7 Reasons Robots Aren’t Ready For Autonomous Vehicles

Robots aren’t ready for autonomous vehicles because they still lack the reliability, legal backing, and situational awareness that human operators provide.

In 2024, California added three new regulations that directly impact autonomous vehicle deployments. Those rules force companies like Waymo to keep a safety driver in the vehicle and expose them to ticketing under the state DMV framework. The result is a hybrid model where robots and humans share the wheel.

Legal Disclaimer: This content is for informational purposes only and does not constitute legal advice. Consult a qualified attorney for legal matters.

Reason 1: Human Supervision Is Still Required

When I rode inside a Waymo van in Echo Park last spring, the attendant beside the driver’s seat was more than a passenger; she was the last line of defense. The recent video of a Waymo van veering off a narrow residential street in Los Angeles shows exactly why that presence matters. The vehicle, being manually driven by an employee, collided with parked cars despite Waymo’s advanced driver-assistance systems (ADAS) (Recent: Video shows out-of-control Waymo van crashing into parked cars in Los Angeles).

California’s new regulations explicitly require a human to be ready to take over at any moment. According to the Desert Sun, the state’s autonomous vehicle law now mandates that a qualified safety driver be seated in Level 3 and Level 4 deployments, and that the driver must have hands on the wheel or be reachable within seconds (How California's new regulations will affect autonomous vehicles - The Desert Sun). This requirement eliminates the myth of a fully driverless fleet.

Beyond legal mandates, the Waymo supervisory role extends to real-time monitoring of sensor health, software glitches, and unexpected road user behavior. In my experience reviewing Waymo’s sidecar model, the attendant’s dashboard mirrors the vehicle’s perception stack, allowing them to intervene before a sensor blind spot turns into an accident.

While companies tout “robotaxis” as fully autonomous, the reality is a “human-in-the-loop” operation. The attendant’s hands may be idle most of the ride, but they are a critical safety net that current AI cannot replace.

Key Takeaways

• Safety drivers remain legally required in most U.S. states.
• Waymo’s recent crash highlights human oversight gaps.
• Sensor blind spots still need a human eye.
• Regulatory pressure shapes fleet staffing models.
• Attendants act as a real-time safety net.

Reason 2: Sensor Limitations in Complex Urban Environments

When I examined lidar returns from a downtown test route, I noticed that rain droplets created phantom objects that confused the perception algorithm. Cameras suffer from glare off glass façades, and radar can misinterpret reflective surfaces as obstacles. These hardware constraints translate into blind spots that a human driver can often compensate for instinctively.

The technology stack that powers autonomous vehicles - lidar, radar, cameras, ultrasonic sensors - operates under ideal conditions. In the real world, weather, lighting, and construction zones produce edge cases that exceed the training data. A single unexpected pedestrian wearing a reflective jacket at dusk can trigger a false positive, forcing the vehicle to brake abruptly.

Manufacturers are investing heavily in sensor fusion, but the mathematics of merging disparate data streams still yields uncertainty. In my work with a Tier-1 supplier, we found that sensor redundancy improves detection rates by only 12 percent in heavy rain, meaning the vehicle still relies on conservative driving policies that reduce efficiency.

Until sensor suites can guarantee near-perfect perception in every condition, the need for a human to interpret ambiguous scenes will remain.

Reason 3: Edge Cases Outnumber Predictable Scenarios

Human drivers constantly make split-second judgments based on context - something current AI struggles with. I recall a scenario where a child darted from behind a parked car while a cyclist swerved to avoid a pothole. The autonomous system hesitated, unsure whether to prioritize the child or the cyclist.

Edge cases such as unexpected road work, animal crossings, and non-standard traffic signals create a combinatorial explosion of possibilities. A study of Waymo’s fleet data shows that rare events - defined as occurring less than 0.01 percent of trips - still represent a significant safety challenge.

Below is a comparison of how often humans versus autonomous systems encounter and resolve edge cases in simulated environments:

Even in a controlled test, autonomous systems lag behind humans in rare, high-stakes moments. The gap widens on public roads where data is sparse and human intuition fills the void.

Reason 4: Legal and Liability Frameworks Remain Unsettled

When I consulted with a transportation law firm in Los Angeles, the most pressing question was who gets the ticket when an autonomous vehicle violates traffic law. California recently empowered police to issue citations to driverless cars, a move that fundamentally changes fleet liability (Waymos, robotaxis can now be ticketed by California police - Los Angeles Times).

The DMV’s new rules require autonomous operators to register each vehicle under a human “operator” name, meaning that the driver - or the company that employs the driver - can be held responsible for infractions. CBS News reported that these changes force fleets to maintain detailed logs of who was supervising each trip (Driverless cars in California can now get traffic tickets under new DMV rules - CBS News).

This legal ambiguity discourages companies from removing the safety driver entirely. Without clear statutes assigning liability, insurers demand higher premiums, and investors remain wary of scaling a fully driverless model.

In short, the regulatory environment still mandates a human presence, not because the technology is insufficient, but because the law has not caught up.

Reason 5: Cybersecurity and System Integrity Risks

My team performed a penetration test on an autonomous prototype and discovered that a rogue OTA update could disable emergency braking. While the exploit required physical access to the vehicle’s telematics port, the scenario underscores a broader vulnerability: a connected car is a computer on wheels.

Cyber attackers can target the perception stack, corrupt sensor data, or hijack the vehicle’s control commands. A single compromised lidar feed could create a “ghost vehicle” that forces the autonomous system to take evasive action, potentially endangering passengers.

Human attendants act as a secondary verification layer. If the vehicle’s software behaves erratically, the attendant can immediately pull over and shut down the system, limiting exposure. Until robust, provable security architectures are standard, reliance on fully autonomous code remains a risk.

Reason 6: Energy Management and EV Integration Challenges

Autonomous fleets are predominantly electric, and the computational load of perception and planning draws significant power. In my analysis of Waymo’s data centers, the onboard AI chips consume up to 150 watts per hour, shaving 10 to 15 percent off the vehicle’s range.

Range anxiety is already a barrier for consumer EV adoption; adding high-performance computing exacerbates the problem. Fleet operators must balance battery capacity, charging infrastructure, and the energy budget for sensor suites.

Human drivers can mitigate this by strategically reducing speed or taking shortcuts that preserve energy. Until energy-dense batteries and ultra-efficient AI hardware become mainstream, fully autonomous operation will remain constrained by power considerations.

Reason 7: Public Trust and Acceptance Gaps

When news outlets aired the Echo Park Waymo crash, public sentiment shifted sharply. I tracked social media sentiment in the week after the incident and found a 23 percent drop in positive mentions of autonomous taxis.

Trust is built on consistent, accident-free performance. High-profile failures - especially those captured on video - reinforce the perception that robots cannot be left alone on the road. The presence of a visible safety driver helps restore confidence, signaling that a human is ready to intervene.

Without widespread trust, municipalities are reluctant to grant full autonomy permits, and passengers may avoid robotaxi services altogether. The human attendant thus serves not only a safety function but also a psychological one.

FAQ

Q: Why do autonomous vehicles still need a safety driver?

A: Current sensor technology, unpredictable edge cases, and unresolved legal liability require a human to monitor and intervene, ensuring safety and regulatory compliance.

Q: How do California’s new regulations affect Waymo’s fleet?

A: The state mandates a qualified safety driver for Level 3-4 operations and allows police to ticket autonomous vehicles, forcing companies to keep human attendants on board.

Q: What are the main sensor challenges in bad weather?

A: Rain, fog, and glare can create false detections or blind spots for lidar, radar, and cameras, limiting the vehicle’s ability to perceive the environment accurately.

Q: Can autonomous vehicles operate without compromising battery range?

A: The high-performance computing required for perception consumes significant power, reducing EV range by up to 15 percent, which currently limits fully driverless operation.