamic Driving Task (DDT) as the sum of ALL real-time operational and tactical functions required to operate a vehicle in on-road traffic. At its core, the DDT consists of three sub-tasks that must run continuously and simultaneously:

1. Lateral Control: The steering-angle commands that keep the vehicle centered in its lane or execute a deliberate lane change. Physically delivered to the EPS (Electric Power Steering) motor.
2. Longitudinal Control: The throttle/torque and brake commands that regulate speed and following distance. Executed as an inverter torque command or e-Booster brake pressure.
3. OEDR (Object and Event Detection and Response): Detecting and classifying objects and events in the environment, prioritizing them, and preparing and executing the appropriate response. OEDR is the driving-task counterpart of the Perceive + Plan stages from AD-01.

DDT Sub-Task	Input	Typical Actuator/Module	AD Pipeline Counterpart (AD-01)
Lateral Control	Lane markings, road curvature	EPS steering motor	Act
Longitudinal Control	Target speed, lead-vehicle gap	Inverter, e-Booster	Act
OEDR	Camera/radar/LiDAR fusion	Perception + behavior planner	Perceive + Plan

CRITICAL — SAE’s formal DDT definition deliberately excludes the strategic layer (route selection, destination choice, charging/rest-stop planning — the third layer touched on in AD-01). The rationale: deciding where to go does not directly affect safety in the current instant, whereas lateral control, longitudinal control, and OEDR determine physical safety every single millisecond. This split is not arbitrary — every SAE level is defined by who executes the DDT (excluding the strategic layer), not by who plans the trip.

---

2. SAE J3016’s Three Architectural Pillars: DDT Execution, DDT Fallback, ODD

FACT — What determines a system’s SAE level is not its marketing name but the answer to three questions. AD-03 will apply these three pillars level-by-level, from L0 to L5; here we only establish their definitions.

1. DDT Execution: Who, at any given moment, is carrying out lateral control + longitudinal control + OEDR — the human, the system, or a shared split between them?
2. DDT Fallback (Recovery): When the system detects a failure (sensor contamination, unreliable perception) or the vehicle is about to exit its ODD, who brings the vehicle to a Minimal Risk Condition (MRC)? Does the human take over, or does the system reach a safe state on its own?
3. ODD (Operational Design Domain): The set of conditions the system is designed to operate under — road type (highway/urban), speed range, weather, lighting, geographic boundary. Outside the ODD, the system either never engages or triggers fallback.

INTERPRETATION — These three pillars really ask one question from three angles: “Before, during, and after something goes wrong, who is responsible?” DDT execution covers responsibility during normal operation, DDT fallback covers responsibility when something breaks, and the ODD defines the boundaries of what counts as “normal operation” in the first place. AD-03’s L0–L5 table shows exactly how these three pillars change hands across levels.

---

3. “Autonomous” or “Automated”? Why a Word Choice Matters

FACT — SAE J3016 itself deliberately avoids the word “autonomous” throughout the standard, using the systematic term “driving automation system” instead.

INTERPRETATION — In robotics and AI literature, an “autonomous” agent is one that sets its own goals and devises independent strategies to reach them. A driving automation system (ADS) does not work that way: the route is chosen by a human occupant, the system is bound by traffic law and an externally defined ODD, and it never decides where to go on its own initiative — it simply executes the transportation task it has been given (from point A to point B). “Autonomous” therefore carries a misleading connotation of will or agency that these systems do not actually possess.

WARNING — The loose, everyday use of “autonomous” in media and marketing is not a harmless simplification: it primes drivers to mentally model an L2 system as if it were L4. When a driver hears “autonomous,” their mental model calibrates toward “the system sees and decides everything” — yet in L2, OEDR responsibility remains 100% human (Section 4). This is one of the most common triggers of mode confusion, discussed below and grounded in the “Tesla naming controversy” case study in AD-03.

---

4. Human Factors at the L2/L3 Boundary: Automation Complacency and Mode Confusion

FACT — Automation Complacency is the phenomenon where an operator’s trust in automation degrades their active monitoring of that automation (Parasuraman & Riley, 1997). The more reliable and “seamless” automation appears, the less attentional budget a human is willing to allocate to it — precisely when that automation needs supervision the most.

CRITICAL — The L2/L3 boundary is especially hazardous for automation complacency because L2 physically feels almost identical to L3: at both levels the vehicle simultaneously manages steering and throttle/brake, and the drive “feels” effortless. Yet the underlying responsibility matrix is fundamentally different — in L2, OEDR is 100% human; in L3, OEDR is the system’s (AD-03, Section 1). If a driver cannot mentally distinguish that difference, they begin using L2 as if it were L3 (looking away from the road); this behavior is called Mode Confusion.

EXAMPLE — Across several US National Transportation Safety Board (NTSB) investigations into Tesla Autopilot crashes, a recurring finding pattern emerges: drivers, driven by excessive trust rooted in the system’s branded name and observed “seamless” performance, disengaged visual attention beyond the system’s actual perceptual limits (e.g., failing to distinguish an oncoming truck’s white trailer against a bright sky). NTSB has characterized this pattern as “foreseeable misuse” and has called on manufacturers for both driver monitoring (DMS, AD-18) and clearer communication of system limits.

ASSUMPTION — The NTSB finding pattern above is grounded in real, publicly available investigation reports; this lesson does not cite specific case numbers or dates because the same pattern is known to recur across multiple investigations (including cases from 2016 and 2018, in different states) — readers wanting a specific case citation should consult NTSB’s public case database (ntsb.gov).

The practical consequence of mode confusion can be measured as a re-engagement latency: for a driver whose eyes have left the road, the time to regain full control equals the sum of perceptual reorientation, decision, and action time:

$T_{reengage} \approx T_{perceive} + T_{decide} + T_{act}$

For a driver absorbed in a secondary task (phone, screen), this total climbs into the multi-second range and grows with how deeply attention was diverted — which is exactly why L3 must guarantee a fixed, bounded Take-Over Request (TOR) window, while L2 tolerates zero delay whatsoever (AD-03, Section 3).

---

5. Comparison: Driver Assistance vs. Conditional/High/Full Automation

Criterion	Driver Assistance (L0–L2)	Conditional/High/Full Automation (L3–L5)
DDT Execution	Human alone (L0/L1) or human+system jointly (L2, both axes)	System, fully within the ODD
OEDR Responsibility	Human, at all times	System, within the ODD
Fallback Trigger	The driver is assumed ready to intervene at any moment, with or without a warning	The system detects its own limit → hands off via TOR (L3) or executes its own MRC (L4/L5)
Legal Liability	100% human driver	Manufacturer/operator while inside the ODD (partial and conditional at L3, full at L4/L5)
Eyes Off the Road	Prohibited; enforced technically via DMS (AD-18)	L3: legally permitted within the ODD; L4/L5: conceptually moot (there may be no driver’s seat)
Real-World Example	Tesla Autopilot / FSD (Supervised) — Manufacturer Claim: “Full Self-Driving”; engineering reality: SAE Level 2. GM Super Cruise (hands-free, eyes-on, limited to mapped highways). Ford BlueCruise (hands-free, eye-tracked, “Blue Zone” ODD).	Mercedes-Benz Drive Pilot — certified genuine L3 within a narrow ODD (highway, limited speed, traffic congestion). Waymo — geofenced, driverless robotaxi service, genuine L4.

WARNING — The “Manufacturer Claim” label in the table is not arbitrary: Tesla’s own US regulatory filings (including correspondence with the California DMV) classify FSD (Supervised) software as SAE Level 2 — the “Full Self-Driving” name is marketing language, not a regulatory classification. This distinction is examined as its own case study in AD-03.

---

Frequently Asked Questions (FAQ)

Q: If a vehicle manages both steering and speed at the same time, does that automatically make it Level 3?

FACT — No. Simultaneous automation of both axes (lateral+longitudinal) is exactly the definition of L2 (Section 5). What makes a system L3 is not that both axes work together — it’s that OEDR responsibility sits with the system and the human is legally permitted to redirect attention elsewhere. AD-03 lays out the full matrix behind this distinction.

Q: Is the word “autonomous” completely off-limits?

INTERPRETATION — No — the word remains common in everyday language and popular writing, and this lesson does not reject it outright. In engineering and regulatory contexts (specifications, safety cases, type-approval documents), “driving automation” is preferable because it carries no ambiguity about the responsibility matrix. This academy uses “automation” throughout its technical material for that reason.

---

Summary

• The DDT consists of three sub-tasks a driver runs continuously and simultaneously: lateral control, longitudinal control, and OEDR; strategic route planning falls outside the DDT.
• SAE J3016’s level definitions rest on three pillars: DDT execution, DDT fallback, and ODD — AD-03 applies these pillars to L0 through L5.
• The word “autonomous” is deliberately avoided by SAE because it implies an independent goal-setting capacity these systems do not possess; “automated” is preferred.
• The L2/L3 boundary is especially prone to automation complacency and mode confusion because of near-identical physical feel but fundamentally different OEDR responsibility — a risk addressed by Driver Monitoring Systems (DMS) in AD-18.
• Driver assistance (L0-L2) versus conditional/high/full automation (L3-L5) is measured by who executes, who monitors, who takes over on failure, and which liability regime applies — never by the product’s name.

---

Sources and Verification Note

Sources: SAE J3016 — Taxonomy and Definitions for Terms Related to Driving Automation Systems, ISO 26262:2018 Road Vehicles — Functional Safety, Winner et al. — Automotive Handbook on ADAS, Parasuraman & Riley — “Humans and Automation: Use, Misuse, Disuse, Abuse” (Human Factors, 1997), NHTSA — Automated Driving Systems 2.0: A Vision for Safety. The DDT definition and the three architectural pillars in this lesson follow SAE J3016 directly. Automation complacency and mode confusion are drawn from human-factors literature (Parasuraman & Riley); NTSB investigation findings are reported as general patterns, and the section marked ASSUMPTION explicitly notes that no specific case number or date is cited.

Next Lesson

• AD-03 — SAE J3016 Levels of Driving Automation (Deconstructing L0–L5): Applying the DDT execution, DDT fallback, and ODD pillars established here to each level from L0 to L5; the Tesla FSD naming controversy as a worked case study; and the legal liability shift that explains why L3 has been so slow to reach market.