Supporting Wilderness Search and Rescue with Integrated Intelligence: Autonomy and Information at the Right Time and the Right Place

Lanny Lin, Michael Roscheck, Michael A. Goodrich, Bryan S. MorseAAAI 2010 · DOI

Wilderness Search and Rescue (WiSAR) is a high-stakes, time-critical domain involving diverse personnel, from mission planners to ground search teams. While small Unmanned Aerial Vehicles (UAVs) offer a low-cost method for obtaining aerial perspectives, their introduction often creates a "cognitive tax" on searchers. This paper addresses the integration challenge: how to deploy autonomy so that it handles the drudgery of flight and data processing, allowing human intelligence to remain focused on high-level decision-making and clue detection.

The Integrated Intelligence Framework

The authors propose a taxonomy for "Integrated Intelligence" that balances the capabilities of individual tools with the needs of a distributed system. This framework is organized across three primary attributes:

  1. Capability: Moving from basic flight to high-level autonomy and modularity.
  2. Information Management: Transitioning from raw data feeds to flexible, fused communication streams.
  3. Performance Evaluation: Shifting focus from individual tool goals to collective progress and search quality.

Advanced Autonomy and Path Planning

To reduce the operator's workload, the system employs a Bayesian model to generate a "probability of area" map. This model incorporates terrain data (topography, vegetation, elevation) and historical lost-person behavior to predict where a subject is most likely to be. Intelligent path-planning algorithms then automatically guide the UAV to maximize the probability of detection within the search area.

Probability distributions based on terrain and historical data, alongside an autonomous search path.

Beyond flight paths, the system introduces the concept of "See-ability." This metric quantifies search quality by calculating how well each point on the ground was observed by the camera, accounting for distance, viewing angle, and sensor resolution. This allows mission managers to determine if an area has been truly "cleared" based on data rather than intuition.

Human-Centric Interface Design

A core insight of the paper is that the person flying the drone and the person analyzing the video have distinct needs. The system splits these roles into specialized interfaces:

  • Phairwell: A "fly-by-mouse" interface for the operator. It allows for seamless switching between manual control and autonomous patterns, such as loitering or following terrain contours, without requiring expert piloting skills.
  • Wonder Client: A dedicated tool for video analysts to annotate clues and for mission managers to track the overall progress of the distributed team.
The Phairwell interface, featuring the 'fly-by-mouse' controls, UAV status, and integrated video feed.

To further assist analysts, the system utilizes video mosaicing. By stitching video frames together in real-time, the system creates a stable, wide-view image that is significantly easier to parse for clues than a shaky, narrow-angle raw feed.

Field Results and Evaluation

The system was validated during a live field exercise in Elberta, Utah, involving professional searchers from Utah County Search and Rescue. The results highlighted the effectiveness of the integrated approach:

  • Rapid Onboarding: A non-expert searcher was able to fly the UAV and navigate the interface effectively after only 30 minutes of training.
  • Mission Success: The team successfully located a simulated missing person (a dummy) in under 35 minutes.
  • Enhanced Awareness: The human operator identified the target in the video feed before the specialized video analyst, suggesting that the integrated UI successfully enhanced the operator's situational awareness rather than distracting from it.

Conclusion and Future Directions

This research demonstrates that autonomy in WiSAR is most effective when viewed as a way to organize work rather than simply replace humans. By offloading flight mechanics and data stabilization to the machine, humans can focus on the nuanced task of finding clues. Future work aims to improve "sliding autonomy"—allowing users to modify search paths dynamically as new information emerges—and integrating UAV data directly into the handheld devices used by ground search teams.

Summary updated

BibTeX
@inproceedings{lin2010supporting,
  author = {Lanny Lin and Michael Roscheck and Michael A. Goodrich and Bryan S. Morse},
  title = {{Supporting Wilderness Search and Rescue with Integrated Intelligence: Autonomy and Information at the Right Time and the Right Place}},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year = {2010},
  doi = {10.1609/aaai.v24i1.7573},
}
Figure