| name | machine-autonomous-road-vehicle-robotaxi |
| description | Non-humanoid autonomous machine: **Autonomous road vehicle (robotaxi)** — carry passengers point-to-point with no human driver. Best in: geofenced urban and suburban roads. A self-driving/self-operating platform whose planning brain issues physical actions as tool calls over a perception-to-control stack (trained on world models, simulation, and RLAIF) inside a defined ODD with teleoperation fallback. Use this skill to plan or operate the platform anywhere this physical job is needed, even if the user only describes the underlying need. |
Autonomous Machine — Autonomous road vehicle (robotaxi)
Layer: Non-humanoid autonomous machine (cross-economy) · Best environments: geofenced urban and suburban roads
Operated by: ../../embodied-ai-stack/ roles (autonomy, fleet ops, teleoperation, safety) · Shared concepts: ../../../00-framework/SKILL.md
Primary job to be done
Carry passengers point-to-point with no human driver.
What it is
An SAE L4 self-driving car operating within a mapped ODD; supervised by remote operators with a verified safe-stop.
When to use this skill
When a task needs the physical job "carry passengers point-to-point with no human driver" in environments such as geofenced urban and suburban roads. Pair with the relevant operating-system skill (01–23) for domain rules and the human accountability boundary, and with _catalogs/embodied-ai-stack/ for the roles that build, operate, and keep it safe.
Cognitive and control architecture (assumed)
These are non-humanoid autonomous machines — vehicles and equipment that drive, fly, or operate themselves. They share the project's brain-and-tool-calls model, adapted for mobility and heavy equipment:
- Cognitive core (the autonomy "brain"). A foundation/LLM-based planner handles mission-level reasoning, natural-language tasking, and long-tail edge cases, sitting over a perception → prediction → planning → control autonomy stack. The brain decides what and where; learned and classical controllers execute how at high frequency. A fleet may share one model or specialize by platform.
- Actions are tool calls. The machine exposes actuation primitives as tools — e.g.
follow_route, set_speed, change_lane, lower_header, dump_bucket, take_off, survey_area, spray_zone, return_to_base — which the brain invokes and low-level controllers carry out.
- Trained on world models + simulation. Planners and policies are trained against world models (learned simulators that predict vehicle dynamics, terrain, weather, and the behavior of other agents) and large-scale driving/field simulation (robot gyms), then transferred to hardware with fleet data and imitation learning.
- Many training paths (RLAIF is one). Behavior is learned through imitation from human driving, model-based and offline RL, sim-to-real, and RLHF/RLAIF, then distilled into the SLMs and tiny models that run on-vehicle — with deterministic planners and controllers (MPC, search) for the safety-critical loop. The autonomy brain is right-sized per function; see
_catalogs/capability-optimization/.
- ODD + safety case. Each machine operates inside a defined Operational Design Domain (the geography, weather, speed, crop, or site it is certified for) and a documented safety case, rated on the SAE levels of automation (L0–L5) for road vehicles or equivalent for off-road and aerial platforms. A verified safety layer can trigger a minimal-risk maneuver (controlled safe-stop / return-to-base / hover) independently of the planning brain.
- Teleoperation fallback. A remote operator supervises and takes over for situations outside the ODD or below a confidence threshold.
Operating implication: physical-world failures are high-consequence, so the safety layer, ODD boundary, and teleop fallback are mandatory and independent of the planning brain. Public-road and airspace operation additionally require regulatory authorization (e.g. SAE-level / FMVSS treatment for road vehicles; FAA Part 107 and BVLOS waivers for drones).
Division of labor and safety
- Human owner / fleet operator — owns the safety case, the ODD, and stop authority; accountable for incidents.
- Autonomy brain — perceives, predicts, plans, and issues actuation as tool calls within the ODD.
- Low-level controllers — execute motion/actuation at high frequency.
- Verified safety layer — triggers a minimal-risk maneuver (safe-stop / return-to-base / hover) independently of the brain.
- Remote operator (teleop) — supervises and takes over beyond the ODD or below a confidence threshold.
Architecture-specific failure modes
- Long-tail / edge cases — rare scenarios the planner mishandles. Mitigation: conservative ODD, teleop fallback, continuous scenario mining.
- ODD exit — conditions drift outside the certified domain (weather, dust, lighting, unmapped area). Mitigation: detect-and-degrade to a minimal-risk maneuver.
- Sensor degradation / spoofing — rain, dust, glare, GPS jamming, adversarial markings. Mitigation: sensor fusion, redundancy, anti-spoofing, conservative fallback.
- Sim-to-real gap — world-model/simulation training diverges from reality. Mitigation: shadow mode, staged deployment, real-world validation.
- Mixed-traffic / human interaction — misreading pedestrians, livestock, ground crew, or other drivers. Mitigation: predictable behavior, low-speed zones, explicit right-of-way rules.
- Teleop latency / link loss — remote takeover delayed or lost. Mitigation: onboard safe-stop, bounded autonomy, comms redundancy.
- Fleet model-monoculture — a shared brain fails in lockstep. Mitigation: model diversity, staged rollout, geofencing.
Adapting to any nation (context modifiers)
Ownership ranges from fleet-as-a-service to cooperatively shared or rented machines; regulation (road approval, airspace/BVLOS, mine/site rules) and infrastructure (maps, connectivity, GPS/RTK) gate where it can run. In low-connectivity settings, on-board autonomy and safe-stop matter more than teleop. Re-read through:
- Scale (city-state → federation): whether this role is unified or layered across local/regional/national tiers.
- State capacity (fragile → high-capacity): whether the owning institution exists and can be held to account, or the job is met by markets, households, NGOs, or donors.
- Income level (low → high): affordability of automation and the balance of subsistence vs. wage work.
- Formality (informal → formal): whether the people and assets this role acts on appear in any registry at all.
- Resource & geography: which hazards and dependencies dominate (water-scarce, flood-prone, landlocked, trade-dependent).
- Political system & legitimacy: where the human-accountability boundary actually binds and who may hold power to account.
