UAV search and rescue
Our analysis of common decision-making models used by UAV-supported search and rescue crews shows the relative strengths of each approach. Our recommendations will also help crews make more effective decisions as improved UAV technology becomes available.
Project team: Sophie Hart, Vicky Steane, Seth Bullock and Jan Noyes
Teams in action
Search and rescue (SAR) teams rely upon strong partnerships. Uncrewed aerial vehicles (UAVs) have transformed SAR in many situations, but clear decisions during SAR operations require good communication between the human crew and the UAV, and between the pilot and payload operator, who analyses UAV sensor outputs.
There is a tendency during the design process to focus on the technical requirements of a UAV system with little consideration for the human operator interacting with these complex devices, and how the technology and new information provided may affect human decision-making.
Our work addresses this gap in understanding. We have used decision models to understand the information and decision-making processes used by traditional and UAV-supported SAR teams, and shown how this learning can help design future UAV systems that promote reliable decision making in human-UAV SAR teams.
We interviewed five SAR responders to detail their decision-making processes when operating in a remote wilderness environment with and without a UAV.
The trends observed in the interviews were applied to three decision models commonly used where human factors are present. This enabled us to analyse each of the approaches, capture the human factors in decision making, and consider simple changes that would result in more robust decisions.
The key features and relative strengths of each decision model are shown as follows, together with our recommendations for their future use.
Model 1: Recognition Primed Decision Model (RPDM)
The payload operator makes decisions based on live information from UAV sensors that is displayed on the user interface.
The RPDM highlighted the importance of the payload operator understanding the search area terrain and judging the likelihood of apparent search sightings being correct. This places a high emphasis on cognitive processes of an expert payload operator.
The RPDM is most suitable for short incidents and less appropriate over an extended time window with a changing environment, as often happens with SAR operations.
Recommendations
The payload operator should be assisted with technology to reduce cognitive load, e.g., augmented displays that show sighting locations.
Model 2: Perceptual Cycle Model (PCM)
The PCM suggests that decisions are based on a knowledge map of the SAR operation that is continuously updated with live information. This creates an objective operational view that allows justifiable decision making and reduces the degree of immediate individual judgement compared to using the RPDM.
The iteration of action and resampling of evidence allows all evidence to be used, including the absence of the search target.
The PCM demonstrated the need for efficient teaming between the UAV pilot and payload operator, which highlights the importance of training and considering how this partnership might be affects by systems design changes.
This highly adaptive approach is well suited to live and dynamic situations.
Recommendations
The payload operator should be guided to the most appropriate information for making decisions, e.g., recommending the most suitable UAV sensor based on different lighting levels.
Model 3: Decision Ladder
Decisions are based on integrating information from UAVs and other disparate resources, such as the location of other teammates.
The Decision Ladder always considers a complete range of possible actions following a sighting.
The Decision Ladder can be used to forecast the impact of novel technology on the decision-making processes of human operators.
Recommendations
The level of visualisation support and automation should be increased to further reduce the reliance on operator experience, for example, use of terrain databases.
The future of UAV-supported SAR
Our work finds that using the PCM (model 2) and Decision Ladder (model 3) in tandem provides a holistic understanding of an end-user’s decision-making process. System designers should apply these models at the earliest stage of the design lifecycle to identify interface design recommendations and training requirements.
Improved technology promises to help search teams but must not interfere with communications between pilot and payload operator. Our process analysis can help assess how planned new UAV technologies might affect decision making during SAR operations.
SAR teams have always relied on strong partnerships, teamwork, and communication to achieve their aims. Our work formalises decision-making processes to understand how best to design new technology in a way that helps UAV-supported SAR teams.