Toward Cognitive and Energy-Aware Multi-Agent UAV–UGV (1) (1)

1.

Toward Cognitive and Energy-Aware Multi-Agent
UAV–UGV Collaboration under Limited
Communication
Bauman Moscow State Technical University (BMSTU),
Moscow, Russia

2.

RESEARCH MOTIVATION : HETEROGENEOUS UAV–UGV TEAMS
O perational advantag es
Unmanned Aerial Vehicles (UAVs): rapid aerial observation, mapping, and
wide-area situational awareness.
Unmanned Ground Vehicles (UGVs): long endurance, close-range
inspection, payload transportation, communication support, and mobile
charging.
R epres entative m is s ion
Inspect a bounded hazardous area within a limited time, build or update a
map, and detect damaged, unknown, or mission-relevant objects.
Main challeng es
Communication partitioning, limited UAV energy, dynamic obstacles, and
heterogeneous capabilities.
Interaction between one or more operator and a multi-agent team.
R es earch objective
Develop an integrated coordination approach that preserves mission
feasibility, safety, energy reserve, and operator situational awareness.
1

3.

RELATED WORK IN HETEROGENEOUS UAVUGV COORDINATION
Representative Prior Work:
Cooperative Navigation & Mapping
Chatziparaschis et al. (2020); Jati et al. (2025)
Cooperative mapping and aerial support for ground navigation.
Task Allocation & Multi-Agent Coordination
Nazarova & Ryzhova (2012); Ramezani et al. (2024)
Task allocation and multi-agent coordination.
Energy-Aware Routing & Mobile Charging
Ramasamy et al. (2022); Mondal et al. (2025); Diller et al. (2025)
Energy-aware routing and mobile UAV charging.
Dynamic Safety & Human–Multi-Robot Interaction
Zhou et al. (2022); Chen & Barnes (2014)
Motion prediction, safety, and human–multi-robot interaction.
Remaining Gap
2
Existing studies integrate several of these aspects, but their joint
treatment under temporary network partitioning, energy constraints,
dynamic hazards, task interruption/resumption, and operator
interaction remains insufficiently addressed.

4.

Research Gap and Position Of This Work
W hat is already addressed
Existing studies already integrate several aspects of UAV–UGV
cooperation, such as:
• SLAM (simultaneous localisation & mapping),
• task allocation with communication,
• routing with energy constraints,
• motion prediction with collision avoidance.
Remaining Gap
The joint treatment of:
• temporary network partitioning,
• energy-critical mobile charging,
• dynamic route safety,
• task suspension and resumption,
• human–multi-agent interaction
remains insufficiently addressed in one coordination framework.
Position of This W ork
Proposed contribution: a cross-layer architecture linking
component-local task allocation + priority energy service +
dynamic-risk prediction + route revalidation + operator interaction.
Operator interaction is defined at the architectural level; its
interface and workload are not yet experimentally evaluated.
3

5.

General Problem Formulation
Solution direction
Communication components → feasibility filtering
→ local task allocation → priority energy service →
dynamic-risk assessment and route revalidation
4

6.

General Solution: Cross-Layer Coordination Architecture
5

7.

Mission Scenario Used For Validation
PROBLEM FORMULATION
Scenario
• 1 UAV — aerial survey
• 2 worker UGVs — ground inspection
• 1 mobile charging UGV
• 1 aerial-survey task + 3 ground-inspection tasks
• 2 moving obstacles
• Temporary communication partition
• Critical UAV energy event requiring mobile
charging
• Other robots continue their assigned tasks during
charging
Purpose of the scenario
• To test whether the general coordination
architecture can maintain task execution, safety,
communication-aware allocation, and UAV energy
reserve under simultaneous disturbances.
6

8.

Component-Local Auction For Task Allocation
W hy an auction?
Suitable for dynamic multi-robot task allocation with
heterogeneous agents.
Requires only local bids rather than global optimization.
Can be applied independently inside each connected
communication component.
Gerkey & Matarić (2002); Dias et al. (2006)
Auction-based multi-agent task allocation: Nazarova & Ryzhova (2012)
7

9.

Priority Mobile Charging without Stopping the Team
ENERGY MANAGEMENT & EXECUTION LAYER
8

10.

State Estimation, Dynamic-Risk Prediction and Route Revalidation
Perception, Mapping and Dynamic SAFETY LAYER
9

11.

Paired Numerical Experiment
Experimental Setup
The goal is to coordinate the team to complete all required tasks safely while preserving the UAV energy reserve.
10

12.

Paired Numerical Experiment
RESULTS
11

13.

Conclusions, Limitations and Future Work
INTERPRETATION
Mission Robustness
Energy and Safety
Statistical Comparison
without
UAV reserve preservation increased
Proposed vs. full risk-aware method:
charging, the proposed full method
from 0% to 1 00% due to priority
87% vs. 84% mission success.
increased mission success from 65%
charging,
to 87%, safe success from 0% to 87%,
resumption, dynamic-risk prediction,
and collision-free missions from 67%
and route revalidation.
Compared
with
auction
task
suspension
/
to 87%.
Main contribution: integrated cross-layer coordination linking task allocation, dynamic safety, energy service, communication
constraints, and mission-level human interaction.
Future work: distributed ROS 2 implementation, real UAV–UGV experiments, larger teams, realistic localization/radio/battery
models, and human–multi-agent interaction, including psychophysiological limitations of operators, workload, interface
design, event prioritization, and intervention strategies.
12

14.

Thank you for listening.