Teams of unmanned aerial vehicles (UAVs) deployed for search and monitoring missions frequently operate as partially connected networks, forcing each robot to trade off exploring the environment against relaying information to teammates. This tradeoff is especially acute when robots are semantically heterogeneous: an observation that appears uninformative to the robot that made it may be critical to a teammate with complementary detection capabilities. In this work, we formalize this setting as the heterogeneous mission-aware coverage (HMAC) problem, which couples complete multi-robot coverage of an area with capability-constrained mission-relevant target (MRT) discovery under intermittent communication. We then present DORA, a divergence-oriented data-relay algorithm that drives communication by the value of information to the team rather than by discovery alone. DORA quantifies the mission-relevant divergence between a robot's current information state and its estimate of each teammate's knowledge, capturing mission relevance, discovery novelty, sensor uncertainty, and the age of information. We evaluate DORA in simulation across four environments with differing object densities and spatial structure, and validate it on a physical UAV platform. Our results show that DORA improves MRT resolution delay by up to 74.8% over traditional time-based communication scheduling methods.
Figures & tables
Fig. 1: Finite state machine depicting our decentralized information relaying algorithm.
Fig. 2: The four experimental areas used in simulation.
Environment
Area [km2]
Cars
Buildings
Roads
Camp
0.12
11
31
55
Suburbia
0.17
364
187
131
Farm
0.24
10
22
64
Business Park
0.37
78
20
189
TABLE I: Object inventory of the four simulation environments.
Class
Base
UAV 1
UAV 2
UAV 3
UAV 4
UAV 5
Cars
0.7
0.9
0.3
0.5
0.0
0.2
Road Segments
0.3
0.0
0.0
0.5
0.5
0.8
Buildings
0.0
0.1
0.7
0.0
0.5
0.0
TABLE II: Object class weights αki for each UAV and base.
Fig. 3: Pareto fronts between total mission time versus mean divergence H(Ij,Iji) (top row), and mean weighted Δt (bottom row), for each arena.
Algorithm
Robots
DORA
MA
TB
WC
Camp
2
64.9 ± 68.1
73.0 ± 77.0
99.9 ± 71.5
117.8 ± 103.0
3
29.7 ± 37.6
35.3 ± 39.9
62.1 ± 60.7
43.6 ± 39.9
4
30.0 ± 40.0
34.7 ± 37.5
52.3 ± 60.2
36.8 ± 44.6
5
20.9 ± 25.2
33.4 ± 50.8
31.4 ± 41.6
25.7 ± 30.9
TABLE III: Simulation showing the time required to identify all MRTs, Δˉmrt (mean ± standard deviation), over 100 trials per cell.
Arena
Camp
Suburbia
Farm
Business Park
Hˉ
TB
% of front ↑
14.3%
20.0%
12.5%
0.0%
within 5% ↑
23.1%
46.2%
20.0%
6.7%
WC
21.4%
15.0%
31.2%
27.3%
33.3%
46.2%
35.7%
26.7%
TABLE IV: Pareto front statistics per algorithm and arena for both objective metrics: the percentage of the Pareto front belonging to each algorithm (front share), and the percentage of that algorithm’s runs within 5% of the front.
Fig. 4: GPS trace and decision points for our physical robot experiment. Red markers show the moment when communication is triggered ( H(UAV,base)≥φ ) and blue markers show when the UAV has synchronized its database with the base station. The emulated perception stack “discovers” four cars at time marks: 50, ∼ 95, 132, and ∼ 330 seconds and communicates with the base three times at: 58, 270, and ∼ 575 seconds.
1Dept. of Artificial Intelligence, Sogang University. · 2Dept. of Mechanical Engineering, College of Design and Engineering, National University of Singapore. · 3Dept. of Electronic Engineering, Sogang University.