The AI landscape doesn't move in one direction — it lurches. Some techniques leap from experiment to table stakes in a single quarter; others stall against regulatory walls, technical ceilings, or organisational inertia that no amount of hype can dislodge. Knowing which is which is the hard part. The State of Play cuts through the noise with a rigorously maintained index of AI techniques across every major business domain — classified by maturity, evidenced by real-world adoption, and updated daily so you always know where you stand relative to the field. Stop guessing. Start knowing.
A daily newsletter distilling the past two weeks of movement in a domain or two — delivered to your inbox while the index updates in the background.
Each dot marks the weighted maturity of practices within a domain — hover for a brief summary, click for more detail
AI that automates train operation and optimises rail network scheduling and capacity allocation. Includes GoA 3/4 automated train control and predictive schedule optimisation; distinct from rail infrastructure inspection which monitors physical assets rather than operating services. Scope begins at AI-based ATO optimisation and ML-driven predictive scheduling; deterministic GoA-rated automation without ML components is out of scope.
Autonomous rail operations sits at the boundary between proven metro automation and the much harder problem of open-network autonomy. GoA4 driverless trains run in production on closed urban systems — Riyadh, Lyon, Delhi — and a procurement pipeline worth over 3 billion euros is confirmed across European metros through 2030 (Copenhagen, Berlin, Paris, Dublin, Lausanne). That success has not translated to freight or regional rail, where obstacle detection on open track, fragmented ERTMS infrastructure, and regulatory resistance remain unresolved. AI-driven scheduling optimisation is further along operationally, with systems like Deutsche Bahn's dispatching algorithms and Singapore's SMRT Overwatch delivering measurable delay reductions. The defining tension is structural: metropolitan deployments keep advancing because they operate in controlled environments with dedicated infrastructure investment, while the broader network — where the largest capacity and efficiency gains sit — faces safety certification gaps, 17-year average infrastructure project delays, active labour opposition, and emerging cybersecurity vulnerabilities in signalling infrastructure. This remains a bleeding-edge practice: real deployments exist and are accelerating globally, but they remain concentrated in controlled metro environments. Freight and regional autonomy remain experimental despite clear potential.
Metropolitan GoA4 procurement accelerates globally with major Q2 2026 contract announcements. Siemens-Stadler's confirmed 3 billion euro contract for 226 driverless trains on Copenhagen's S-Bane targets a 35% capacity increase by 2030, with phased transition to unattended operations in early 2030s. Dublin MetroLink's M500 DBFOM procurement explicitly specifies GoA4 fully automated signalling and 25-year O&M, adding to Paris Line 13, Berlin U5/U8, and Lausanne m2 contracts. Alstom is upgrading Lausanne m2 with 295 million euros in CBTC investment, while FS Group validated GoA4 integration on ETCS Level 2 infrastructure in Bologna test ring (April 2026), confirming multi-technology compatibility. India's adoption pipeline is widening: Lucknow Metro tendered 15 advanced driverless train sets with AI-based predictive maintenance and CBTC for East-West Corridor; combined with Delhi's two operational autonomous corridors, this validates bleeding-edge adoption across three continents. China's urban rail market demonstrates ecosystem maturity: 69 fully automated metro/light-rail lines totaling 1,983.84 km deployed across 25 cities by end-2025, with 33.383 billion passenger trips.
Mainline and regional autonomy test coverage broadens. Siemens' Nordic demonstration on Finland's 19 km mainline achieved 30-80 cm stopping precision with ETCS Level 2 and confirmed 2029 commercial rollout on Tampere-Pori. Poland's PKP state railway continues GoA2 testing; Intramotev's captive-industrial autonomous trains accumulate operational hours on Pennsylvania mining routes. Freight scheduling automation shows concrete returns: UK Catapult's prototype demonstrated 9.2% fleet reduction potential on real Q4 2025 operational data. However, infrastructure barriers remain stubbornly entrenched. The April 2026 EU ERTMS coordinator report to the European Parliament confirms only 50% core network and 40% of trains will be equipped by 2030, with fragmented standards and cross-border gaps blocking autonomous scaling. Critical infrastructure vulnerabilities emerged in May 2026: a Taiwan high-speed rail security incident exposed that digitalized signalling systems rely on static, unencrypted parameters enabling signal spoofing via commodity SDR hardware—highlighting a fundamental risk for autonomous operations that depends on these digital systems.
Governance frameworks solidify alongside deployment and security gaps. The UK Office of Rail and Road published formal safety principles for GoA4 unattended train operations in March 2026. Yet labour and regulatory opposition persists globally. In the U.S., freight autonomy remains blocked by union objections and FRA certification gaps; labour unions (SMART-TD) documented April 2026 that FRA regulatory push favors autonomous technology deployment while vendor dispatch systems show documented reliability failures. Indonesia's parliamentary transport commission acknowledged (May 2026) that current systems operate at GoA0 (fully manual) and lack ATP/ATO/ATC—a critical gap, though the expert explicitly validated that higher automation levels are safer. Real-world incidents — January 2026 Spanish derailments, control-centre software failures, cybersecurity vulnerabilities in rail signalling — underscore that autonomous operations require integrated socio-technical systems and robust infrastructure quality, not algorithmic maturity alone. Metropolitan success masks a widening competence gap: where governance and infrastructure investment exists (Europe, Asia), deployments advance; where fragmented ownership, legacy networks, and cybersecurity gaps dominate (freight corridors, regional lines), autonomy remains experimental.
— Siemens-Stadler consortium delivering 226 GoA4 driverless trains for 170 km Copenhagen S-Bane with CBTC, obstacle detection, regenerative braking; phased transition to unattended operations targets early 2030s with 35% capacity increase.
— Uttar Pradesh Metro procuring 15 advanced driverless train sets with UTO mode, CBTC signalling, AI-based predictive maintenance, and sensor-based safety systems; demonstrates India's adoption of full GoA4 automation on new metro infrastructure.
— Taiwan HSR experienced 48-minute disruption after student spoofed TETRA signals using commodity SDR hardware; 19-year static parameters enabled exploitation—critical vulnerability in digitalized rail infrastructure that autonomous operations depend on.
— Dublin MetroLink procurement explicitly specifies Grade of Automation 4 (GoA4) fully automated signalling system with CBTC control, rolling stock, and 25-year O&M—major European metro commitment to unattended autonomous operations.
— 295 million euro contract to deploy Alstom Urbalis Fluence CBTC on operational Lausanne m2 metro, enabling closer train spacing and higher capacity through train-centric architecture with onboard intelligence.
— Indonesian parliament expert documents current GoA0 (fully manual) automation levels and lack of ATP/ATO/ATC deployment as critical safety gap; acknowledges automation as proven safety solution not yet deployed, validating adoption barriers in legacy rail networks.
— FS Group demonstrated integrated GoA4 (automatic driving, remote operation, obstacle detection) on ETCS Level 2 infrastructure at San Donato 5.7 km test ring, validating multi-technology interoperability for autonomous mainline operations.
— London Underground Four Lines Modernisation deploys CBTC with ATO/ATS/ATP across 150 km, 200+ trains, 100+ stations; detailed RF architecture for safety-critical operations on dense heritage infrastructure validates large-scale metro CBTC deployment.
2019: ERTMS infrastructure deployment accelerates across Europe with commitment announcements; ATO research demonstrates quantified benefits (17-45% energy savings) but commercial deployments remain limited to niche pilots; safety incidents highlight certification and design challenges; Alstom and SNCF advance autonomous freight train retrofit toward 2023 trials.
2020: Rio Tinto AutoHaul reaches full-scale production with 98% fleet autonomous; EU standardization advances (X2Rail-4 integrates GoA2 into CCS TSI 2023); vendor tests announced for 2021 (Germany passenger, Netherlands shunting); infrastructure defects and signalling failures demonstrate safety gaps; research progresses on ML scheduling and AI perception but open-network deployment remains constrained.
2021: Deutsche Bahn/Siemens deploy first fully autonomous urban passenger trains on Hamburg S-Bahn (December 2021, 4 units, 30% efficiency gains); Fraunhofer BerDiBa consortium launches €13.7M real-world testing program with 12 partners; Kelana Jaya LRT collision (213 injuries) exposes safety risks in mixed automated/manual operations; U.S. FRA resists automation approvals; retrofit cost estimates exceed £7 billion with 7-9 year timelines, constraining adoption despite technical proof-of-concept.
2022-H1: EU Horizon Europe program launches structured R&D (TRL 5/6 targets by 2025) for autonomous operations; Alstom advances German regional train pilot (€5.5M government funding) with real-world signal recognition and obstacle detection testing; York University/Thales/Lumibird complete successful autonomous perception trials on 16 km track; RL-based scheduling algorithms demonstrate rapid disruption recovery optimization; U.S. FRA and labor unions intensify regulatory and safety opposition, citing compliance failures and job threats.
2022-H2: Autonomous rail market grew to $8.3B with vendors continuing development; Alstom demonstrated GoA4 shunting autonomy in Netherlands with operational partners; however, persistent challenges dominated: Transportation Safety Board recommended expedited physical fail-safes after freight accidents, Kelana Jaya LRT suffered ATC reliability failures mid-service, research confirmed obstacle detection remained at TRL 5 (unchanged for 20 years), and WMATA began return-to-automation planning after 12-year hiatus following the deadly 2009 crash — illustrating that technical demonstrations coexist with entrenched safety, reliability, and regulatory barriers.
2023-H1: Sensors4Rail R&D project concluded with 500+ hours of testing and 450 TB of multisensor perception data from Hamburg S-Bahn. Quantinuum and Deutsche Bahn demonstrated 17% improvement in rail rescheduling using quantum algorithms. Siemens launched safe.trAIn for AI/ML safety assurance in driverless trains. However, critical infrastructure barriers intensified: ERTMS deployment faced €3.9B in spending yet remained incomplete, and Norway's rollout slipped by one year due to supply-chain and testing issues.
2023-H2: Lyon metro Line B achieved full GoA4 automation with 36 operational trains and extension serving 25,000+ daily passengers, confirming large-scale mainline deployment viability. German AutomatedTrain project received €42.6M funding for fully automated dispatch and parking development with ETCS integration (prototypes by 2026). Safety assurance methodologies advanced with scenario-based validation research and SNCF's structured ATO safety analysis. However, ERTMS remained critical bottleneck: only 13% deployment on Orient/East-Med corridor with major delays in Germany (2025-2028) and Greece; peer-reviewed analysis emphasized AI regulatory frameworks remained inadequate for industry-wide adoption. Sector showed technical momentum and growing commercial confidence alongside persistent infrastructure and governance barriers.
2024-Q1: Safety methodology research accelerated with University of York's SACRED framework (March 2024) for GoA4 systems validation informed by Berlin S-Bahn plans. Siemens announced major deployment contract for Copenhagen S-bane GoA4 upgrade (170 km, 350k daily commuters, 84 trains/hour capacity, operational by 2030). Signaling X cloud platform launched with ATO integration and 30% claimed energy savings. Stuttgart Digital Node and AutomatedTrain project advanced prototypes toward 2026/2030 operational milestones. Japan's RTRI demonstrated autonomous operation with obstacle detection. However, U.S. union opposition to FRA-waived freight autonomy tests intensified (March 2024), citing safety concerns over sensor capability and crew absence; ERTMS infrastructure remained critical bottleneck with deployment stalled. Regulatory frameworks for AI/ML in safety-critical systems remained inadequate, persisting as governance barrier.
2024-Q2: German government published five-stage automation roadmap confirming GoA2 feasibility and staged deployment by 2026; AutomatedTrain project advanced obstacle detection testing on two vehicle types. TMH (Russia) scheduled mainline GoA3 by 2026; Intramotev's ReVolt autonomous railcars accumulated 1,000+ miles on Pennsylvania mining route. However, U.S. unions filed formal safety objections (April) to FRA on autonomous freight tests, citing unresolved sensor/stopping-distance gaps; Rio Tinto's AutoHaul system experienced two derailments in Australia (Feb/May), exposing operational reliability risks; ERTMS infrastructure remained incomplete with cost overruns unresolved.
2024-Q3: Siemens Mobility secured €200M Berlin U5/U8 metro contract and major Copenhagen S-bane GoA4 upgrade contract (170 km, 350k daily commuters, operational 2030). InnoTrans 2024 saw major European stakeholders renew ERTMS commitment with focus on cost/migration strategies; EU confirmed mandatory ERTMS deployment deadlines (2030/2040/2050). However, critical gaps persisted: peer-reviewed research identified AI scheduling methods still inadequate for dynamic autonomous dispatch; ERTMS remained bottleneck (only 13% deployed on Orient/East-Med corridor); Rio Tinto derailments (Feb/May) and industry survey highlighted data quality and implementation barriers as top adoption challenges.
2024-Q4: Major production deployments affirmed sector maturity: Siemens Riyadh Metro (67 trains, 64 km, full GoA4 launched December 2024) and DC Metrorail Red Line ATO reactivation (December 2024 after 15-year hiatus); CAF completed 40,000 km ATO endurance testing and autonomous shunting demonstrations with Nederlandse Spoorwegen; Hitachi's Florence autonomous tram won InnoTrans 2024 award for two-year field deployment. Scheduling optimization advanced through Siemens Optrail acquisition (real-time automated dispatching, operationally deployed). Yet unresolved barriers persisted: peer-reviewed ERTMS research documented network variations and interoperability challenges; Rio Tinto autonomous freight derailments continued; labor opposition to FRA freight autonomy waivers remained; industry data sourcing and implementation barriers unresolved. By Q4 2024, large-scale mainline GoA4 deployments (Riyadh, Berlin, Copenhagen) and U.S. urban reactivation signaled commercial confidence alongside persistent infrastructure, safety assurance, and labor-regulatory barriers limiting adoption pace.
2025-Q1: Siemens Mobility and S-Bahn Berlin advanced obstacle detection testing on ten trains for full autonomous driving capability (March 2025); industry associations formally documented ERTMS infrastructure crisis with only 15% European deployment in fragmented state, blocking autonomous operations scaling; Europe's Rail research identified organizational barriers as critical adoption constraint; AutomatedTrain project confirmed 2026 prototype milestones for autonomous regional trains; market growth continued (USD 9.2B in 2025) but gap widened between deployment progress and systemic barriers (infrastructure chaos, organizational complexity, regulatory resistance).
2025-Q2: Scheduling optimization algorithms advanced with peer-reviewed research (PLOS ONE) demonstrating AI/DRL methods outperforming traditional ATO on Beijing Metro data; Denmark's €2.7B ERTMS rollout achieved 96.4% on-time performance on 2,600 km, validating infrastructure-led autonomy enablement; German safety standards (DIN DKE SPEC 99002, 99004) published for AI in autonomous rail; distributed autonomous agent research (TU Delft) advanced real-time timetable rescheduling. Market sizing: USD 10B in 2025 (8% CAGR to 2034). However, U.S. freight autonomy heavily contested — labor unions opposed FRA permit for autonomous pilot, citing aging infrastructure and cybersecurity vulnerabilities; gap widened between metropolitan automation success and freight/regional regulatory deadlock.
2025-Q3: Metropolitan rail automation expanded with Siemens' Paris Metro Line 13 contract (24 km, 550K passengers, GoA4 by 2032) joining Copenhagen and Berlin pipeline (€500M+ through 2032). AutomatedTrain project advanced toward 2026 prototypes with digital twin training achieving 20x testing efficiency gains. Fraunhofer safe.trAIn completed DIN DKE SPEC 99002/99004 standards for AI safety in autonomous rail. However, infrastructure complexity barriers persisted: Dutch ERTMS programme reported deployment delays due to longer component development; adoption survey showed 63% beyond pilot phase but with persistent data readiness, legacy integration, and ROI barriers documented by practitioners. Regulatory gridlock continued in U.S. freight autonomy. Gap widened between metropolitan success and freight/regional obstacles.
2025-Q4: Freight and regional rail testing accelerated with ProRail launching one-year ATO evaluation on Betuweroute (October 2025) targeting 20-30% capacity and 10-15% energy gains, and Parallel Systems receiving FRA approval for battery-powered autonomous freight train testing in Georgia (May 2025). Research methodology advanced: RWTH Aachen developed AI obstacle detection for regional driverless services addressing staff shortages (December 2025); peer-reviewed study demonstrated feasibility of ETCS integration with ATO via RAMS analysis. Infrastructure modernization progressed: European transition from GSM-R to 5G FRMCS for safety-critical communications began live testing (December 2024, funded €13.5M) with FRMCS v3 targeted for 2028 deployment on 130,000+ km. Europe's Rail research characterized GoA2-4 use cases for regional lines emphasizing cost reduction. However, same structural barriers persisted: no resolution of U.S. labor/regulatory gridlock on freight autonomy; ERTMS deployment remained incomplete with cost overruns; implementation barriers (data readiness, legacy integration, ROI uncertainties) continued constraining scale. By end-2025, deployment activity (Paris Line 13, Berlin U5/U8, Copenhagen S-bane) remained concentrated in metropolitan systems with strong governance and infrastructure commitment; freight and regional scaling faced unresolved regulatory, technical, and organizational barriers.
2026-Jan: Depot and remote operations advanced with RemODtrAIn project (Siemens/Rheinmetall, €17M funding) equipping ICE 4 trains with AI-based obstacle detection and 5G remote control from Deutsche Bahn depot (January 2026). Autonomous freight rail market expanded beyond Europe with three independent U.S. startups (Parallel Systems, Railspire, Intramotev) conducting pilots in ~$200B market with mainline, yard, and captive-industrial segmentation. Safety governance matured: EU AI Act classification framework for ATO systems formalized regulatory expectations; DIN standards established AI terminology and ODD taxonomy. Critical infrastructure barriers persisted despite governance advances: ECA 2026 audit documented 17-year average delays on transnational rail projects with massive cost escalations (Rail Baltica +291%, Lyon-Turin +127%), blocking autonomous operations scaling. Real-world safety incidents (January 2026 Spanish derailment and software failures) demonstrated autonomous operations require integrated socio-technical systems—not just algorithmic control—with robust infrastructure, software quality, and regulatory oversight. Metropolitan automation maintained momentum (Paris Line 13, Copenhagen S-bane, Berlin U5/U8) while freight and regional scaling faced unresolved regulatory, infrastructure, and organizational barriers.
2026-Feb: Metropolitan automation deployment solidified with Siemens-Stadler €3 billion contract for 226 GoA4 trains on Copenhagen S-Bane (February 2026), targeting 35% capacity increase by 2030. Infrastructure modernization progressed: EU Third ERTMS Work Plan (February 2026) documented 10% TEN-T deployment (12,400 km) and 19% fleet equipping by end-2024, flagging slow progress requiring accelerated industrial-scale rollout; Singapore LTA announced accelerated signalling, power, and train system renewals with fault-bypass feasibility study. R&D infrastructure advanced: Europe's Rail R2DATO Data Factory delivered operational synthetic ML training data generation (February 2026); AutomatedTrain progressed toward 2026 prototypes. Scheduling optimization demonstrated operational impact: Singapore SMRT Overwatch reduced delays 30%; Deutsche Bahn AI dispatching compensated for 8-minute delays (February 2026). Systemic barriers persisted unresolved: ERTMS deployment remained at 10-13% despite critical importance; January 2026 Spanish rail safety incidents underscored socio-technical system requirements; U.S. freight regulatory gridlock continued; data readiness remained adoption constraint.
2026-Apr: Infrastructure gaps sharpened as the EU ERTMS coordinator reported to the European Parliament that only 50% of the core network and 40% of trains will be equipped by 2030, citing fragmented standards and cross-border interoperability as the binding barriers. Governance matured at the project level: the UK Office of Rail and Road published formal safety principles for GoA4 unattended passenger train operations, and Hitachi Rail secured a €481.6M contract for GoA4 driverless trains on Turin Metro Line 2, extending the European procurement pipeline. Siemens' Nordic ATO trial on Finland's mainline achieved 30-80 cm stopping accuracy under ETCS Level 2 and confirmed a 2029 commercial rollout, while a UK Catapult freight-planning prototype demonstrated 9.2% fleet reduction potential using AI optimizers on real operating data—concrete proof points for open-network benefits that the infrastructure deficit continues to delay at scale. Additional April signals: DC Metro's board unanimously approved a $913M project to fully automate the 32-mile Red Line, the largest single ATO modernization commitment in the U.S.; the Dutch ERTMS Programme Directorate warned its rollout risks missing the 2050 target by 10-20 years, deepening the infrastructure gap across Europe; Europe's Rail R2DATO project published 156 system requirements for remote-driving automation across 26 use cases; and U.S. labor unions backed bipartisan legislation to hold rail technology vendors accountable for documented dispatch-system reliability failures, formalizing regulatory pressure on autonomous operations quality.
2026-May: Metropolitan GoA4 procurement pipeline extended with major new commitments: Dublin MetroLink tendered a 25-year GoA4 DBFOM contract, Alstom signed a €295M CBTC upgrade for Lausanne m2, India's Lucknow Metro tendered 15 driverless train sets with AI-based predictive maintenance and CBTC, and China documented 69 fully automated lines totaling 1,984 km across 25 cities. FS Group validated GoA4 integration on ETCS Level 2 infrastructure at a Bologna test ring, confirming multi-technology interoperability for mainline autonomy. A critical cybersecurity incident emerged: a student spoofed Taiwan HSR TETRA signals using commodity SDR hardware, causing a 48-minute service disruption and exposing static, unencrypted signalling parameters as a systemic vulnerability in digitalized rail infrastructure that autonomous operations depend on.