Searching and Tracking People with Cooperative Mobile Robots

Alex Goldhoorn, Anaís Garrell, René Alquézar and Alberto Sanfeliu

Autonomous Robots

Abstract— Social robots should be able to search and track people in order to help them. In this paper we present two different techniques of coordinated multi-robots for searching and tracking people. A probability map (belief) of a target person location is maintained, and to initialize and update it, two methods were implemented and tested: one based on a reinforcement learning algorithm and the other based on a particle filter. The person is tracked if visible, otherwise an exploration is done by making a balance, for each candidate location, between the belief, the distance, and whether close locations are already being explored by other robots of the team. The validation of the approach was accomplished throughout an extensive set of simulations using up to five agents and a large amount of dynamic obstacles; furthermore, over three hours of real-life experiments with two robots searching and tracking were recorded and analysed.


On this page videos are shown which give an overview of the experiments done with the Multi-agent HB-PF Explorer to find and follow people. Like explained in the article, the robot has to search and follow the person. The person to follow is recognized using AR Markers. More information about the location of the experiments and the map can be found on the map page


In the video the experiment is shown in three sections:
  1. Left: map and probability maps.
  2. Right-top: video focusing on Tibi.
  3. Right-bottom: video focusing on Dabo.
The left image shows a map and a probability map:


Like explained in the article the robots detect the persons in two phases: 1) by laser detection of the legs, and 2) by AR tags. Since the first can only be used to detect people, we added the second to recognize the person. The AR tag should be sufficient, but results in some false positives, therefore we only accept detected tags if there was a person detection by laser close enough. This however still results in some false positive if for example another person is close to the position of a falsely detected tag.
When a false positive detection occurs, the probability for the person being on that location increases, and therefore the robots go and explore that area, but since the false positive normally is detected only for a short time, the probability propagates to other places, and the probability map recovers to the correct area.

The videos

We did different type of experiments: exploration, search, and following with static and dynamic obstacles.
The following table shows the statistics of the different experiment types:
Exploration Search & Track Tracking Total
Distance per robot (km) 1.2 1.2 0.7 3.2
Measured distance person (km) - 0.4 0.5 0.9
Total time (h) 1.1 1.2 0.9 3.2
Avg. visibility (%) 0 16.3 36.4 15.3
Avg. time connected (%) 95.0 79.5 85.8 86.6
Avg. distance person (m) - 8.4 ± 6.4 8.4 ± 5.6 8.3 ± 5.9
Avg. number dynamic obstacles 2.0 ± 1.5 0.6 ± 1.3 3.9 ± 2.8 1.9 ± 2.2
Avg. time found (s) - 106.8 ± 138.7 23.5 ± 42.5 72.9 ± 117.6
Avg. time recovered (s) - 19.6 ± 39.0 12.0 ± 28.3 15.3 ± 33.6

If you cannot see the (complete) video, please right click it and choose the option Copy URL or Copy Location, and then try to view it with an external viewer (e.g. VLC) or another browser.