Responding to disaster using autonomous systems

Responding to disaster using autonomous systems
9 June 2015 admin

We exist within a society where we often take for granted our heavy reliance on instant information through the use of pervasive technology. However, we have too often seen that there are still occasions where events can suddenly descend into chaos and disorder, through natural events such as floods and earthquakes, or through calculated terrorist attacks. In such instances, we may turn to technology for assistance in achieving resilience and the ability to provide security.

In the aftermath of a disaster, we are confronted with uncertainty and it is often the first responders that begin to systematically process and build a coherent picture of the event. In order to resolve emergency situations it is critical that the uncertainties that exist are quickly ameliorated. When an emergency alert is called, the more robust and effective the team is at collecting and analysing information, the greater potential they have for success in resolving the situation. The Human-Agent Mission Planning for Emergency Responders (HAMPER) project examines the use of autonomous agents in dealing with complex emergency situations and their effectiveness in assisting human responders to manage the issues that arise in dealing with such situations.

In a disaster, the ability to respond quickly will depend on a number of factors, the first and most critical of which is the successful establishment of an early response network in order to gather and process information. This network will allow the situation to be analysed effectively and will ensure that resources are directed to the problems that require them the most. The increasing availability of unmanned aerial vehicles (UAVs) is making this technology more attractive, as UAVs allow large areas to be searched more quickly, efficiently and cost-effectively when compared to equivalent, limited and more valuable manned assets, especially in safety-critical environments. For example, it was possible to use UAVs to monitor radiation levels following the Fukushima Daiichi incident. Similarly, unmanned ground vehicles (UGVs) have been deployed to identify and locate designated points of interest to a single operator who is remote from the area of operations and out of harm’s way. Several UGVs were used during the September 11 attacks on New York.

While the use of unmanned technology provides a degree of endurance and enhanced effectiveness, the introduction of autonomy to those platforms provides the human with a wider capability and flexibility in their response. The provision of intelligent software agents into that network can not only aid the human in terms of supporting critical decision making but also allow the human to delegate some of the more mundane tasks in order to free up their own cognitive resources to focus on the overall global objectives.

Unmanned systems should provide an emergency team with an extension of capability, in essence providing greater reach and quality of situation awareness. Unmanned systems provide significant advantages for allowing the human to see beyond their physical limitations by entering locations that would be hostile to humans, for example, structurally unsound or contaminated areas. The sensors that these unmanned platforms can carry extend the reach of the command team and can play a critical role in understanding the state of the situation.

Rather than individually directing each unmanned asset, the introduction of autonomy allows the operator to not only delegate tasks to individual assets under their control but also to designate tasks to a team of ‘agents’ with the power to then choose which asset within their own functional team is best matched to the task. This capability presents a newly emerging paradigm of a collective team made up of both humans and agents. Traditionally, agents have been designated to fill the gaps in capability and raise effectiveness of the human operators, but there is now the ability to seamlessly integrate software agents into the decision-making team. Ultimately, this integration should afford the end user a greater degree of flexibility in terms of freeing up not only cognitive resources, but also physical assets at their disposal.

The use of multiple unmanned autonomous systems (UAS) can be seen as a significant benefit when those systems are mixed with existing human teams. The use of unmanned systems, whereby tasks are delegated via the use of a well-defined autonomy framework, can provide the human commander with a degree of flexibility. ‘Dull, dirty and dangerous’ tasks can be delegated to systems as opposed to other people within the command team.

When examining an incident at a site where high risks are associated, such as a nuclear power plant or gas storage facility, the main goal of any team is the resolution of the incident to a state of stability that is parallel to that which existed before the accident occurred. Traditionally, well-rehearsed training procedures and events would ensure that in the event of a real life emergency, all components of the team would have clearly defined roles and responsibilities. In such circumstances, the effective use of all the team members, assets and resources is critical in achieving a successful outcome. The ability to replace a human team member with an unmanned asset increases the endurance and, in some instances, the capability of the team, whilst reducing risk to human life. However, people remain a critical factor in tactical awareness and dynamic decision-making in order to maintain the required flexibility to deal with a rapidly changing event such as a fire that has the potential to escalate seriously.

A critical aspect of the HAMPER project is the interaction between humans and agents, which allows the operator to command the unmanned systems to achieve a goal, rather than manually dropping waypoints and tasking individual assets to perform multiple missions. The operator, in this instance a Commander or Senior Incident Officer, generates a mission goal for the system of agents so that the system decides how best to achieve their goal via the use of both unmanned and manned elements. Within this framework it would therefore be possible for a Commander to define a top level goal, such as “search a location and report back by sending images”. With available human and agent assets within the team, the Commander has the ability to allocate multiple tasks to specific human or agent elements. This approach will have an associated cost in terms of the cognitive loading on the Commander. However, if the Commander is focused primarily on assessing the situation and gathering more information, then simply designating a goal to the human-agent collective will allow the human and agent to decide between them who is best placed to achieve the goal. In some cases, this goal may be shared between both the human and agent team members in terms of dividing tasks from an available library, or ‘task schema’, in order to allow different team members to perform tasks.

When human-agent interactive teams are commanded to carry out a task, the intent of the agent system is more likely to be interrogated by the human element, rather than the agent second-guessing what the human is doing. Therefore, the Commander of the team needs to possess a good understanding of what the agent is currently engaged in, and also have the ability to predict the autonomous agent’s intent.

The information that the unmanned agents offer will inform the Commander and wider team as to the collective progress towards achieving the goal, and in doing so provide the human team members with a degree of trust in the overall system. Within the HAMPER project, the human-machine interface for delegating such commands is therefore a key component in building the trust required.

The key to achieving mission effectiveness is to provide the decision-maker with good tactical situation awareness. In the midst of an emergency response, as events unfold and data is processed through the system, team members must not only make sense of what this data means but also be able to achieve a collective awareness of the reality of the situation and of how best to resolve or contain the consequences of the event using all the resources they have available. Therefore, we can see the importance that data and information have on breaking down the uncertainties that challenge our wellbeing, safety and security.

By Dale Richards, Alex Stedmon, Siraj Ahmed Shaikh & Dan Davies

Dr Dale Richards is a Senior Research Fellow in Human Factors
Dr Alex Stedmon is a Reader in Human Factors
Dr Siraj Ahmed Shaikh is a Reader in Cyber Security
Dale, Alex and Siraj all work in the Faculty of Engineering and Computing at Coventry University
Dan Davies is the Military and Security Lead at Serious Games International Ltd

This article first appeared in issue 534 of The Ergonomist, December 2014.