Previous Abstract Return to Session D3 Next Abstract

Session D3: Aerial Vehicle Navigation

Using Unmanned Aerial Vehicles (UAVs) in Locating Wandering Patients with Dementia
D. Hanna, A Ferworn, M. Lukaczyn, A. Abhari, Lum, J. Ryerson University, Canada
Location: Spyglass

This paper presents findings from three experiments concerning the use of Unmanned Aerial Vehicles (UAVs) for locating a wandering person whose behavior resembles the behavior of a wandering patient with dementia. The characteristics of the critical form of wandering associated with patients with dementia eloping are discussed. This is one form of Alzheimer disease and its severity varies from one person to another; thus the wandering patterns vary. By using test subjects simulating a lost patient with dementia, along with current Search and Rescue (SAR) operation methods, experiments were performed using drones to find the wandering persons. The data used are based on the analysis from incidents taken from the International Search and Rescue Incident Database (ISRID) which contains thousands of international and national police records on lost persons; both the search and the rescue data were obtained. The data is widely used by first responders in SAR as well as for police training in this field. All experiments were set up with an initial planning point (IPP); this point resembles the position where the person was last seen by an eyewitness or could be the last known point which is identified by strong evidence. The experiments revealed that UAVs, if used with the pre-determined path, could expedite the search process thus enhancing the survivability of the lost person. The first experiment was conducted with police officers familiar with SAR operations acting as observers as part of an ASTM International Standards Committee on Homeland Security Applications; Response Robots Exercise to validate test methods for remotely operated and autonomous ground, aerial, and aquatic systems. In this experiment, three drones were used.
From the three complete tests, two were successful and one failed to find the wandering person. The second trial was done in an open urban park. The test subject was instructed to follow the same directions; however, they were asked to record their route using their mobile device so that a ‘Find Location’ and details could be determined by the path. For the third experiment, four complete tests were administered, three of the tests were successful in locating the person and one failed. As with the other two experiments, a pilot flew the drone in each case, and a person provided directions to the test person acting as the wandering patient.
The paper lists the time needed to detect the person, duration of the complete mission, the differential longitude and latitude analysis from Initial Planning Point (IPP) and the time taken to find the test subject. Challenges and recommendations are presented to inform future experiments. This is the initial step to support the development of a theoretical model or an algorithm to predict the path of a wandering patient using drones. Testing, as well as deeper analysis of available data, will establish a benchmark for the algorithm design. Finally, based on direct communication with personnel from search and rescue teams, more in-depth research in this field is needed to support the SAR teams in finding lost people with dementia safely.
Objectives:
1. Apply existing analyzed data to form an initial algorithm to test UAVs efficiency in locating a wandering person.
2. Conduct field tests using drones on wandering people using test subjects using current SAR statistical models.
3. Develop a methodology/an algorithm to predict the path for wandering people.
4. Test the automation of the UAV search for wandering people based on that model.
5. Analyze existing data using data science methods to determine the similarity in behavior.
Actual Results:
Experiment 1: Three different drones were used; One with very capable technical abilities operated by trained police officers. One with body heat detection via a FLIR camera and a DJI Mavic.
• First search failed to find the person due to a malfunction in the drone’s camera and poor vision on a small mobile device (iPhone).
• Test subject found in less than 7 minutes (Recording to indicate route is available).
• Search using body heat. Wandering person found in 6 minutes.
Experiment 2: Only one test was performed due to restriction in flight zones. An alternative flight detection app was used to detect the wandering person.
• Record path of wanderer using a mobile device.
• Used DJI Phantom 4 Pro drone with the Pix4D manual flight mode app, the test subject was found in 10 min.
• Find location was within an error of approximately 5 meters.
Experiment 3:
• Used DJI Phantom 4 Pro drone
• Apply the same strategy: Fly various paths (4 tests, 1 failed to find the wandering person) based on ISRID data listed in Koester (2008): straight until find object, then ping-pong, travel far and travel alongside a road.
• Calculated differential longitude and latitude from initial planning point, duration to find the test person (minutes), duration of the drone focusing on the test person and duration of the complete mission. Table and graphs are provided.
Conclusions
The use of UAVs, if used with the pre-determined path, can expedite the search process thus enhancing the survivability of the lost person. Knowing the behavioral patterns of lost people with dementia is important in determining the search path. Using UAVs to locate wandering persons can prove much faster and more efficient than standard ground searches. Our findings support the need to develop a model that could be transformed technically to automate the search process using UAVs. Future work will create simulations to test the algorithm(s) for finding people with dementia via UAV search and then develop software that can be used to automate the search process. Furthermore, these simulations could be used to train first responders and SAR teams through game-based learning.



Previous Abstract Return to Session D3 Next Abstract