Presentation Inkonova AB at Berlin Mining Forum 2019.
In the last years, the use of Unmanned Aerial Vehicles (UAVs, also known as “drones”) have found application in different environments that are dangerous or inaccessible by humans like inspection or mapping of underground mining stopes or shafts. During a drone mission it is often required to maintain connectivity with the ground station (referred hereinafter as GS). Even in autonomous flights, real-time communication provides several advantages like active operator supervision and eventual mission correction, in-flight mapping data transfer in case of drone crash inside an inaccessible area and others. In this context, we are interested in using a drone “leader” to explore unknown, dangerous and/or inaccessible underground areas, while keeping constant communication with the GS.
In this paper, we address the problem of using a swarm of autonomous drones, “repeaters”, as a relay chain to maintain communication between a GS and the drone leader responsible for exploration and data acquisition. We propose a sampling-based solution for dynamical positioning of the relay chain. Our method is fully decentralized, scalable and can deal with the case when the trajectory of the main drone is unknown. Simulation results are provided to show the performance of the proposed algorithm.
To simulate the behavior of the relay chain, we use a 2D simulation environment where the trajectory of the leader is predefined but not provided to the repeaters. The model used for the drone’s motion is based on a control signal that is provided as an acceleration and velocity that are bounded, and the drone is modeled as a point in space without orientation (also known as “headless” or “head-free”). In trivial situations, our algorithm can position the relay chain from the current and past mapping data from the leader. Further exploration and analysis of the utility functions to evaluate the sampled positions could drastically improve the performance. A higher level coordination for the whole drone repeaters’ chain could be achieved by using Behavior Trees, which would also increase the robustness and reliability of the whole system.
Sampling-Based Positioning of Unmanned Aerial Vehicles as Communication Relays in Underground Environments
1. Sampling-Based Positioning of Unmanned
Aerial Vehicles as Communication Relays in
Underground Environments
Alex Hermansson*
Gianluigi Silvestri*
Pau Mallol**
* Royal Institute of Technology, KTH ** Inkonova AB
2. Inkonova AB (Stockholm, Sweden)
”3D laser mapping and inspection of underground
inaccessible, hard-to-reach and GPS-deprived, areas”
3. Current solutions for 3D mapping
• Scanner-on-a-stick approach
• Off-the-shelf drones in:
• First Person View (FPV)
• Beyond Visual Line Of Sight (BVLOS)
• Custom or underground drones like:
• TILT Scout (Inkonova)
• TILT Ranger (Inkonova)
• Batonomous (Inkonova)
4. Two examples
”3D laser mapping and inspection of underground
inaccessible, hard-to-reach and GPS-deprived, areas”
Joint mission with
< 5 min flight Courtesy of Clickmox
5. Inspection problems at inaccessible UG areas
• Communication range for inspection vehicles (even if autonomous)
• GPS cannot be used for georeferencing
• GPS cannot be used for autonomous vehicles
• Zones’ variability requires rebustness and adaptability for comms
6. Consequences
• Limited inspection capabilities and reach due to comm
• Single-point failure for hardware and mapped data
• Heavy scanners >
> large drones >
> expensive units
> cannot access through narrow areas
7. Comm solution: chain of repeater drones
• Autonomous deployment
• Repeater drones follow the leader, scanning, drone
• Repeaters optimize their relative position to maximize signal range
and quality
• The repeaters are hierarchic: automatic substitution
• Crash
• Battery depletion
• Other
• Fully decentralised and scalable
8.
9. Current limitations
• We are still at conceptual stage simulating different repeater chain
autonomy approaches
• Current model is in 2D only
10. Conclusions
• Current simulations show potential
• It can be implemented in real hardware and tested in ground
vehicles first
• The vast majority of the pending work is on software
• Repeater hardware and vehicles are fully available
11. Conclusions and future work
• Evaluate better performing utility functions for the sampled positions
• Signal propagation models for constricted spaces
(reflections, constructive/destrutive interferences, signal
absorption/dampening…)
• Behavior trees for ”intelligent” link loss restoration for increased
redundancy and robustness
12. Future work (cont.)
• Multi-agent SLAM mapping
• Several vehicles collaboratively mapping an area
• No need for large, single, mapping drone
• Share data > redundancy and robustness
• Smaller vehicles > access small areas
• No pilot intervention but oversee
• Highly scalable and versatile