Governing Robots: The Questions Cities Miss

Author: Bern Grush Date Published: August 9, 2026

A new paper by Lina Moe and Clint Andrews, "Make Room for Robots,"[1] gives policymakers a diagnostic framework for critical governance questions that cities must answer:

a) which existing rules apply to robots? and b) where do gaps remain?

Their Framework

Moe and Andrews start from a simple observation: "Robots" is not one category. A Roomba, a hospital delivery bot, and a delivery drone share little except mobility and autonomy. Applying one policy to all three makes no more sense than one traffic law for forklifts, bicycles, and 18-wheelers.

To sort this out, they propose three diagnostic questions:

1. what kind of space is the robot in?

2. what kind of interaction does it create? and

3. what infrastructure does it depend on?

Each question routes the practitioner to a different governance toolkit.

In their framework, "space" is split into four layers:

Private spaces, such as hospital operating rooms and building interiors, are situated inside mature governance systems including building codes, product liability, applicable safety rules, and insurance.

Semipublic spaces, e.g., malls, subway platforms, and campus corridors, are conditionally open to the public, with an institution setting the rules (such systems can fail if public legitimacy remains unresolved, as happened in a 2023 security robot trial in New York's subway[2]).

Public rights-of-way, especially streets and sidewalks, are the toughest case as such pathways are shared by strangers, layered with overlapping rules, and now contested at the curb by pedestrians, wheelchairs, bikes, ride-hail, and robotaxi drop-offs.

Airspace remains under federal authority but still produces local effects like noise and privacy concerns that cities may be able to address through zoning and other ordinances.

The authors’ second and third questions — interaction type and infrastructure coupling — sharpen the picture further. One-to-one robots (exoskeletons, companion devices) map onto consent and product-liability law. Robots moving among non-involved bystanders rather than collaborating with consenting users expose a critical gap: rights-of-way laws assume human or human-operated agents, and no framework tells a robot how to yield or defer.

The December 2025 Waymo blackout case study shows what happens when infrastructure coupling compounds this gap at scale — nearly 1,600 robotaxis frozen on San Francisco streets during a signal outage,[3] tying up 911 dispatch and exposing that no regulatory framework (so far) anticipates a monoculture fleet failure as opposed to an isolated device or vehicle failure.

Where We'd Push Further

Many existing governance regimes work to some degree for private spaces and airspace, but semipublic spaces and public rights-of-way are far more challenging. It is for these two layers that the Urban Robotics Foundation has spent years focussed on governance tools. The Moe-Andrews diagnostic — both needed and correct — undersells how much structure these spaces still need.

Semipublic spaces are highly unstructured and variable. The authors are right that an institution sets initial rules and public legitimacy is the open question. But "public legitimacy" is not a single hurdle — it recurs at every stage of fleet growth.

Our staged deployment model (Trials → Early Deployment → Scaled Deployment → Expansion) treats legitimacy as a moving target: a hospital delivery fleet of three robots on one floor raises different stakeholder concerns than the same fleet expanded to twenty robots across a hospital network sharing Wi-Fi with the patient record system.

Institutions that authorize a pilot rarely have the governance capacity to manage what that pilot becomes. The NYPD subway case was not simply a semipublic-space failure — it was a readiness failure, where an institution deployed at a maturity level its own oversight processes, and its user cohort, weren't prepared for. NBC News Story: https://www.nbcnewyork.com/news/local/nyc-retires-420-pound-nypd-subway-surveilling-robot-after-pilot-ends/5099973/

Public rights-of-way need more than "existing traffic law, licensing, and liability regimes." That toolkit describes 20th-century vehicles, not fleets of mixed-operator, mixed-task robots competing for the same sidewalk and curb.

Two gaps stand out: Orchestration and Monetization.

First, orchestration. When two independent delivery operators run robots through the same right-of-way, per-operator routing algorithms cannot prevent conflict at shared chokepoints — a problem already documented at U.S. crosswalks. Air traffic control offers the working analogy: standardized protocols and coordinated timing windows, not vehicle-by-vehicle self-management. Multi-fleet, hierarchical orchestration becomes necessary well before the "significant" fleet count that triggers it — and cities that wait until congestion appears are already behind.

Second, monetization as a governance tool (rather than just revenue). Fleet licensing fees, certification fees, and usage-based pricing all shape behavior. Dynamic monetization of orchestration systems can create equitable sidewalk, crosswalk, and curb access, control congestion, and fund the enforcement systems these deployments require.

A maintenance robot de-icing a public sidewalk for a public works department is a different monetization case than a robot from a commercial delivery fleet occupying the same space. Cities that treat monetization purely as license fee collection miss its role as a demand-management lever, i.e., the same role variable road pricing plays for private road vehicles.

The Real Governance Gap

Moe and Andrews close with three actions cities can take now:

1. use procurement as leverage,

2. build a coordination layer, and

3. publish permit data for transparency.

These are sound first moves, and align closely with the guidance we've compiled in URF's own PMR governance work. But procurement leverage and a single point of contact solve the early deployment problem. They do not solve what happens once fleets scale, multiply, have multiple device designs and purposes all while interfacing with adjacent systems.

Orchestration, certification, and monetization design need to start well before a city needs them, not after the first multi-fleet conflict forces the issue.

The road-vehicle rulebook took decades to mature and came at a cost the paper pointedly notes: 40,000 US traffic deaths a year, a toll we are far from solving for. Smaller mobile robots differ from motor vehicles in scale and lethality, but not in the risk of governing them reactively.

The diagnostic Moe and Andrews offer is a genuinely useful start. The harder work — staged orchestration, certification, and monetization design for semipublic and right-of-way deployments — is where cities need to start focussing. URF works with cities and institutions on exactly this kind of staged deployment and governance design. If any of this maps onto a project you're tackling now, reach out — we're happy to provide guidance and support.

---

[1] Moe, L., Andrews, C. (2026) “Make Room for Robots” https://issues.org/urban-robots-moe-andrews/

[2] Rubinstein, D., Meko, H. (2024) "Goodbye for Now to the Robot That (Sort Of) Patrolled New York’s Subway" https://www.nytimes.com/2024/02/02/nyregion/nypd-subway-robot-retires.html

[3] Wang, F. (2026) "San Francisco leaders press Waymo on stalled robotaxis during December power outage" https://abc7news.com/post/san-francisco-leaders-press-waymo-stalled-robotaxis-during-december-power-outage/18669520/