How to prepare a supervised automated-vehicle trial for a UK public road

A closed-track success is not a public-road release decision

A development team has completed simulation and controlled-track runs. The vehicle follows its planned lane, responds to a cyclist surrogate and stops when the communications link fails. The team proposes a public route next week. It has a driver with the right conventional licence and insurance for the prototype. Those are important conditions, but they do not prove that the proposed trial is ready. The organisation must show that the actual vehicle configuration, operating domain, takeover arrangement, road environment, service model and response plan have been examined together.

The Department for Transport's June 2026 automated-vehicle trialling code is the core primary guidance for this page. It says public trials can take place under existing law with a driver ready, able and willing to resume control, a roadworthy vehicle and appropriate insurance. It also sets out expected safety-case, engagement, data, driver and vehicle practices. The code is guidance, not a general AV trial permit. Failing to follow it may still matter in later liability proceedings, and following it does not guarantee immunity.

This page owns the go or no-go decision for the whole supervised trial. The separate safety-driver draft owns recruitment, training, monitoring and handover competence. A future safety-case page can examine argument structure. The no-safety-driver pilot guide owns the different government applicant route. Do not use this checklist to claim that removing the safety driver leaves the trial under the same legal path.

Confirm that this is the right route

Write one sentence that describes the planned activity without marketing terms: “We will trial a named automated function in a named vehicle on named public roads, with a trained safety driver monitoring and able to resume control.” If that sentence is false because the vehicle will move without a driver ready to take over, assess the no-safety-driver pilot and vehicle permission path instead. If passengers or freight will be carried, examine the existing service licensing regime as well. The trialling code says a service may need appropriate licensing even where the trial vehicle itself satisfies road law.

Do not assume a permit is needed merely because the technology is automated. The code states that ordinary public-road trials do not require a special AV trial permit or surety bond when conducted under existing law. Equally, that statement does not waive conventional rules on vehicle use, insurance, driver licensing, public-service operation or roadworthiness. Some prototype vehicles require a separate vehicle special order or exemption because their design does not meet ordinary requirements. The vehicle configuration determines that question. A legal reviewer should document the path rather than answer it from the project's title.

The 2026 code discusses the UK and notes relevant Northern Ireland differences. A programme crossing national boundaries should check the road and service rules that apply to each operation. The UK AV regulations guide compares the regimes. For the full Automated Vehicles Act authorisation scheme, the government implementation programme still describes development toward 2027. A present supervised trial should not be called an authorised full-framework deployment merely because the Act is enacted.

Define the trial as an evidence object

Before asking whether a vehicle is safe, specify what is being tested. Record the vehicle identifier, hardware build, software and map versions, automated functions, route, time of day, speed range, weather limits, road class, passenger or freight activity, safety-driver station and communications dependencies. Record which data will be collected and which supplier controls each system. The code expects a safety case proportionate to the trial and representative of its risks. A case about a different software release or a broad city area cannot quietly stand in for a narrow route that has changed.

Use a map that marks junctions, crossings, cycle facilities, bus stops, temporary works and locations where the vehicle may stop safely. Describe how the route will be inspected before each trial period. A planned limit of dry daylight operation should appear in the driver briefing and the launch decision. If rain begins or roadworks change the path, there must be an understandable rule for pausing, returning to manual control or ending the run. The map is not just an illustration for an authority meeting. It sets a boundary for the trial's evidence.

Agree measurable success and stop criteria in advance. Examples may include completing a defined number of safe runs without a particular failure, demonstrating a reliable intervention procedure, or investigating every unplanned disengagement. These are internal trial criteria, not statutory thresholds supplied by this article. Identify who can stop the trial immediately and who can authorise restart after a defect, collision, near miss or software change. An engineer focused on learning from failure and an operations lead focused on protecting road users may perceive the same event differently. The decision rule should settle that conflict before the vehicle enters traffic.

Build a safety case that matches the route

The code expects a detailed safety case before public trials. It points to the specific activity, vehicles, operating domain, safety-driver arrangements, training, management responsibilities, law, engagement and progress reporting. A useful case makes a claim, shows the evidence for it and identifies limits. “The vehicle has completed 10,000 kilometres” is weak if those kilometres did not include the junction type or lighting condition on the intended route. Record the scenario coverage and residual uncertainty, not only the total distance.

The safety case should connect simulation, bench, closed-track and public-road evidence. The code says software versions and revisions should be recorded and that extensive testing normally starts away from public roads. The transfer from a test track is a decision about the evidence gap. Ask which hazards remain untested, what additional protective control exists on the public road and what observation will trigger a review. If a low-speed shuttle is expected to encounter horse riders or children near a school, the case should address those road users rather than rely on motorway examples.

Keep an abridged version suitable for public release when appropriate. The code recommends making a public-facing safety case freely available and sending it to the Centre for Connected and Autonomous Vehicles. Protect sensitive security and personal information without removing the essential explanation of what is being trialled, where, how the driver supervises it and how the public can contact the organisation. A publication decision still needs legal and security review. The recommendation is not a blanket command to disclose proprietary engineering files.

Check vehicle legality and return-to-service control

A roadworthy vehicle is a core legal condition. The vehicle must meet applicable rules for the place and use. A prototype's unusual bodywork, controls, sensors or dimensions may raise questions that ordinary passenger-car paperwork cannot answer. The code discusses national in-service requirements, testing, exemptions and when a special order may be needed. A vehicle under a special order may have an MOT exemption in specified circumstances while still having to remain roadworthy. Do not turn an exemption into a statement that no inspection or maintenance is needed.

Create a configuration record that an independent reviewer can understand. Include registration or identification, approved modifications, software and firmware release, sensor calibration, tyres, braking, steering, emergency-stop mechanism and the safety driver's controls. The maintenance team should know which defect grounds a vehicle immediately and who signs the return-to-service decision. The trial team should compare the vehicle presented on launch day with the version covered by the safety case and insurance. A late sensor or control change deserves a fresh decision even if it seems like an improvement.

Insurance must fit the actual vehicle, location, people and use. The code says the existing statutory motor-insurance requirement applies to trials on public roads and places. A policy document should be read for prototype, automated mode, passenger or commercial operation, named drivers and any testing exclusions. This article cannot determine whether a particular policy is sufficient. Obtain written confirmation from the insurer or broker on the specified trial if the wording is unclear. Record who checked it and when it must be reviewed after a change.

Prove the safety-driver arrangement, then look beyond it

The safety driver needs the correct driving entitlement for the vehicle on a public road and must be able to monitor the road and system while ready to intervene. The code distinguishes these legal conditions from further recommendations on experience and training. A signed roster is not enough. The organisation should test takeover under the trial's vehicle configuration, at the speeds and conditions planned, and document how it will manage fatigue and distraction. Keep the driver's training and practical assessment with the launch record so that a reviewer can see which vehicle and system version were used.

The driver is part of a wider system. Who watches for a road closure? Who checks that the route remains within the tested domain? Who answers an emergency-service call, and who can tell the driver to stop? Communications may be useful, but a remote instruction should not be assumed to substitute for the driver's legal control of the vehicle. Plan what the vehicle and driver do if radio or data service fails. Test the failure during preparation rather than first discovering it at a busy crossing.

The trial should not place the driver in a role with conflicting duties, such as supervising an automated function while collecting detailed observations on a laptop or providing passenger assistance. A separate observer may help, subject to seating, distraction and vehicle rules. The safety-driver page should examine this human-factors detail. This readiness guide keeps the launch team focused on whether the combined arrangement is credible for the route.

Engage road and enforcement bodies before the launch date

The code says trial organisations should inform the Centre for Connected and Autonomous Vehicles before public trials. It recommends early engagement with highways and local authorities, police, landowners and other relevant bodies. The specific list depends on the route and service. A trial that affects a bus stop, a pedestrianised area or a school crossing may need a different conversation from one on an industrial estate road. Record whom the team approached, what they said and how the route or controls changed.

Engagement is more useful than merely sending a generic deck. Provide the area, dates, vehicle dimensions, stopping behaviour, contact point, safety-driver arrangement and incident plan. Ask the highway authority about planned roadworks or traffic changes. Ask the police and emergency responders how they would identify and disable the vehicle after an incident, who can provide technical information, and how data can be preserved. Infrastructure changes such as signs or parking adjustments need agreement with the responsible road authority. Do not assume that an informal conversation gives permission to alter the road.

Keep one public contact route during the trial. Residents and other road users may see unfamiliar vehicle behaviour and need a way to report concerns. A public explanation can describe the trial limits and what the safety driver does without promising that the vehicle is already authorised as self-driving under the future framework. The code encourages communication with vulnerable road users, including people with reduced mobility, cyclists and pedestrians. Where feedback identifies a new hazard, feed it back into the safety case and route decision.

Prepare data, incidents and a controlled pause

The code expects trial vehicles to record information that can show who or what controlled them and to preserve data after an incident. It recommends data on system configuration, mode, speed, steering, braking, location, connectivity and driver intervention. The exact hardware and sampling design require technical and privacy review. A procurement claim that Complys itself captures vehicle sensor data would be unsupported. The programme needs a reliable primary recorder and a way to connect its files to the incident and vehicle configuration record.

Rehearse a stopped vehicle, a collision and an unplanned driver takeover. The immediate response should protect people and the road before anyone tries to finish a test run. The plan should name the person who contacts emergency services, secures the scene, preserves vehicle data, informs the insurer, assesses reportability and suspends operations. The code recommends planning with authorities and maintaining access to incident data in a forensically sound manner. Reporting duties depend on the event and the applicable law or permission. Do not copy a fixed 24-hour pilot reporting number into every supervised trial.

Personal data may appear in video, audio, location and passenger records. The code points to data-protection duties and recommends a privacy assessment for trials. Decide what is captured, who can access it, how long it is needed and how to answer an investigation request without exposing unrelated people unnecessarily. A security incident and a road collision may require different specialists and response paths. The launch review should identify both rather than leave a vague instruction to “call IT”.

Hold a launch review with explicit outcomes

The final readiness meeting should include someone accountable for the trial, vehicle engineering, safety, the safety driver, road operations, insurance and legal or regulatory issues. A chair should ask for source evidence, not only verbal reassurance. Show the exact vehicle configuration, route map, safety case, driver authorisation, insurance confirmation, roadworthiness record, authority engagement, incident plan and data access arrangement. Record open issues with an owner and due date. A missing critical condition should produce a no-go decision even if the public launch has been announced.

Use three possible outcomes: go within the documented boundary, go only after named actions are closed, or no go pending rework. This is an internal decision framework, not a government certificate. The decision log should say who approved the run, which version of each document was considered and what conditions apply. A conditional go should never be treated as permission to start before the condition is actually met. The trial manager should confirm closure with evidence and keep a record of the final release.

Review the decision after a material change. A new route, different vehicle, updated automated-driving software, altered weather limit, new driver, insurer condition or stakeholder objection can invalidate part of the earlier case. The change owner should identify affected evidence and obtain fresh approval before operating outside the previous boundary. This is how a trial remains a controlled experiment rather than a one-time launch ceremony.

Where Complys could help

Complys can be considered for organising the trial's permission, safety, training, vehicle, supplier and incident evidence. This is a proposed workflow, not a verified claim that the current product contains a dedicated AV trial release module. The team should request a demonstration that starts with one route, one vehicle and one incident, then introduces a changed software version. Ask whether the system can preserve previous evidence, assign a review owner and export what an authority or insurer asks to see. Product-owner confirmation is needed before any specific feature claim appears in published copy.

The AV software pillar contains a buyer's demonstration script. The free AV checker gives a nonbinding route pointer, but its existing result copy has an integration handoff for legal and house-style corrections. Neither tool grants a trial permission. A launch team should begin with the trialling code and the particular road, vehicle and service law, then use software only to control the evidence and review process it can actually demonstrate.