Transportation Robotics Use-Case Priority Study

Executive summary

• Start with work-zone traffic-control-device inspection. It combines direct worker exposure, a task that robots can meaningfully substitute, and continuity with ScholarMind's current CARLA and perception prototypes.
• Rail track inspection is the most mature adjacent market. FRA already operates autonomous inspection vehicles, so the opportunity is differentiated change detection, interpretation, and reporting rather than proving that automated inspection is possible.
• Roadside incident and towing scenes may be more dangerous, but full autonomy is not the right first promise. Begin with protected remote reconnaissance, risk-zone mapping, and debris detection.
• Danger alone is insufficient. A first use case should remove real exposure, be technically achievable, and fit the operating and regulatory environment.

Selection rule: prioritize exposure that a robot can actually remove

The model separates six considerations: worker risk exposure (30%), the share of exposure a robot could directly substitute (20%), technical feasibility (20%), operational value (15%), deployment and regulatory readiness (10%), and evidence strength (5%).

The composite score is not an accident probability or an objective truth. It is a transparent screening model. Every 1-5 judgment, weight, and calculation is available for review and recomputation.

Work zones and rail form the leading tier

Work-zone TTC scores 4.45 and rail track inspection scores 4.40. Work zones lead on risk, exposure substitution, and strategic continuity; rail leads on technology maturity and established operational value. Their proximity means partner access and data availability could still change the investment order.

The top five require different market-entry strategies

1. Work-zone TTC inspection and restoration: start with inspection, localization, and alerts; add manipulation only after perception is validated.
2. Rail track inspection: contribute change detection, anomaly explanation, and human-review workflows to the existing inspection ecosystem.
3. Roadside incident and towing response: begin with a protected remote first look, hazard mapping, and debris detection; vehicle hookup is a later stage.
4. Bridge and high-mast inspection: combine UAS and crawling platforms as supplements to qualified human inspection, not replacements for required tactile checks.
5. Tunnel and culvert inspection: validate communications, localization, and failure recovery under darkness, confinement, and short access windows.

The leading tier is stable under reasonable weight changes

Across 20,000 draws around the baseline weights, work-zone TTC appears in the top three 99.2% of the time and rail track inspection 95.5%. Roadside response and bridge inspection form a second competitive group whose ordering depends on whether a decision maker emphasizes maximum risk or immediate deployability.

Accident evidence supports removing people from moving-vehicle exposure

BLS reported 1,937 fatal occupational transportation incidents in 2024, representing 38.2% of all occupational fatalities. The cross-sector signal is exposure to moving vehicles and equipment.

At task level, NIOSH reported that 650 of 1,462 road-construction-site occupational fatalities in 2011-2022 involved a worker struck by a vehicle in a work zone. A separate NIOSH-supported study estimated a towing-industry fatality rate of 60.4 per 100,000 workers. These statistics use different denominators and periods, so they cannot be combined into one accident-rate ranking.

Recommended research and product path

1. Use work-zone robotic inspection as the flagship validation case. Build a 90-day evidence package covering public data, multi-condition CARLA experiments, quantitative change detection, and failure cases.
2. Run a narrow rail-inspection study in parallel. Test whether the same temporal change-detection methods transfer to fasteners, geometry, or right-of-way objects.
3. Treat roadside incident response as a high-risk exploration. Keep the first prototype to protected remote reconnaissance; do not claim automated towing or open-traffic manipulation.
4. Apply one evidence gate to every use case. Require official or real data, a reproducible baseline, failure taxonomy, deployment constraints, and a field partner before calling a study company experience.
5. Interview buyers before expanding hardware investment. Validate inspection frequency, downtime cost, liability boundaries, and procurement paths with DOTs, railroads, airports, and incident-response operators.

Questions still to answer

• Which partner can provide the fastest access to a closed site and real defect data?
• How much worker exposure time can each use case remove, beyond improving detection accuracy?
• Does the available budget sit with safety, inspection, maintenance, or IT, and who owns the purchase decision?
• Should work-zone manipulation follow a fixed or vehicle-mounted inspection product?
• Can consistently mapped CFOI research data support a true task-level rate comparison?

Caveats and assumptions

• National public data do not provide a common accidents-per-work-hour denominator for all 12 use cases; the model therefore scores risk, substitution, maturity, and evidence separately.
• The 1-5 values include research judgment. They support portfolio screening, not direct estimates of return on investment, fatalities prevented, or regulatory approval.
• Technology maturity often means assisting a qualified human inspector, not unattended autonomy. Bridge, airport, rail, and roadway operations retain explicit qualification and safety boundaries.
• Robots introduce collision, pinch, control-loss, and human-interaction hazards. High-risk deployments require separation, speed limits, emergency stops, audit logs, and progressive validation.