Skip to main content

Advanced Training Environment for Crisis Scenarios

Final Report Summary - PANDORA (Advanced Training Environment for Crisis Scenarios.)

Executive Summary:

The PANDORA project has developed a smart, novel digital support environment and crisis simulation system to enhance and expand training exercises for Gold Commanders in crisis management. Gold Commanders are specifically engaged in the development of strategic plans to deal with a wide range of potential crisis situations that can arise in civil society. These crisis situations could include:

• Natural events e.g. extreme weather.
• Transport events e.g. plane, train or vehicle crashes
• Service failures e.g. water supply failure
• Health crises e.g. pandemics
• Technology failures e.g. breakdown of automated control systems
• Policing and terrorism events

The role of individuals identified as Gold Commanders is explicitly strategic, rather than tactical (Silver) or operational (Bronze), although in practice some individuals may also have tactical or operational responsibility. In an actual crisis situation Gold Commanders from different areas, e.g. police, fire, health, will be expected to work together as a team and therefore need to be trained as a team to help develop their:

• Collaborative, Strategic thinking, Negotiation and Decision-making skills
• Flexibility and capability under pressure
• Ability to recognise and challenge assumptions, both their own and those of others in the team
• Ability to think of innovative solutions to problems when pre-prepared plans are no longer appropriate

Currently, the primary model of training offered to Gold Commanders consists of group-based, table-top activities led by an expert trainer. The bulk of the information provided to the trainees is paper-based, with some limited audio-visual input, and the activities take place during an intensive, time constrained training event. These can be seen as unrealistic, make it difficult to engage the Gold Commanders in the stress of a real crisis situation, and are almost entirely dependent on the ability of the trainer to engage and motivate the trainees, and to assess their performance subjectively in the training event. The PANDORA system addresses the shortcomings of the existing training model, enhances the range and scope of the training events, and offers the potential for future development by:

• Offering a fully-featured multimedia environment to provide information to the trainees, including audio, video, maps, texts, email, graphics and text
• Developing a crisis scenario model based on a structured, timeline-based, event network, running in a computer-based simulation environment controlled by the trainer
• Providing real-time operational inputs demonstrating strategic decision outcomes to trainees, asking them to dynamically revise strategic plans and decisions
• Capturing trainee behaviour and emotional state, through the use of pre-event information capture, direct sensor inputs, self-reporting by trainees, and trainer inputs, and using affective media to induce changes to those behavioural and emotional states
• Providing a graphical virtual representation of the training environment to support on-line distributed training events
• Providing virtual characters, in any form from textual through to full animation, to engage in the event, including replacements for missing trainees, to ensure the full scenario enactment is supported in all training events
• Providing the trainer with a full control system for the training event, including the ability to change events, add new events, expand and compress timelines, provide direct interventions into the scenario, and increase or decrease the emotional stress applied to individual trainees
• Maintaining a detailed log of the training event, to permit rerun of some or all events, modelling of individual trainee performance, and capture of relevant and useful events as exemplars for future training
• Maintaining configurable scenario models, knowledge, multimedia asset and data bases to enable the system to build a wide range of crisis scenarios, to use as training events for those involved in crisis management at all levels

Project Context and Objectives:

Context and Background

It is often human behaviour alone that prevents a disaster from becoming a crisis, however major crises are rare events and although those required to deal with them will have pre-prepared plans, they may have little or no experience of a real crisis situation or any understanding of how they might personally react in the event of a real crisis situation. Training is therefore extremely important in order to prepare people to take on leadership roles in a crisis situation. The vast majority of the training systems currently available focus on preparing people at the tactical (Silver) and operational (Bronze) levels, however it is the strategic (Gold) level personnel who make the key decisions and set the direction for managing a crisis, and, where necessary, make life and death decisions. Typical crisis situations will present complex problems to solve, and require a coordinated, multi-agency response. Examples of crises can include one or more of the following situations:

• Policing and terrorism events - fire or explosions
• Natural events - extreme weather, earthquake, landslides, storms, wind, heat-wave etc.
• Transport events – sea, air, train or vehicle crashes.
• Service failures - electrical power plant failure, water supply failure, etc.
• Health crises - pandemics, epidemics, containment conditions, disease (including animal disease), toxic chemical release impacting the environment etc.
• Technology failures - breakdown of automated control systems, central services

Gold Commanders are typically involved in a crisis when an event threatens health, life, service infrastructures etc. Their remit is quite wide and can include:

• Relieving suffering, saving and protecting life and promoting / facilitating self-help
• Containing the emergency, ensuring the health and safety of staff are protected
• Protecting property and safeguarding the environment
• Maintaining/restoring critical services
• Facilitating the investigation/inquiry and evaluating and identifying lessons learned.
• Providing the public with information as appropriate, including management of the media.

Strategic level commanders, henceforward referred to as Gold Commanders (a UK terminology), are typically trained in one of two ways. The first, and dominant approach, is through the use of table-top exercises. These are where a group of trainees (typically representing different emergency services e.g. fire, police, health) sit around a table, and are given a written description of an initial crisis scenario e.g. lorry carrying inflammable liquid has crashed into a tunnel wall, which caused a fire to break out and the loss of mains electricity to the local town. The group might be presented with the occasional audio or video clip of some news for example but fundamentally it is a paper-based exercise and it is down to the enthusiasm of the trainer to try and engage the trainees in the scenario and make it feel as realistic as possible. Another key problem with this approach is that management of a crisis scenario will often require that key emergency roles are represented in order to make a scenario realistic. However, if, for example, the Fire Chief was unable to attend the training session that day, then the trainer will often have to act out that role which is hard to do as they know the implications of the decisions but it also detracts from their role as a trainer, making the training session considerable more challenging to run. The second approach is a live training exercise where part of a scenario is realistically simulated. Whilst these will inevitably engage trainees in the emotion and stress of a real crisis situation, they are very expensive, time consuming to run and can only ever simulate part of a real crisis situation with a limited number of outcomes.

The key outcomes planned for training events designed for Gold Commanders are:

• Collaborative, Strategic thinking, Negotiation and Decision-making skills.
• Ability to work in teams, demonstrating flexibility and capability under pressure and to come up with solutions in a time constrained manner.
• Ability to understand the impact of their decisions on other emergency services.
• Ability to recognise and challenge assumptions, both their own and those of other team members, ensuring that they can provide a full justification after the event, as if something goes wrong during the management of the crisis assumptions will be a key focus of any investigation.
• Ability to think of innovative and alternative solutions to problems when pre-prepared plans are no longer appropriate.
• Ability to reflect on whether the decisions made during the course of the crisis management made the situation worse or better.
• Ability to deal with the media, which is inevitable in the event of a crisis.

From the trainer’s perspective, the outcomes will be:

• Monitoring the risk taking behaviour of the group, given research shows that appetite for risk tends to increase in a group situation.
• Assessing the capabilities of each individual and determining whether it would be appropriate to involve them in a real crisis situation.

Research tells us that many of the characteristics identified above and carried out by individuals in everyday life, e.g. the ability to consider alternative solutions before making a decision, may not occur in very stressful situations. For example when time is precious and lives are at stake it can be all too easy to implement the first solution thought of. A key part of the reason that training must be realistic is so that people can discover how they might really behave in a real crisis situation.

The context for the development of the Pandora system, named the Pandora Advanced Training Environment (PATE), is to bridge the gap between the table-top exercises and live exercises, providing an environment that can run a crisis scenario that feels realistic and can engage trainees at an emotional level, ensuring that they experience the stress of a real crisis situation.

The Pandora Training Environment

In the Pandora system, training takes place in a crisis training room which has two interfaces: a software client that can be downloaded and installed on a local workstation, and a 3D virtual room delivered through a web browser in which trainees are represented by avatars. Both of these can be used to deliver training in the following ways:

i) At a training site which would be typically used for training Gold Commanders. In this mode, the trainer and trainees are co-located in the same room. In the middle of the room there is a table around which the trainees will periodically meet, at the request of the chair (one of the trainees) to discuss strategy. However, for the majority of the training, the trainees are physically separated, working at their own workstations and using the Pandora system to communicate. In addition, several large screens may be used to display multimedia information to all the trainees such as sound, pictures, maps, animations, videos, to simulate receiving information about the crisis, such as news broadcasts.
ii) Same as i) however delivered at a non-specialised site such as the offices of one of the Gold Commanders (portable mode).
iii) The trainer and the trainees and are geographically dispersed and communicate only through the Pandora system (distributed mode).

A crisis scenario is described in terms of a sequence of events that occur as the crisis unfolds, e.g. event 1 = plane crashes into a dam, event 2 = flood in local town, event 3 = town loses electricity supply etc. All of these events are represented on a timeline within an event network, and can be delivered to the trainees through a variety of media, e.g. via text message, through a news broadcast, or through a multimedia clip.

Trainer Support Framework

The crisis room is managed through the Trainer Support Framework, which provides a number of facilities to setup and configure the system both prior to, and during a training session. For example: loading a scenario; the ability to interject new events into a scenario prior to execution or dynamically as the scenario is executing; rollback a scenario to a previous decision point, or jump forward to a future point; and speed up or slow down the execution of the scenario as it is running. The trainer is also able to configure the scenario to set up non-playing characters (NPC), for example to role play an emergency service not represented within the group of trainees (some scenarios require that certain roles are filled e.g. Chief of Police, if no physical trainee is available to undertake that role); subject matter experts; higher control (HICON), such as Government ministers; and lower control (LOCON) - representing the lower levels of command within the crisis team. The system can be configured to play the roles automatically, but the trainer can also take over and simulate responses by these NPCs at any time during the execution of the scenario if they so desire.

A record of every event, trainee decision, action, email, use of the chat facility etc. is logged and can be viewed or printed at the end of the training session to support an objective post-training analysis, reflection and feedback by both the trainer and the individual trainee.

Emotional Modelling and Use of Augmented Reality to create Affect

In order to impact the emotions of trainees and provide personalisation of a training plan, various psychological traits that have been shown to impact decision-making under stress are assessed prior to the training event. Examples of these psychological traits are background experience (known as a static trait, as this cannot change) and stress and anxiety (which are dynamic traits, which can vary during a training session). These traits are ascertained through a pre-training questionnaire, and are used in conjunction with goals set by the trainer for each individual trainee (e.g. trainee profile shows insecurity so the system needs to put them through a confidence building exercise) to adapt the training as the scenario unfolds.

During scenario execution, the system runs through a continuous cycle of planning, execution and re-planning. Throughout the training, the stress level of individual trainees is measured through a number of means such as self-reporting, physiological sensors (such as a heart rate monitor), and trainer observation. At some point the trainer might decide that one or more trainees appears too relaxed or too stressed and that the stress levels for one or more trainees should be increased or decreased, to better reflect a real world crisis scenario. This can be done through a variety of manual interventions such as compressing timelines thus forcing decisions to be taken faster or by interjecting new events into the system manually. More importantly, however, it can be addressed in an automated mode by the Pandora system.

Within the Pandora training environment, the relationship between affective state and learning is key, and in order to optimise that the project has to have techniques to determine the affective state of the trainee, described above, and make appropriate changes to the environment and scenario to address that state. Pandora uses a W3C standard on emotional modelling, to which the project has contributed and which is still under development, called EmotionML, to develop the Pandora affective model and mark-up affective content. Having determined the emotional and behavioural state of each trainee, the calculation of the desired state may be pre-defined within the training scenario, dynamically generated by the system, or as a result of direct trainer input. If the desired state differs from the existing state, the system has to plan suitable events to affect trainee emotions towards the target state. The information related to the existing and target states is passed to an internal rule-based mash-up engine, which can construct multimedia content representations, using affectively marked content, in a multi-channel, polyphonic, time-constrained model driven by a trajectory definition to move from an existing trainee state towards a desired state.

The key argument for the use of affect in the Pandora scenario is in the creation of affective ambience, by which we mean the use of multimedia assets and information management and manipulation to engender requisite levels of emotional impact on the trainees at the decision making points. Although there is some consideration of the use of avatar representations of emotion in the distributed mode of the Pandora Box, the key affective representation is associated with the creation of ambience or atmosphere. A number of techniques have been taken from film and TV production, combined with a number of known issues in information distribution and management, to create a set of rules by which multimedia mashups can be created and made available to the trainees across a set of configured media channels. This provides the potential to induce stress in the trainees through the use of vocative inflections, video and image representations of crisis situations, and textual updates from those situations. An example of this might be to show a picture of a hospital that is flooded, and overlay an appropriate piece of audio either saying that the water is receding, if the desired effect is to reduce the stress of a trainee, or that the water is rising and people are about to drown, if the desired effect is to increase the stress of a trainee.

However, the system can also induce stress through the corruption of media channels, incomplete information, missing information, noisy channels, and a variety of other techniques to diminish the certainty of information available to trainees at those points in the scenario where they are required to make decisions. Using this kind of induced stress the system can impact on trainees in terms of their self-efficacy, leadership capability, and existing personality traits, to determine their capability to make effective and appropriate decisions under stress.

Testing and Evaluation

The Pandora system was thoroughly tested over three days of user trials with 13 different Gold Commanders with a range of experience, at the UK Government Emergency Planning College. All trainees were completely immersed in the solving the crisis situation they had been presented with for the entire day of training. The feedback was overwhelmingly positive and, without exception, the trainees described the training experience as far more realistic than any other training models they had experienced. Similarly positive feedback was received from the trainers. There was also considerable enthusiasm from both trainees and trainers that the Pandora system be made available to them for practical use.


Unlike existing systems, the Pandora system has been developed both as an augmented and virtual reality environment, with a view to offering both blended learning and a fully virtual experience, where this is appropriate and desired. The system has the potential to encompass crisis situations of all types and perspectives and to support strategic level training in all of them. The Pandora box (PATE), being the set of tools and the methodological approach to configure the system and environment, have intentionally been developed to be generic rather than specific, and could therefore be applied to event-based scenarios in domains other than crisis management, e.g. inter alia stock portfolio planning or health service delivery planning.

Project Results:

Please see attachment.

Potential Impact:

Please see attachment.

List of Websites:


Contact details:

Dr Liz Bacon, Email: